Without arguing the merits of the Altera investment or divestment, a common pattern for Intel seems to be a wild see-sawing between an aggressive and a defensive market posture - it’s a regular occurrence for Intel to announce a bold new venture to try to claim some new territory, and just as regular that they announce they’re halting that venture in the name of “consolidating” and “focusing on their core.” The consequence is that they never give new ventures time to actually succeed, so they just bleed money creating things they murder in the cradle, and nobody born before last Tuesday is investing in bothering to learn the new Intel thing because its expected lifespan is shorter than the average Google product.
Intel either needs to focus or they need to be bold (and I’d actually prefer they be bold - they’ve started down some cool paths over time), but what they really need is to make up their goddamn minds and stop panicking every other quarter that their “ten-year bets” from last quarter haven’t paid off yet.
Speaking from personal experience, many director-level and above positions at Intel, especially in growth related areas are filled through nepotism and professional connections. I've never seen a headline about Intel’s decline and thought, 'Wow, how could that happen?'
I had a business partner that I agreed on a lot of things with but not about Intel. My assumption was that any small software package from Intel, such as a graph processing toolkit, was trash. He thought they could do no wrong.
Intel really is good at certain kinds of software like compilers or MKL but my belief is that organizations like that have a belief in their "number oneness" that gets in their way of doing anything that it outside what they're good at. Maybe it is the people, processes, organization, values, etc. that gets in the way. Or maybe not having the flexibility to know that what is good at task A is not good at task B.
I saw always intel as a HW company making terribly bad SW. Anywhere I saw intel SW I would run away. Lately I used a big open source library from them, which is standard in the embedded space. Work great, but if you look the code you will be puking for a week.
In my experience Intel's WiFi and Bluetooth drivers on Linux are, by far, the best. They're reliably available on the latest kernel and they actually work. After having used other brands on Linux, I have no intention of getting non-intel WiFi or Bluetooth any time soon. The one time that I found a bug, emailing them about it got me in direct contact with the developers of the driver.
I had a different non-Intel WiFi card before where the driver literally permanently fried all occupied PCIe slots -- they never worked again and the problem happened right after installing the driver. I don't know how a driver such as this causes that but it looks like it did.
Yes, their open source drivers had a painful birth, but they are good once they're sanded and sharpened with the community.
However, they somehow managed to bork e1000e driver in a way that certain older cards sometimes fail to initialize and require a reboot. I have been bitten by the bug, and the problem was fixed later by reverting the problematic patch in Debian.
I don't know current state of the driver since I passed the system on. Besides a couple of bad patches in their VGA drivers, their cards are reliable and works well.
From my experience, their open source driver quality does not depend on the process, but on specific people and their knowledge and love for what they do.
I don't like the aggressive Intel which undercuts everyone by shady tactics, but I don't want them to wither and die, either, but seems like their process, frequency and performance "tricks" are biting them now.
I have found bluez by far the hardest stack to use for Bluetooth Low Energy Peripherals. I have used iOS’s stack, suffered the evolution of the Android stack, used the ACI (ST’s layer), and finally done just straight python to the HCI on pi. Bluez is hands down my least favorite.
that's only because their hardware is extremely simple.
so the driver have little to screw up. but they still manage to! for example, the pci cards are all broken, when it's literary the same hardware as the USB ones.
The affliction he’s imputing is born of absolute dominance over decades. Apple has never had the same level of dominance, and NVidia has only had it for two or three years.
It could possibly come to haunt NVidia or TSMC in decades to come.
A friend who developed a game engine from scratch and is familiar with inner workings and behavior of NVIDIA driver calls it an absolute circus of a driver.
Also, their latest consumer card launches are less then stellar, and the tricks they use to pump up performance numbers are borderline fraud.
As Gamers Nexus puts it "Fake prices for fake frames".
My response is somewhat tangential: When I look at GPUs strictly from the perspective of gaming performance, the last few generations have been so underwhelming. I am not a gamer, but games basically look life-like at this point. What kind of improvements are gamers expecting going forward? Seriously, a mid-level GPU has life-like raytracing at 4K/60HZ. What else do you need for gaming? (Please don't read this as looking down upon gaming; I am only questioning what else gamers need from their GPUs.)
To me, the situation is similar with monitors. After we got the pixel density of 4K at 27 inches with 60Hz refresh rate (enough pixels, enough inches, enough refresh rate), how can it get any better for normies? Ok, maybe we can add HDR, but monitors are mostly finished, similar to mobile phones. Ah, one last one: I guess we can upgrade to OLED when the prices are not so scandalous. Still, for the corporate normies, who account for the lion's share of people siting in front of 1990s-style desktop PCs with a monitor, they are fine with 4K at 27 inches with 60Hz refresh rate forever.
I can't answer the first part, since I'm not playing any modern games, but continuously visit RTS games like C&C & Starcraft series.
However, I can talk about monitors. Yes, a 27" 4K@60 monitor is really, really good, but panel quality (lighting, uniformity and color correctness) goes a long way. After using Dell and HPs "business" monitors for so long, most "normal monitors for normies" look bad to me. Uncomfortable with harsh light and bad uniformity.
So, the monitor quality is not "finished" yet. I don't like OLEDs on big screens, because I tend to use what I buy for a very long time, and I don't wany my screen to age non-uniformly, esp. if I'm looking to it everyday and for long periods of time.
See the funny thing is, even with all of this stuff about Intel that I hear about (and agree with as reported), I also just committed a cardinal sin just recently.
I'm old, i.e. "never buy ATI" is something that I've stuck to since the very early Nvidia days. I.e. switched from Matrox and Voodoo to Nvidia while commiserating and witnessing friend's and colleagues ATI woes for years.
The high end gaming days are long gone, even had a time of laptops where 3D graphics was of no concern whatsoever. I happened to have Intel chips and integrated graphics. Could even start up some gaming I missed out on during the years or replay old favourites just fine as even a business laptop Intel integrated graphics chip was fine for it.
And then I bought an AMD based laptop with integrated Radeon graphics because of all that negative stuff you hear about Intel and AMD itself is fine, sometimes even better, so I thought it was fair to give it a try.
Oh my was that a mistake. AMD Radeon graphics is still the old ATI in full blown problem glory. I guess it's going to be another 25 years until I might make that mistake again.
It's a bummer you've had poor experiences with ATI and later AMD, especially on a new system. I have an AMD laptop with Ryzen 7 7840U which includes a Radeon 780M for integrated graphics and it's been rock solid. I tested many old and new titles on it, albeit at medium-ish settings.
Built a PC with a top-of-the line AMD CPU, it's great. AMD APUs are great in dedicated gaming devices like the XBOX ONE, PS 4 and 5 and Steam Deck.
On the other hand I still think of Intel Integrated GPU in "that thing that screws up your web browser chrome of if you have a laptop with dedicated graphics"
AMD basically stopped supporting (including updating drivers) for GPUs before RDNA (in particular GCN), while such GPUs were still part of AMD's Zen 3 APU offerings.
Well back when, literally 25 years ago, when it was all ATI, there were constant driver issues with ATI. I think it's a pretty well known thing. At least was back when.
I did think that given ATI was bought out by AMD and AMD itself is fine it should be OK. AMD always was. I've had systems with AMD CPUs and Nvidia GPUs back when it was an actual desktop tower gaming system I was building/upgrading myself. Heck my basement server is still an AMD CPU system with zero issues whatsoever. Of course it's got zero graphics duties.
On the laptop side, for a time I'd buy something with discrete Nvidia cards when I was still gaming more actively. But then life happened, so graphics was no longer important and I do keep my systems for a long time / buy non-latest gen. So by chance I've been with Intel for a long time and gaming came up again, casually. The Intel HD graphics were of course totally inadequate for any "real" current gaming. But I found that replaying some old favs and even "newer" games I had missed out on (new as in, playing a 2013 game for the very first time in 2023 type thing) was totally fine on an Intel iGPU.
So when I was getting to newer titles, the Intel HD graphics no longer cut it but I'm still not a "gamer" again, I looked at a more recent system and thought I'd be totally fine trying an AMD system. Exactly like another poster said, "post 2015 should be fine, right?! And then there's all this recent bad news about Intel, this is the time to switch!".
Still iGPU. I'm not going to shell out thousands of dollars here.
And then I get the system and I get into Windows and ... everything just looks way too bright, washed out, hard to look at. I doctored around, installed the latest AMD Adrenalin driver, played around with brightness, contract, HDR, color balance, tried to disable the Vari-Brightness I read was supposed to be the culprit etc. It does get worse once you get into a game. Like you're in Windows and it's bearable. Then you start a game and you might Alt-Tab back to do something and everything is just awfully weirdly bright and it doesn't go away when you shut down the game either.
I stuck with it and kept doctoring for over 6 months now.
I've had enough. I bought a new laptop, two generations behind with an Intel Iris Xe for the same amount of money as the ATI system. I open Windows and ... everything is entirely totally 150% fine, no need to adjust anything. It's comfortable, colors are fine, brightness and contrast are fine. And the performance is entirely adequately the same as with the AMD system. Again, still iGPU and that's fine and expected. It's the quality I'm concerned with, not the performance I'm paying for. I expect to be able to get proper quality software and hardware even if I pay for less performance than gamer kid me back when was willing to.
It's Lenovo. FWIW, one thing I really didn't like much either was that I found out that AMD really tries to hide what actual GPU is in there.
Everything just reports it as "with Radeon graphics", including benchmarking software, so it's almost impossible to find anything about it online.
The only thing I found helped was GPU-Z. Maybe it's just one of the known bad ones and everything else is fine and "I bought the one lemon from a prime steak company" but that doesn't change that my first experience with the lemon company turned prime steak company is ... another lemon ;)
It's a Lucienne C2 apparently. And again, performance wise, absolute exactly as I expected. Graphics quality and AMD software? Unfortunately exactly what I expected from ATI :(
And I'm not alone when I look online and what you find online is not just all Lenovo. So I do doubt it's that. All and I mean all my laptops I'm talking about here were Lenovos. Including when they were called IBM ThinkPads and just built by Lenovo ;)
Laptops have really gone to hell in the past few years. IMO the only sane laptop choices remaining are Framework and Apple. Every other vendor is mess, especially when it comes to properly sleeping when closing the lid.
I bought an AMD Ryzen Thinkpad late last year, and I had the same issue with bright/saturated colours. I fixed it by running X-Rite Color Assistant which was bundled with the laptop, and setting the profile to sRGB. I then turned up the brightness a little.
I think this a consequence of the laptop having HDR colour, and the vendor wanting to make it obvious. It's the blinding blue LED of the current day.
Yeah, I read HDR might be the issue. Didn't know X-Rite and did not come with the laptop, but did play with disabling / trying to adjust HDR, making sure sRGB was set etc. Did not help. Also ran all the calibrations I could find for gamma, brightness and contrast many many times to try and find something that was better.
What I settled on for quite some time was manually adjusted color balance and contrast and turning the brightness down. That made it bearable but especially right next to another system, it's just "off" and still washed out.
If this was HDR and one can't get rid of it, then yeah agreed, it's just bad. I'm actually surprised you'd turn the brightness up. That was one of the worst things to do, to have the brightness too high. Felt like it was burning my eyes.
If the diagnosis is that AMD GPUs can't do HDR properly then yes. There was not a single setting anywhere in Windows itself nor the Adrenalin driver software that allowed me to configure the screen to a comfortable setting. Even when specifically trying to disable anything HDR related.
My work Macbook on the other hand has zero issues with HDR and its display.
To be fair, you can still blame the OEM of course but as a user I have no way to distinguish that, especially in my specific situation.
I think I found X-Rite by just searching for color with the start menu.
Before I used that tool, I tried a few of the built-in colour profiles under the display settings, and that didn't help.
I had to turn the brightness up because when the display is in sRGB it gets dimmer. Everything is much more dim and muted, like a conventional laptop screen. But if I change it back to say, one of the DICOM profiles, then yeah, torch mode. (And if I turn the brightness down in that mode, bright colours are fine but dim colours are too dim and everything is still too saturated).
AMD is appropriately valued IMO, Intel is undervalued and Nvidia is wildly overvalued. We're hitting a wall with LLMs, Nvidia was at one point valued higher than Apple which is insane.
Also CUDA doesn't matter that much, Nvidia was powered by intense AGI FOMO but I think that frenzy is more or less done.
Nvidia is valuably precisely because the software, which is also why AMD is not so valuable. CUDA matters a lot (though that might become less true soon). And Nvidia's CUDA/software forward thinking most certainly predated AGI FOMO and that is the CAUSE of them doing so well with this "AI boom".
It's also not wildly overvalued, purely on a forward PE basis.*
I do wonder about the LLM focus, specifically whether we're designing hardware too much for LLM at the cost of other ML/scientific computing workflows, especially the focus on low precision ops.
But..
1) I don't know how a company like Nvidia could feasibly not focus on designing for LLM in the midst of this craziness and not be sued by shareholders for negligence or something
2) they're able to roll out new architectures with great improvements, especially in memory, on a 2 year cycle! I obviously don't know the counterfactual, but I think without the LLM craze, the hypothetical generation of GPU/compute chips would be behind where they are now.
I think it's possible AMD is undervalued. I've been hoping forever they'd somehow catch up on software. They do very well in server business, and if Intel continues fucking up as much as they have been, AMD will own CPU/servers. I also think what deepseek has done may convince people it's worth it programming closer to the hardware, somewhat weakening Nvidias software moat.
*Of course, it's possible I'm not discounting enough for the geopolitical risk.
> It's also not wildly overvalued, purely on a forward PE basis.*
Once you start approaching a critical mass of sales, it's very difficult to keep growing it. Nvidia is being valued as though they'll reach a trillion dollars worth of sales per year. So nearly 10x growth.
You need to make a lot of assumptions to explain how they'll reach that, versus a ton of risk.
Risk #1: arbitrage principle aka. wherever there's profit to be made other players will move in. AMD has AI chips that are doing quite well, Amazon and Google both have their own AI chips, Apple has their own AI chips... IMO it's far more likely that we'll see commodification of AI chips than that the whole industry will do nothing and pay Nvidia's markup. Especially since TSMC is the one making the chips, not Nvidia.
Risk #2: AI is hitting a wall. VCs claim is isn't so but it's pretty obvious that it is. We went from "AGI in 2025" to AI companies essentially adding traditional AI elements to LLMs to make then useful. LLMs will never reach AGI, we need another technological breakthrough. Companies won't be willing to keep buying every generation of Nvidia chip for ever-diminishing returns.
Risk #3: Geopolitical, as you mentioned. Tariffs, China, etc...
Risk #4: CUDA isn't a moat. It was when no one else had the incentive to create an alternative and it gave everyone on Nvidia a head start. But now everything runs on AMD now too. Google and Amazon have obviously figured out something for their own accelerators.
The only way Nvidia reaches enough revenue to justify their market cap is if Jensen Huang's wild futuristic predictions become reality AND the Googles, Amazons, Apples, AMDs, Qualcomms, Mediateks and every other chip company all fail to catch up.
What I see right now is AI hitting a wall and the commodification of chip production.
I've used Linux exclusively for 15 years so probably why my experience is so positive. Both Intel and AMD are pretty much flawless on Linux, drivers for both are in the kernel nowadays, AMD just wins slightly with their iGPUs.
Yet my AMD APU was never properly supported for hardware video decoding, and could only do up to OpenGL 3.3, while the Windows 10 driver could go up to OpenGL 4.1.
That's actually something I have not tried at all again yet.
Back in the day, w/ AMD CPU and Nvidia GPU, I was gaming on Linux a lot. ATI was basically unusable on Linux while Nvidia (not with the nouveau driver of course), if you looked past the whole kernel driver controversy with GPL hardliners, was excellent quality and performance. It just worked and it performed.
I was playing World of Warcraft back in the mid 2000s via Wine on Linux and the experience was actually better than in Windows. And other titles like say Counter Strike 1.5, 1.6 and Q3 of course.
I have not tried that in a long time. I did hear exactly what you're saying here. Then again I heard the same about AMD buying ATI and things being OK now. My other reply(ies) elaborate on what exactly the experience has been if you're interested.
I wish I had an AMD card. Instead our work laptops are X1 extremes with discrete nvidia cards and they are absolutely infuriating. The external outputs are all routed through the nvidia card, so one frequently ends up with the fan blowing on full blast when plugged into a monitor. Moreover, when unplugging the laptop often fails to shutdown the discrete graphics card so suddenly the battery is empty (because the discrete card uses twice the power). The Intel card on the other hand seems to prevent S3 sleep when on battery, i.e. the laptop starts sleeping and immediately wakes up again (I chased it down to the Intel driver but couldn't get further).
And I'm not even talking about the hassle of the nvidia drivers on Linux (which admittedly has become quite a bit better).
All that just for some negligible graphics power that I'm never using on the laptop.
That’s not specific to Intel though. That’s how Directors and above are recruited in any big company.
For example, Uber hired a VP from Amazon. And the first thing he did was to hire most of his immediate reports at Amazon to Director/Senior Director positions at Uber.
At that level of management work gets done mostly through connections, favors and networking.
Major companies like that become infected with large hierarchies of scum sucking middle management that eat revenue with bonuses.
Of course they are obsessed with shrinking labor costs and resisting all downsizing until it reaches comical levels.
Take a company like health insurance that can't show a large dividend because it would be a public relations disaster. Filled to the gills with vice presidents to suck up extra earnings. Or medical devices.
Software is also very difficult for these hierarchies of overpaid management, because you need to pay labor well to get good software, and the only raison d'etre of these guys is wage suppression.
Leadership is hard for these managers because the primary thing rewarded is middle management machiavellianism, turf wars, and domain building, and any visionary leadership or inspiration is quashed.
It almost fascinates me that large company organizations basically are like Soviet style communism, Even though there are opportunities for internal competition. Like data centers and hosting and it groups. They always need to be centralized for" efficiency".
Meanwhile, they are like 20 data centers and if you had each of them compete for the company's internal business, they'd all run more efficiently.
> it’s a regular occurrence for Intel to announce a bold new venture to try to claim some new territory, and just as regular that they announce they’re halting that venture in the name of “consolidating” and “focusing on their core.” [...] [Intel's new thing's] expected lifespan is shorter than the average Google product.
You got there in the end. You get the same outcome with the same corporate incentive.
Both Intel and Google prioritize {starting something new} over {growing an existing thing}, in terms of corporate promotions and rewards, and therefore employees and leaders self-optimize to produce the repeated behavior you see.
The way to fix this would be to decrease the rewards for starting a new thing and increase the rewards for evolving and growing an existing line of business.
> Both Intel and Google prioritize {starting something new} over {growing an existing thing}, in terms of corporate promotions and rewards, and therefore employees and leaders self-optimize to produce the repeated behavior you see.
I cannot speak for Intel, but Google has done very well by "growing an existing thing" in AdWords and YouTube. Both account for the lion's share of profits. They are absolute revenue giants. Many have tried, and failed to chip away at that lead, but Google has managed to adapt over and over again.
It is the only two things that google has regularly maintained, one of which with one of the biggest moats (youtube, the to go video service), and the other connected to the homepage of the internet.
It's really hard to fuck these things up. Which they have been trying hard, given the state of youtube and the search engine.
It's similar to sales vs dev in software. Sales are always prioritizing new features to attract new users instead of fixing the known issues that are pissing off your current users.
New feature attracts new users and allows for fancy press releases. Nobody cares about press releases about an existing product getting a bug fix are become more stable.
Our society is nothing but "ooh look, shiny!" type of short attention span
But, well, it was a ten-year bet: Altera was acquired in 2015.
If they could not figure how to make it profitable, maybe somebody else should try. (Of course I don't think that the PE company is going to do just that.)
It was a ten-year bet, but they spent the first several years actively sabotaging Altera by trying to move their whole product stack over to non-functional Intel fabs.
Doesn't purchase by a PE company pretty much guarantee the death of it? At least the selling off of the most profitable parts and pieces? Has there ever been a story of a PE purchase and the company grew under the new owner?
PE’s buy companies to increase the company’s value then sell it. There’s been many successes. Powerschool, Hilton, Dunkin’ Brands, Dollar General, Beats by Dre, Petco, GoDaddy, BJ’s Wholesale Club, Neiman Marcus, Panera Bread, Allegro, Guitar Center, Nielsen, McAfee…
I wouldn't call that a roaring success. Funnily enough, Intel played a major role in running McAfee into the ground.
With proper leadership, McAfee could've ended up in the position CrowdStrike is now.
Trying not to piss off the Chinese government, and in particular its intelligence services (in order to sell chips) is unfortunately not a good model for an antimalware business.
> Additionally, managers may prefer mergers because empirical evidence suggests that the size of a company and the compensation of managers are correlated.
Yeah, that's where my mind went. Executive and upper management salaries seem to be a function of revenue, not profit.
A lot of compensation works that way, to be honest. First order is that you get a percentage of whatever river of money you sit close to, regardless of effort or skill.
Esp when you are talking about software. Revenue means you have a customer that is locked up. Once you are ready to get profit, reduce costs/jack up prices and profit comes rolling in.
If I hold stock in a company, then my company acquires that company, the stock rises, and I liquidate my position in it after 6 months or whatever the cool-down period is, is this considered insider trading?
If you hold stock in company A, and your current company B acquires company A, that's not insider trading if you already owned the stock in company A before you had any information that company B was going to make that decision.
It is, however, a conflict of interest for you to be involved in company B's acquisition of company A (e.g. influencing company B to buy company A), and might even rise to the level of a breach of your fiduciary duty to company B.
I know a woman who was part of a M&A team. On her first day, she was told her days of owning individual stocks in the industry were over. She could only purchase aggregate funds. Although, I do wonder if the same rules apply to the VPs who actually have to sign off on the deals.
Insider trading is all about information held by "insiders", not about who owns what. So it would depend on whether you know something material and nonpublic when you liquidate your position (e.g., you know the acquisition is going terribly and the acquiring company is going to write it off).
This seems to be common for corporate America in general. I used to work at a YC startup. We kiiiiiinda maaaaaaaybe ran out of money (not my department) and happened to get bought by a large investor that also happens to be a US-based hardware manufacturer. Two years and countless reorgs later, they laid everyone off and as far as I know, are no longer in the business of selling the software products they bought. They never figured out how software worked, never had anyone managining the division for more than 6 months, and got bored. I think they thought by moving everyone over to Microsoft Word and Windows laptops (peppered with a half-hearted threat about RTO), they would just magically make billions of dollars the first month. It didn't happen.
I am beginning to think M&A are just some sort of ego thing for bored megacorp execs, rather than serious attempts to add efficiency and value to the marketplace. (Prove me wrong, bored megacorp execs. I'll wait.)
Having been through a few acquisitions myself, I think there is a perverse incentive where buying and destroying any competition (real or imagined) leads to positive enough outcomes that it doesn't matter if the underlying asset is destroyed. Nobody would come out and say that, but when an acquisition is tossed aside there may not be enough repercussions to prevent it from happening again.
Intel bought a drone company that was producing the only drone that was good enough for my real estate inspection company to use. They acquired it and then killed it a year or two after. The inspection industry didn't have a proper drone for years after that until DJI started getting serious about it and produced the M30E.
It was just senseless, Intel doesn't have real or imagined competition from a drone company, it wasn't even close to being in the same market. They just believed the hype about drones being the next big thing and when they found out they were too early they decided they didn't have the patience to wait for drones to become a thing and they killed it. There was no long term vision behind it.
The high-end Falcon models were an engineering marvel and, as you say, nothing else in the market was even close.
I don't know about "real estate inspection", but another use case was for them to be used in oil rigs in the North Sea to inspect the structure of the rig itself. They had to be self-stabilizing under high winds and adverse weather conditions, and they had to carry a good enough camera to take detailed photos.
Unfortunately, while the technology was there, the market wasn't. Not many wanted to get a $35K drone to be able to sustain this business.
Exactly, but surely they weren't actually losing money? Why not keep the business afloat selling expensive drones to specialty companies until the broader market picked up as they envisioned it would when they bought the company. I think we paid more like $25k for our Falcons, though buying them wasn't on my side of the company. We would have gladly pay $35k for a next gen Falcon if they ever made one. Now DJI makes good enough drones for less than $10k, but there's a chance we still would've went with Intel just we could tell our customers we weren't flying Chinese drones.
In real estate inspection, we had the same sort of concerns, can't fly too close to the object for safety reasons, and we need high resolution photos to determine quality of the masonry, paintwork and roofing etc..
The company is still thriving, you can check out their website to find out more about what real estate inspection is about (in The Netherlands): https://www.aeroscan.nl/
I believe it could be the weight of the camera and lens you would desire for good looking photos (think Sony a7 size). Good looking photos sell houses.
EDIT I just noticed the “inspection” part. Maybe they wanted good zoom to spy on the tenants? (Or maybe that’s a really uncharitable take).
To me, high quality photos for real estate inspection means (e.g.) being able to take high resolution photos of a specific part of the roof so you can understand why there’s a leak. Not having to climb is a big deal!
This is one of the main reasons we added anti monopoly provisions to our laws more than 100 years ago. Market dominance is a recognized factor in allowing this inversion of rewards to occur.
That's the face of it. Labor is a market as well. The impacts of these arrangements on our labor pool is extraordinary. It's a massive displaced cost of allowing these types of mergers to occur born out by the people who stand to gain the least from the merging of business assets.
My personal theory is that desktop / laptop / server x86 (usually) is such a giant money printer that a) Intel can invest in anything (Altera, antivirus, Optane...) but b) when they do, they quickly realise that this isn't a giant profit margin machine like x86, so why bother?
They fuck their customers when they do that. A good friend of mine had a product designed around Quark that was about to go into production when Intel pulled the rug out from under him.
I worked for a former Fortune 300 company that had an active internal investment strategy. They wanted the next billion dollar business, guaranteed, in 12 months. And wanted to invest more than 1 million dollars. Sadly they are now bankrupt and owned by PE.
Rest in Peace Altera I guess? I still drink out of my color changing Altera mug (that's long stopped changing color) most days. PE ruins everything so it's only a matter of time before they're gutted and sold for scraps by the vultures at Silver Lake. (though honestly the writing was on the wall since the Intel acquisition I had held onto some hope) If only we had a functioning government interested in actually maintaining our technological dominance and enforcing/expanding antitrust legislation. I wrote my first Verilog on an Altera chip and I'll remember them fondly.
Tariffs done well could be a boon for a sector. But they are tricky to do well, and the current administration doesn't show the slightest hint of being capable of doing it well.
You have to treat tariffs not as a moat to protect an industry for good, but a runway to give a nascent industry enough room to take off. In a mature industry, tariffs are more likely to keep incumbents uncompetitive and disincentivize investments that would make them more competitive, especially if those are capital-heavy.
Also, tariffs aren't going to be effective if other structural issues exist in an industry that prevent or sharply limit expansion. Like key components having a sole worldwide supplier with a full order book. Or if capital investment to set up a new factory are beyond the ability of the financial markets to provide.
I don't think people are opposed to tariffs, at least they weren't before. Bernie Sanders has historically been for tariffs when used properly. Used properly being the important phrase. When you have someone who doesn't understand what a trade deficit is imposing tariffs based on the difference between deficit and surplus, you pretty effectively turn people against tariffs on top of the whole destroying the global economy thing.
Congress delegated tariff power to the president after Smoot-Hawley caused such a disaster.
> The reason the President doesn't have this power is because the economy should not rest on the whims or understanding of any one person.
It's interesting to see that it just takes time for lessons to be un-learned. The reason Smoot-Hawley was such a disaster is that it took hundreds of people to agree that it was good policy in the house, which meant adding tariffs to the bill in favor of the districts they individually represented. The result was an egregiously long list of things being tariffed. They delegated it to the one person specifically because they weren't similarly beholden to so many conflicting pressures.
I don't mean any of this to defend Trump's actions, in fact the opposite: he's essentially managing to do the same thing even without politic pressures to do so. I just mean to say that it is reasonably sane for congress to have delegated tariffs in a limited capacity when this flaw was revealed.
In this case it's not Trump but Peter Navarro [0] who doesn't understand how tariffs work, because he's apparently never looked into multiparty game theory.
Exhibit A: Navarro being sidelined and Scott Bessent put in charge of running tariff negotiations, after the bond markets spooked.
No it's Trump. If it wasn't made clear to you during his first term that he doesn't understand tariffs or trade deficits, he spent most of 2023 and 2024 campaigning and showing you he doesn't understand it.
Trump doesn't understand most things he does, hence why his advisors have so much power over outcomes, and policies shift as different advisors fall in and out of favor.
I mean there are several issues at play here. He is being sued for the illegal tariffs since Republicans are spineless and are cool with him just doing anything so I'm focusing on the practical problems.
As someone who's worked at Xilinx before and after the merger, it's a surprise they were even able to sell it for that much. Altera has been noncompetitive to Xilinx in performance and to Lattice in terms of low-end/low-power offerings for at least the last 2 generations.
I'm concerned about the future of FPGAs and wonder who will lead the way to fix these abhorrent toolchains these FPGA companies force upon developers.
Agree on both.
As things like the PIO on the rp line of micros gets more common, micros will have IO that can match FPGAs. For low end, micros are generally good enough or gain NPU compute cores. It’s the IO that differentiates FPGAs.
So Intel found optimists who think they can make Altera more competitive? It's a success. Success with Intel products would be better, and excellence at M&A is hard to convert into excellence at chipmaking, but it's better than nothing.
I hear this a lot, but in my experience this isn't true at all.
A Versal AI Edge FPGA has a theoretical performance of 0.7TFLOPs just from the DSPs alone, while consuming less power than a Raspberry Pi 5 and this is ignoring the AI Engines, which are exactly the ASICs that you are talking about. They are more power efficient than GPUs, because they don't need to pretend to run multiple threads each with their own register files or hide memory latency by swapping warps. Their 2D NOC plus cascaded connections allow them to have a really high internal memory bandwidth in-between the tiles at low power.
What they are missing is processing in memory, specifically LPDDR-PIM for GEMV acceleration. The memory controllers simply can't deliver a memory bandwidth that is competitive with what Nvidia has and I'm talking about boards like Jetson Orin here.
They're competitive on perf/watt because they're designed to do one thing. But they're much more expensive than an ASIC, which, if also designed to do one thing would be better than the FPGA.
FPGAs are neither here nor there and will always be niche. If you need the same thing many times you make dedicated silicon. If you need many different things available all at once you use a normal CPU. Only if ASICs are too expensive and CPUs are too slow the FPGA can shine.
There is literally no market for FPGA as coprocessor/accelerator and there never was (that was some kind of pipe/hype dream before GPGPU took off). Where there is a market for them (prototyping ASICs, automotive, whatever, network switches, etc) there is no replacement but there is also no growth.
Depends entirely how you define "growth". If you take AI, LLM as your baseline of growth, then yeah, sure. But what else is growing?
FPGAs are getting cheaper with each gen, expanding into low cost, high volume markets that were unthinkable for an FPGA 10 years ago. Lattice has an FPGA family specifically targeted to smartphones, and I've been consulting for a high end audio company that wanted to do some dsp, and a cheap FPGA was the best option in the market for the particular implementation that they wanted to do.
It's not sexy growth, but it's growth. Otherwise, we wouldn't had the explosion of the latest years in low end FPGA companies.
lookup sigmastudio dsp, dsp is insanely cheap todo, there is absolutely no need for fpga, what that guy was doing was either nonsense or it was in 1995. which are both irrelevant points, or rather you provided examples that show fpga are irrelevant, no growth market.
(how many audio devices were using TMS320 dsps even before and after ipod was a thing...)
My point is that FPGAs have become very cheap, competing with microcontrollers. I would agree that high end audio manufacturers are about as rational as they costumers.
If FPGAs are not a growing market, how come we have gone from 2 companies (we'll ignore niche space stuff) to ~10 in the last 20 years? Not many IC fields where there is a growth in manufacturers instead of consolidation...
Yeah I personally wondered if AMD was just copying Intel, because apparently every CPU manufacturer also needs to manufacture FPGAs, or they actually have a long term strategy where it is essential for both the FPGA and CPU departments to cooperate.
I think Xilinx did a fine job with their AI Engines and AMD decided to integrate a machine learning focused variant on their laptops as a result. The design of the intel NPU is nowhere near as good as AMD's. I have to say that AMD is not a software company though and while the hardware is interesting, their software support is nonexistent.
Also, if you're worried about FPGAs that doesn't really make much sense, since Effinix is killing it.
I briefly hoped that, like the integration of GPUs, there would be a broader integration of programmable logic in general purpose CPUs, with AMD integrating Xilinx fabric and Intel integrating Altera fabric. But I could never imagine a real use case and apparently there wasn't a marketable enough one either. Something like high-level synthesis ending up like CUDA always seemed like it would present a neat development environment for certain optimizations.
For those keeping score at home, 51% sold at a total valuation of $8.75B, which means they are bringing in around $4.5B, and recognizing a loss of roughly 50% on what was their biggest deal ever when it took place in 2015.
Sure, it was down 60%. But the real question is whether it outperformed Intel as a whole, and outperformed other internal investments Intel could make. I certainly wouldn't think that a 2015 dollar anywhere else within Intel is worth more than 40¢ today, given how they've been running.
Or they got what they wanted from it and are selling off the rest, like when Google bought Motorola Wireless for the Patents then sold off the non-googly employees, culture, and brand for cheap.
I was excited, too. I was also excited when Intel announced Larrabee.
That was before I learnt about the many and varied ways in which Intel sabotages itself, and released that Intel's underperformance has little to do with a lack of good technical ideas or talent.
I.e. I was young and naive. I am now considerably less young, and at least a little less naive.
There was some hope at the time that FPGAs could be used in a lot more applications in the data center. It is likely still feasible. Remember Hennessy published:
And maybe this is/was a pipe dream - maybe there aren't enough people with the skills to have a "golden age of architecture". But MSFT was deploying FPGAs in the data center and there were certainly hopes and dreams this would become a big thing.
FPGA dev is just much more painful and more expensive than software dev at every step.
That's in no small part because the industry & tools seem to be stuck decades in the past. They never had their "GCC moment". But there's also inherent complexity in working at a very low level, having to pay attention to all sorts of details all the time that can't easily be abstracted away.
There's the added constraint that FPGA code is also not portable without a lot of extra effort. You have to pick some specific FPGA you want to target, and it can be highly non-trivial to port it to a different one.
And if you do go through all that trouble, you find out that running your code on a cloud FPGAs turns out to be pretty damn expensive.
So in terms of perf per dollar invested, adding SIMD to your hot loop, or using a GPU as an accelerator may have a lower ceiling, but it's much much more bang for the buck and involves a whole lot less pain along the way.
It's hard to find places where FPGAs really win. For relatively simple tasks FPGA can beat just about anything in latency. For instance for the serialization/deserialization end of a high frequency trading system. If a problem has a large working set and needs to store data in DRAM it needs a memory controller the same way a CPU or GPU has a memory controller and this can only be efficient if the system's memory access pattern is predictable.
You can certainly pencil out FPGA or ASIC systems that which would attain high levels of efficient parallelism if there wasn't memory bandwidth or latency limits but there are. If you want to do math that GPUs are good at, you use GPUs. Historically some FPGAs have let you allocate bits in smaller slices so if you only need 6 bit math you can have 6 bit math but GPUs are muscling in on that for AI applications.
FPGAs really are good at bitwise operations used in cryptography. They beat CPUs at code cracking and bitcoin mining but in turn they get beat by ASICs. However there is some number of units (say N=10,000) where the economics of the ASIC plus the higher performance will drive you to ASIC -- for Bitcoin mining or for the NSA's codebreaking cluster. You might prototype this system on an FPGA before you get masks made for an ASIC though.
For something like the F-35 where you have N=1000 or so, could care less about costs, and might need to reconfigure it for tomorrow's threats, the FPGA looks good.
One strange low N case is that of display controllers for retrocomputers. Like it or not a display controller has one heck of a parts count to make out of discrete parts and ASIC display controllers were key to the third generation of home computers which were made with N=100,000 or so. Things like
and are already expensive compared to the Raspberry Pi so they tend to use either a microcontroller or FPGA, the microcontroller tends to win because an ESP32 which costs a few buck is, amazingly, fast enough to drive a A/D converter at VGA rates or push enough bits for HDMI!
Rapid product development. Got a project that needs to ship in 6-9 months and will be on the market for less than two years in small volume? Thats where FPGAs go. Medical, test and measurement, military, video effects, telepresence, etc.
I'm not sure about that. In these fields there are plenty of places where you need to ingest or process masses of data (eg. from a sensor in a medical device), and you're only going to sell 5 of these machines a month for 100K each, so $3000+ bill of materials for an FPGA to solve the problem makes sense.
The problem (for Intel) is that you don't sell billions of dollars of FPGAs into a mass market this way.
Most of my knowledge about FPGAs come from ex-FPGA people, so take this with a grain of salt:
First off, clock rates on an FPGA run at about a tenth that of CPUs, which means you need a 10× parallelism speedup just to break-even, which can be a pretty tall order, even for a lot of embarrassingly parallel problems.
(This one is probably a little bit garbled) My understanding is that the design of FPGAs is such that they're intrinsically worse at getting you FLOP/memory bandwidth number than other designs, which also gives you a cap on expected perf boosts.
The programming model is also famously bad. FPGAs are notorious for taking forever to compile--and the end result of waiting half an hour or more might simply be "oops, your kernel is too large." Also, to a degree, a lot of the benefits of FPGA are in being able to, say, do a 4-bit computation instead of having to waste the logic on a full 8-bits, which means your code also needs to be tailored quite heavily for an FPGA, which makes it less accessible for most programmers.
Tooling mostly. To write fast code for CPUs you need a good optimizing compiler, like clang or gcc. Imagine how much work has gone into making them good. We're talking thousands of man years over several decades. You need just as good tooling for FPGAs and it takes just as much effort to produce. Except the market is orders of magnitudes smaller. You can also not "get help" from the open source community since high-end FPGAs are way to expensive for most hackers.
Intel tried to get around this problem by having a common framework. So one compiler (based on clang) with multiple backends for their CPUs, FPGAs, and GPUs. But in practice it doesn't work. The architectures are too different.
There is nothing quite like gcc or LLVM for FPGAs yet. FPGA tooling is still stuck in the world of proprietary compilers and closed software stacks. It makes the whole segment of the industry move slower and have higher friction. This is just starting to break with Yosys and related tools, which are showing wild advantages in efficiency over some of the proprietary tooling, but still only support a fraction of available chips, mostly the smaller ones.
I'm just a casual observer, but I'm pretty sure one hard thing about FPGAs is preventing abuse. A customer could easily set up a ring oscillator that burns out all the LUTs. Another thing is FPGAs are about 10x slower than dedicated logic, so CPUs/GPUs beat them for a lot of applications. Plus, there's not a lot of logic designers in the first place. Software skills don't transfer over very well. For example, a multiplier is about the size of 8kb of RAM, so lookups and complex flow are way more expensive than just multiplying a value again (kinda like GPUs, except if you only had an L1 cache without main memory).
Not sure why I'm being downvoted, would those who downvoted me explain why? I try to be accurate so if I missed any important details I'd like to know :)
I dont know how a Ring Osc specifically could burn out LUTs. All switching contributes to the device temperature. If they get too hot then they will enter thermal protection mode.
We run such oscillators as dummy payloads for thermal tests while we are waiting for the real firmware to be written.
To play devil's advocate, I wonder how well they handle more annoying things.
When a CMOS switches, it essentially creates a very brief short circuit between VCC and GND. That's part of normal dynamic power consumption, it's expected and entirely accounted for.
But I don't know how these cloud FPGAs could enforce that you don't violate setup and hold times all over the place. When you screw up your crossings and accidentally have a little bit of metastability, that CMOS will switch back and forth a little bit, burn some power, and settle one way or the other.
Now if you intentionally go out of your way to keep one cell metastable as long as possible while the neighbors are cold, that's going to be one hell of a localized hotspot. I wouldn't be surprised if thermal protection can't kick in fast enough.
It's just kibitzing though, I'm not particularly inclined to try with my own hardware
It made their stock pop for a while which was all that mattered to Brian Krzanich who took the bonus and left the mess in the hands of Bob Swan who did the same things and left the mess ... (recursion her).
It was a forced acquisition, iirc they made promises to Altera to get them to use their foundry, failed to keep those promises and could either get sued and embarrassed or just buy Altera outright for about what they were worth before the deal.
> Intel soon discovered the obvious, which is that customers with applications well-suited to FPGAs already use FPGAs.
So selling FPGA's was a bad move? Or was the purchase price just wildly out-of-line with the--checking...$9.8B annual market that's expected to rise to $23.3B by 2030?
And less disastrous for Xilinx, given they could basically just keep going as they were before, instead of being significantly diverted onto a sinking ship of a process.
> Intel soon discovered the obvious, which is that customers with applications well-suited to FPGAs already use FPGAs.
Yes, but pairing an FPGA somewhat tightly integrated with an actually powerful x86 CPU would have made an interesting alternative to the usual FPGA+some low end ARM combo that's common these days.
Sure, if they wanted to intel could have done what nvidia did with CUDA: Put the tech into everything, even their lowest end consumer devices, and sink hundreds of millions into tooling and developer education given away free of charge.
And maybe it would have lead somewhere. Perhaps. But they didn't.
I wasn't there, but I've always imagined the conversation went something like this:
Intel: Welcome, Altera. We'd like you to integrate your FPGA fabric onto our CPUs.
Altera: Sure thing, boss! Loads of our FPGAs get plugged into PCIe slots, or have hard or soft CPU cores, so we know what we're doing.
Intel: Great! Oh, by the way, we'll need the ability to run multiple FPGA 'programs' independently, at the same time.
Altera: Ummmm
Intel: The programs might belong to different users, they'll need an impenetrable security barrier between them. It needs full OS integration, so multi-user systems can let different users FPGA at the same time. Windows and Linux, naturally. And virtual machine support too, otherwise how will cloud vendors be able to use it?
Altera: Uh
Intel: We'll need run-time scaling, so large chips get fully utilised, but smaller chips still work. And it'll need to be dynamic, so a user can go from using the whole chip for one program to sharing it between two.
Intel: And of course indefinite backwards compatibility, that's the x86 promise. Don't do anything you can't support for at least 20 years.
Intel: Your toolchain must support protecting licensed IP blocks, but also be 100% acceptable to the open source community.
Intel: Also your current toolchain kinda sucks. It needs to be much easier to use. And stop charging for it.
Intel: You'll need a college outreach program. And a Coursera course. Of course students might not have our hardware, so we'll need a cloud offering of some sort, so they can actually do the exercises in the course.
Altera: I guess to start with we
Intel: Are you profitable yet? Why aren't you contributing to our bottom line?
I think they have tried to improve the software for FPGAs--FPGA backends are part of their oneAPI software stack, for example. And when I was in grad school, Intel was definitely doing courses on building for FPGAs using OpenCL (I remember seeing some of their materials, but I don't know much about them otehr than they existed).
As to why it didn't work, well, I'm not plugged into this space to have a high degree of certainty, but my best guess is "FPGAs just aren't that useful for that many things."
Yes, if they actually made the thing available, maybe people would have used it for something. There were several proofs of concept at the time, with some serious gains, ever for the uses that people ended up using CUDA.
But they didn't actually sell it. At least not in any form anybody could buy. So, yeah, we get the OP claiming it was an obvious technological dead-end.
And if they included it on lower-end chips (the ones they sold just a few years after they brought Altera), we could have basically what the RasPI 2040 is nowadays. Just a decade earlier and controlled by them... On a second thought, maybe this was for the best.
Applications that benefit from the Zynq-style combination (e.g. radio systems) generally take that approach because they have SWaP concerns that preclude the use of big x86 CPUs in the first place.
Size, Weight, and Power. It would be very nice if people would take the two seconds to type those words out instead of using ungoogleable acronyms on a public forum unfamilar with the terms.
GPGPUs ended up becoming the AI/cloud accelerators that FPGAs promised to be back when Intel bought Altera.
FPGAs are not ideal for raw parallel number crunching like in AI/LLMs. They are more appropriate for predictable real-time/ultra-low-latency parallel things like the the modulation and demodulation of signals in 5G base state stations.
FPGAs might not be ideal, but AMD's NPU IP originated with Xilinx.
Intel was an early player to so many massive industries (e.g. XScale, GPGPU, hybrid FPGA SoCs). Intel abandoned all of them prematurely and has been left playing catch-up every time. We might be having a very different discussion if literally any of them had succeeded.
XScale was forced on Intel as a penalty for anticompetitive activities against Digital. It’s no surprise then that they weren’t interested in doing anything with it.
If only someone could have come up with a plausibly profitable use-case for advanced FPGA's and highly performant, efficient, real-time processing or hardware acceleration in those intervening years? What are ya gonna do?
Quick search shows Altera held 30% of the FPGA market. That puts AMD’s $50B acquisition of Xilinx (which holds ~50% of the market) in an awkward light. Using some extremely crude math, Xilinx’s fair market value might now be closer to ~$15B.
Did AMD massively overpay, or has the FPGA market fundamentally shifted? Curious to see how this new benchmark ripples into AMD’s stock valuation.
The FPGA market shifted. For a brief moment they were allowed to be on BOMs of end user devices due to the rest of the computing field lagging behind somewhat. That period, as far as I can tell, is over.
My anecdotal example would be high end broadcast audio processors. These do quite a bit beyond the actual processing of audio, in particular, into baseband or even RF signal generation.
In any case these devices used to be fully analog, then when they first went digital were a combination of DSPs for processing and FPGAs for signal output. Later generations dropped the DSP and did everything in larger FPGAs as the larger FPGAs became available. Later generations dropped the whole stack and just run on an 8 core Intel processor using real time linux and some specialized real time signal processing software with custom designed signal generators.
The high core and high frequency CPUs became good enough and getting custom made chips became exceptionally cheap as well. FPGAs became rather pointless in this pipline.
The US military, for a time, had a next generation radio specification that specifically called for the use of FPGAs, as that would allow them to make manufacturer agnostic radios and custom software for them. That never panned out but it shows the peak use of FPGAs to manage the constraints of this time period.
When Intel acquired Altera, Altera’s market share was at 36% and Xilinx at 51%. Today Xilinx remains at ~50% while Altera’s share has dropped to 29%. Altera has lost share to Microchip and Lattice.
I’ve said it before, Intel is where technology companies go to die. Fortunately while Altera is probably a mess of useless Intel drone MBAs, there’s a decent core that can be salvaged. Best of luck to them.
Apparently, the FPGA industry wasn't large enough for two major players. Maintaining an extremely specialized developer ecosystem for a relatively small niche can't have been cheap. Almost zero cross-over too, since FPGA tooling is much too foreign to be repurposed for other architectures. I suspect this move will make it a bit harder for Intel to collect "developer mindshare" for their other hyped up stuff because no one likes having the rug pulled out from under them. Hope AMD can make a better job with Xilinx than what Intel could with Altera.
Intel FPGA venture made tons more sense than AMD following it. FPGAs are great at filling up your idle fabs and honing engineering skills on reaching high yields.
Selling now also makes sense. There was only one serious competitor in 2015. Now you got Tariffs both ways to the main place where everything is build, and said place has own homegrown vendors like GOWIN, Sipeed, Efinix. But the biggest reason is amount of stuff designed in the West/Taiwan is falling with China taking over actual product design.
>In 2015, China released its “Made in China 2025” (MIC 2025) strategy, which refined some of these targets, setting a goal of achieving 40 percent self-sufficiency in semiconductors by 2020 and 70 percent by 2025.
>In 2024, the majority of MIC 2025's goals were considered to be achieved, despite U.S. efforts to curb the program.
Products coming out of China no longer use STM microcontrollers, Vishay/Analog mosfets/diodes and Altera/Xilinx FPGAs. Its all Chinese semiconductor brands you never heard about. Good example is this teardown of Deye SUN-5K-SG04LP1 5kW hybrid solar inverter https://www.youtube.com/watch?v=n0_cTg36A2Q
Intel was looking to sell Altera for over a year before Trump's Tariff Tourettes. And I bet it wasn't the hardware that was the problem, it was the software. No matter how amazing your FPGA hardware is, it is useless if you can't also produce high-quality software for operating it. For CPUs you can just tell people to use gcc or clang, not so with FPGAs.
I used to work at Intel (around 1999) in their Jones Farm campus in Oregon. My employee stock grants from that time are still underwater.
This was the heyday at Intel. I left within a year because I noticed that the talent that was respected, compensated and influential at Intel was the sales engineers. I can't pretend to have known that would lead to the decline of the company, but I knew that as an engineer uninterested in sales, that it wasn't the place for me.
I'm not surprised by this divestiture - Investor confidence in Alterra has been low ever since the disastrous loss of their brand new capital ship Aurora on its very first mission.
What a waste! I can never understand corporate thinking and how CEOs get such massive fucking pay for decisions like this.
Intel paid $16.7 billion in 2015 and sold it for $8.75 billion?! What about all the money dumped into Altera from 2015 to 2025? How much was that? Is Intel just handing over the FPGA market to AMD?
Beyond ensuring adequate cash flow, they need to be 100% focused on getting 18A shipping in volume as soon as possible rather than financial engineering stuff.
My guess would be no, but I could be wrong. The current administration clearly wants more domestic capability, so even if someone like TSMC/Samsung/etc wanted to acquire as part of their US operations, my gut says it would be challenged.
When AMD spun off their fabs into what became Global Foundries, it was difficult for many to see the upside. However, today, it seems not being tied to any particular fab/tech is one of AMD's biggest advantages.
That whole acquisition always was a head-scratcher to me. OK, you pay nearly 17B for the thing, and then do absolutely nothing with it. How does that work? Even during its best years 17B was big money for Intel.
I left the hardware business in 1992 but seems that nothing has changed. People still think FPGAs are really cool and nobody cab figure out how to make money selling them.
Seems quite cheap. If I was a state I'd buy it. Possibly give stake to the suitable university and then create internships and other learning opportunities. I would also subsidise products to SMEs and then invest more to ensure company can supply defence and other industries, decoupling the country from dependence on other countries from crucial tech.
I think nationalization is usually frowned on in the west, but your comment about universities got me wondering. It seems small enough that the state could donate it to a consortium of research universities. That'd have to be better than PE in terms of serving the national interest, wouldn't it?
The business geniuses of intel bought Altera for nearly $17B in 2015. Now sold control at valuation barely half of that. After official inflation of over 30% during that time. Which means it lost over 2/3 of its value, (take into account also the lost interest on that money). Given that they gave up control for half the money, it’s effectively as if it was relinquished for 1/3 of what Intel paid for it.
So far - no one of the responsible Intel execs paid any price for such atrocious loss of stakeholders’ value. They need to be in jail and lose their personal wealth to repay the stockholders.
The SEC should investigate them, see whether there was any inside trading to benefit from this horrible value loss.
This criminal lack of performance needs to be brought up during the upcoming shareholders meeting. Responsible must pay the price.
Would you hire again the Intel CEOs, head of Intel Capital, any members of Intel’s board of directors after such abysmal performance?
Intel either needs to focus or they need to be bold (and I’d actually prefer they be bold - they’ve started down some cool paths over time), but what they really need is to make up their goddamn minds and stop panicking every other quarter that their “ten-year bets” from last quarter haven’t paid off yet.
Intel really is good at certain kinds of software like compilers or MKL but my belief is that organizations like that have a belief in their "number oneness" that gets in their way of doing anything that it outside what they're good at. Maybe it is the people, processes, organization, values, etc. that gets in the way. Or maybe not having the flexibility to know that what is good at task A is not good at task B.
I had a different non-Intel WiFi card before where the driver literally permanently fried all occupied PCIe slots -- they never worked again and the problem happened right after installing the driver. I don't know how a driver such as this causes that but it looks like it did.
However, they somehow managed to bork e1000e driver in a way that certain older cards sometimes fail to initialize and require a reboot. I have been bitten by the bug, and the problem was fixed later by reverting the problematic patch in Debian.
I don't know current state of the driver since I passed the system on. Besides a couple of bad patches in their VGA drivers, their cards are reliable and works well.
From my experience, their open source driver quality does not depend on the process, but on specific people and their knowledge and love for what they do.
I don't like the aggressive Intel which undercuts everyone by shady tactics, but I don't want them to wither and die, either, but seems like their process, frequency and performance "tricks" are biting them now.
I have found bluez by far the hardest stack to use for Bluetooth Low Energy Peripherals. I have used iOS’s stack, suffered the evolution of the Android stack, used the ACI (ST’s layer), and finally done just straight python to the HCI on pi. Bluez is hands down my least favorite.
so the driver have little to screw up. but they still manage to! for example, the pci cards are all broken, when it's literary the same hardware as the USB ones.
Frequent releases, GitHub repo with good enough user interaction, examples, bug fixing and feedback.
It could possibly come to haunt NVidia or TSMC in decades to come.
Also, their latest consumer card launches are less then stellar, and the tricks they use to pump up performance numbers are borderline fraud.
As Gamers Nexus puts it "Fake prices for fake frames".
To me, the situation is similar with monitors. After we got the pixel density of 4K at 27 inches with 60Hz refresh rate (enough pixels, enough inches, enough refresh rate), how can it get any better for normies? Ok, maybe we can add HDR, but monitors are mostly finished, similar to mobile phones. Ah, one last one: I guess we can upgrade to OLED when the prices are not so scandalous. Still, for the corporate normies, who account for the lion's share of people siting in front of 1990s-style desktop PCs with a monitor, they are fine with 4K at 27 inches with 60Hz refresh rate forever.
However, I can talk about monitors. Yes, a 27" 4K@60 monitor is really, really good, but panel quality (lighting, uniformity and color correctness) goes a long way. After using Dell and HPs "business" monitors for so long, most "normal monitors for normies" look bad to me. Uncomfortable with harsh light and bad uniformity.
So, the monitor quality is not "finished" yet. I don't like OLEDs on big screens, because I tend to use what I buy for a very long time, and I don't wany my screen to age non-uniformly, esp. if I'm looking to it everyday and for long periods of time.
I'm old, i.e. "never buy ATI" is something that I've stuck to since the very early Nvidia days. I.e. switched from Matrox and Voodoo to Nvidia while commiserating and witnessing friend's and colleagues ATI woes for years.
The high end gaming days are long gone, even had a time of laptops where 3D graphics was of no concern whatsoever. I happened to have Intel chips and integrated graphics. Could even start up some gaming I missed out on during the years or replay old favourites just fine as even a business laptop Intel integrated graphics chip was fine for it.
And then I bought an AMD based laptop with integrated Radeon graphics because of all that negative stuff you hear about Intel and AMD itself is fine, sometimes even better, so I thought it was fair to give it a try.
Oh my was that a mistake. AMD Radeon graphics is still the old ATI in full blown problem glory. I guess it's going to be another 25 years until I might make that mistake again.
What kind of problems did you see on your laptop?
On the other hand I still think of Intel Integrated GPU in "that thing that screws up your web browser chrome of if you have a laptop with dedicated graphics"
AMD basically stopped supporting (including updating drivers) for GPUs before RDNA (in particular GCN), while such GPUs were still part of AMD's Zen 3 APU offerings.
I did think that given ATI was bought out by AMD and AMD itself is fine it should be OK. AMD always was. I've had systems with AMD CPUs and Nvidia GPUs back when it was an actual desktop tower gaming system I was building/upgrading myself. Heck my basement server is still an AMD CPU system with zero issues whatsoever. Of course it's got zero graphics duties.
On the laptop side, for a time I'd buy something with discrete Nvidia cards when I was still gaming more actively. But then life happened, so graphics was no longer important and I do keep my systems for a long time / buy non-latest gen. So by chance I've been with Intel for a long time and gaming came up again, casually. The Intel HD graphics were of course totally inadequate for any "real" current gaming. But I found that replaying some old favs and even "newer" games I had missed out on (new as in, playing a 2013 game for the very first time in 2023 type thing) was totally fine on an Intel iGPU.
So when I was getting to newer titles, the Intel HD graphics no longer cut it but I'm still not a "gamer" again, I looked at a more recent system and thought I'd be totally fine trying an AMD system. Exactly like another poster said, "post 2015 should be fine, right?! And then there's all this recent bad news about Intel, this is the time to switch!".
Still iGPU. I'm not going to shell out thousands of dollars here.
And then I get the system and I get into Windows and ... everything just looks way too bright, washed out, hard to look at. I doctored around, installed the latest AMD Adrenalin driver, played around with brightness, contract, HDR, color balance, tried to disable the Vari-Brightness I read was supposed to be the culprit etc. It does get worse once you get into a game. Like you're in Windows and it's bearable. Then you start a game and you might Alt-Tab back to do something and everything is just awfully weirdly bright and it doesn't go away when you shut down the game either.
I stuck with it and kept doctoring for over 6 months now.
I've had enough. I bought a new laptop, two generations behind with an Intel Iris Xe for the same amount of money as the ATI system. I open Windows and ... everything is entirely totally 150% fine, no need to adjust anything. It's comfortable, colors are fine, brightness and contrast are fine. And the performance is entirely adequately the same as with the AMD system. Again, still iGPU and that's fine and expected. It's the quality I'm concerned with, not the performance I'm paying for. I expect to be able to get proper quality software and hardware even if I pay for less performance than gamer kid me back when was willing to.
I've seen OEMs do that to an Intel+NVIDIA laptop, too. Whatever you imagine AMD's software incompetence to be, PC OEMs are worse.
Everything just reports it as "with Radeon graphics", including benchmarking software, so it's almost impossible to find anything about it online.
The only thing I found helped was GPU-Z. Maybe it's just one of the known bad ones and everything else is fine and "I bought the one lemon from a prime steak company" but that doesn't change that my first experience with the lemon company turned prime steak company is ... another lemon ;)
It's a Lucienne C2 apparently. And again, performance wise, absolute exactly as I expected. Graphics quality and AMD software? Unfortunately exactly what I expected from ATI :(
And I'm not alone when I look online and what you find online is not just all Lenovo. So I do doubt it's that. All and I mean all my laptops I'm talking about here were Lenovos. Including when they were called IBM ThinkPads and just built by Lenovo ;)
I think this a consequence of the laptop having HDR colour, and the vendor wanting to make it obvious. It's the blinding blue LED of the current day.
What I settled on for quite some time was manually adjusted color balance and contrast and turning the brightness down. That made it bearable but especially right next to another system, it's just "off" and still washed out.
If this was HDR and one can't get rid of it, then yeah agreed, it's just bad. I'm actually surprised you'd turn the brightness up. That was one of the worst things to do, to have the brightness too high. Felt like it was burning my eyes.
You don't like current AMD systems because one of them had an HDR screen? Nothing to do with CPU/GPU/APU?
My work Macbook on the other hand has zero issues with HDR and its display.
To be fair, you can still blame the OEM of course but as a user I have no way to distinguish that, especially in my specific situation.
Before I used that tool, I tried a few of the built-in colour profiles under the display settings, and that didn't help.
I had to turn the brightness up because when the display is in sRGB it gets dimmer. Everything is much more dim and muted, like a conventional laptop screen. But if I change it back to say, one of the DICOM profiles, then yeah, torch mode. (And if I turn the brightness down in that mode, bright colours are fine but dim colours are too dim and everything is still too saturated).
Also CUDA doesn't matter that much, Nvidia was powered by intense AGI FOMO but I think that frenzy is more or less done.
Nvidia is valuably precisely because the software, which is also why AMD is not so valuable. CUDA matters a lot (though that might become less true soon). And Nvidia's CUDA/software forward thinking most certainly predated AGI FOMO and that is the CAUSE of them doing so well with this "AI boom".
It's also not wildly overvalued, purely on a forward PE basis.*
I do wonder about the LLM focus, specifically whether we're designing hardware too much for LLM at the cost of other ML/scientific computing workflows, especially the focus on low precision ops.
But.. 1) I don't know how a company like Nvidia could feasibly not focus on designing for LLM in the midst of this craziness and not be sued by shareholders for negligence or something 2) they're able to roll out new architectures with great improvements, especially in memory, on a 2 year cycle! I obviously don't know the counterfactual, but I think without the LLM craze, the hypothetical generation of GPU/compute chips would be behind where they are now.
I think it's possible AMD is undervalued. I've been hoping forever they'd somehow catch up on software. They do very well in server business, and if Intel continues fucking up as much as they have been, AMD will own CPU/servers. I also think what deepseek has done may convince people it's worth it programming closer to the hardware, somewhat weakening Nvidias software moat.
*Of course, it's possible I'm not discounting enough for the geopolitical risk.
Once you start approaching a critical mass of sales, it's very difficult to keep growing it. Nvidia is being valued as though they'll reach a trillion dollars worth of sales per year. So nearly 10x growth.
You need to make a lot of assumptions to explain how they'll reach that, versus a ton of risk.
Risk #1: arbitrage principle aka. wherever there's profit to be made other players will move in. AMD has AI chips that are doing quite well, Amazon and Google both have their own AI chips, Apple has their own AI chips... IMO it's far more likely that we'll see commodification of AI chips than that the whole industry will do nothing and pay Nvidia's markup. Especially since TSMC is the one making the chips, not Nvidia.
Risk #2: AI is hitting a wall. VCs claim is isn't so but it's pretty obvious that it is. We went from "AGI in 2025" to AI companies essentially adding traditional AI elements to LLMs to make then useful. LLMs will never reach AGI, we need another technological breakthrough. Companies won't be willing to keep buying every generation of Nvidia chip for ever-diminishing returns.
Risk #3: Geopolitical, as you mentioned. Tariffs, China, etc...
Risk #4: CUDA isn't a moat. It was when no one else had the incentive to create an alternative and it gave everyone on Nvidia a head start. But now everything runs on AMD now too. Google and Amazon have obviously figured out something for their own accelerators.
The only way Nvidia reaches enough revenue to justify their market cap is if Jensen Huang's wild futuristic predictions become reality AND the Googles, Amazons, Apples, AMDs, Qualcomms, Mediateks and every other chip company all fail to catch up.
What I see right now is AI hitting a wall and the commodification of chip production.
I've used Linux exclusively for 15 years so probably why my experience is so positive. Both Intel and AMD are pretty much flawless on Linux, drivers for both are in the kernel nowadays, AMD just wins slightly with their iGPUs.
I had a Ryzen 2700u that was fully supported, latest OpenGL and Vulkan from day 1, hardware decoding, etc... but on Linux.
Back in the day, w/ AMD CPU and Nvidia GPU, I was gaming on Linux a lot. ATI was basically unusable on Linux while Nvidia (not with the nouveau driver of course), if you looked past the whole kernel driver controversy with GPL hardliners, was excellent quality and performance. It just worked and it performed.
I was playing World of Warcraft back in the mid 2000s via Wine on Linux and the experience was actually better than in Windows. And other titles like say Counter Strike 1.5, 1.6 and Q3 of course.
I have not tried that in a long time. I did hear exactly what you're saying here. Then again I heard the same about AMD buying ATI and things being OK now. My other reply(ies) elaborate on what exactly the experience has been if you're interested.
And I'm not even talking about the hassle of the nvidia drivers on Linux (which admittedly has become quite a bit better).
All that just for some negligible graphics power that I'm never using on the laptop.
For example, Uber hired a VP from Amazon. And the first thing he did was to hire most of his immediate reports at Amazon to Director/Senior Director positions at Uber.
At that level of management work gets done mostly through connections, favors and networking.
Of course they are obsessed with shrinking labor costs and resisting all downsizing until it reaches comical levels.
Take a company like health insurance that can't show a large dividend because it would be a public relations disaster. Filled to the gills with vice presidents to suck up extra earnings. Or medical devices.
Software is also very difficult for these hierarchies of overpaid management, because you need to pay labor well to get good software, and the only raison d'etre of these guys is wage suppression.
Leadership is hard for these managers because the primary thing rewarded is middle management machiavellianism, turf wars, and domain building, and any visionary leadership or inspiration is quashed.
It almost fascinates me that large company organizations basically are like Soviet style communism, Even though there are opportunities for internal competition. Like data centers and hosting and it groups. They always need to be centralized for" efficiency".
Meanwhile, they are like 20 data centers and if you had each of them compete for the company's internal business, they'd all run more efficiently.
You got there in the end. You get the same outcome with the same corporate incentive.
Both Intel and Google prioritize {starting something new} over {growing an existing thing}, in terms of corporate promotions and rewards, and therefore employees and leaders self-optimize to produce the repeated behavior you see.
The way to fix this would be to decrease the rewards for starting a new thing and increase the rewards for evolving and growing an existing line of business.
It's really hard to fuck these things up. Which they have been trying hard, given the state of youtube and the search engine.
I can see why you have to be "special" to work at these places.
New feature attracts new users and allows for fancy press releases. Nobody cares about press releases about an existing product getting a bug fix are become more stable.
Our society is nothing but "ooh look, shiny!" type of short attention span
If they could not figure how to make it profitable, maybe somebody else should try. (Of course I don't think that the PE company is going to do just that.)
I wouldn't call that a roaring success. Funnily enough, Intel played a major role in running McAfee into the ground.
With proper leadership, McAfee could've ended up in the position CrowdStrike is now.
Trying not to piss off the Chinese government, and in particular its intelligence services (in order to sell chips) is unfortunately not a good model for an antimalware business.
Bonuses by juicing revenue numbers
Bigger next job by doing M&A and having really good-looking resume and interview story.
https://corporatefinanceinstitute.com/resources/valuation/mo...
Yeah, that's where my mind went. Executive and upper management salaries seem to be a function of revenue, not profit.
It is, however, a conflict of interest for you to be involved in company B's acquisition of company A (e.g. influencing company B to buy company A), and might even rise to the level of a breach of your fiduciary duty to company B.
I am beginning to think M&A are just some sort of ego thing for bored megacorp execs, rather than serious attempts to add efficiency and value to the marketplace. (Prove me wrong, bored megacorp execs. I'll wait.)
It was just senseless, Intel doesn't have real or imagined competition from a drone company, it wasn't even close to being in the same market. They just believed the hype about drones being the next big thing and when they found out they were too early they decided they didn't have the patience to wait for drones to become a thing and they killed it. There was no long term vision behind it.
I don't know about "real estate inspection", but another use case was for them to be used in oil rigs in the North Sea to inspect the structure of the rig itself. They had to be self-stabilizing under high winds and adverse weather conditions, and they had to carry a good enough camera to take detailed photos.
Unfortunately, while the technology was there, the market wasn't. Not many wanted to get a $35K drone to be able to sustain this business.
In real estate inspection, we had the same sort of concerns, can't fly too close to the object for safety reasons, and we need high resolution photos to determine quality of the masonry, paintwork and roofing etc..
EDIT I just noticed the “inspection” part. Maybe they wanted good zoom to spy on the tenants? (Or maybe that’s a really uncharitable take).
That's the face of it. Labor is a market as well. The impacts of these arrangements on our labor pool is extraordinary. It's a massive displaced cost of allowing these types of mergers to occur born out by the people who stand to gain the least from the merging of business assets.
It seems like a low risk effort to put a promising inexperienced exec in charge of a recent acquisition.
If they're a screw up and run it into the ground, imagine how much damage they could have done in a megacorp position of power.
Megacorp saved (at the cost of a smaller company)
Stock down again? Sell the company you bought 2 years ago.
From the top to the bottom the problem with late stage capitalism is misaligned incentives.
Edit: I wrote "the problem" and I should have written "among the many, many problems"
tick, tock
I don't think PE is responsible for that one.
https://datasheet.octopart.com/CUBIC-CYCLONIUM-Altera-datash...
You have to treat tariffs not as a moat to protect an industry for good, but a runway to give a nascent industry enough room to take off. In a mature industry, tariffs are more likely to keep incumbents uncompetitive and disincentivize investments that would make them more competitive, especially if those are capital-heavy.
Also, tariffs aren't going to be effective if other structural issues exist in an industry that prevent or sharply limit expansion. Like key components having a sole worldwide supplier with a full order book. Or if capital investment to set up a new factory are beyond the ability of the financial markets to provide.
The President does not have the power to create tariffs. Congress does!
The reason the President doesn't have this power is because the economy should not rest on the whims or understanding of any one person.
> The reason the President doesn't have this power is because the economy should not rest on the whims or understanding of any one person.
It's interesting to see that it just takes time for lessons to be un-learned. The reason Smoot-Hawley was such a disaster is that it took hundreds of people to agree that it was good policy in the house, which meant adding tariffs to the bill in favor of the districts they individually represented. The result was an egregiously long list of things being tariffed. They delegated it to the one person specifically because they weren't similarly beholden to so many conflicting pressures.
I don't mean any of this to defend Trump's actions, in fact the opposite: he's essentially managing to do the same thing even without politic pressures to do so. I just mean to say that it is reasonably sane for congress to have delegated tariffs in a limited capacity when this flaw was revealed.
Exhibit A: Navarro being sidelined and Scott Bessent put in charge of running tariff negotiations, after the bond markets spooked.
[0] https://en.m.wikipedia.org/wiki/Peter_Navarro
I'm concerned about the future of FPGAs and wonder who will lead the way to fix these abhorrent toolchains these FPGA companies force upon developers.
https://www.achronix.com/blog/accelerating-llm-inferencing-f...
There's been neural processing chips since before LLM craze [1].
[1]: https://en.wikipedia.org/wiki/Neural_processing_unit#History
A Versal AI Edge FPGA has a theoretical performance of 0.7TFLOPs just from the DSPs alone, while consuming less power than a Raspberry Pi 5 and this is ignoring the AI Engines, which are exactly the ASICs that you are talking about. They are more power efficient than GPUs, because they don't need to pretend to run multiple threads each with their own register files or hide memory latency by swapping warps. Their 2D NOC plus cascaded connections allow them to have a really high internal memory bandwidth in-between the tiles at low power.
What they are missing is processing in memory, specifically LPDDR-PIM for GEMV acceleration. The memory controllers simply can't deliver a memory bandwidth that is competitive with what Nvidia has and I'm talking about boards like Jetson Orin here.
Honestly I've asked different hardware researchers this question and they all seem to give different answers.
[1] The FFT Strikes Back: An Efficient Alternative to Self-Attention (168 comments):
https://news.ycombinator.com/item?id=43182325
Some FPGA vendors are contributing to and relying, partially or completely, on the open source stack (mainly yosys+nextpnr).
It is still perceived as not being "as good" as the universally hated proprietary tools, but it's getting there.
0. https://ir.quicklogic.com/press-releases/detail/657/quicklog...
1. https://www.designnews.com/semiconductors-chips/is-platypus-...
FPGAs are getting cheaper with each gen, expanding into low cost, high volume markets that were unthinkable for an FPGA 10 years ago. Lattice has an FPGA family specifically targeted to smartphones, and I've been consulting for a high end audio company that wanted to do some dsp, and a cheap FPGA was the best option in the market for the particular implementation that they wanted to do.
It's not sexy growth, but it's growth. Otherwise, we wouldn't had the explosion of the latest years in low end FPGA companies.
lookup sigmastudio dsp, dsp is insanely cheap todo, there is absolutely no need for fpga, what that guy was doing was either nonsense or it was in 1995. which are both irrelevant points, or rather you provided examples that show fpga are irrelevant, no growth market.
(how many audio devices were using TMS320 dsps even before and after ipod was a thing...)
If FPGAs are not a growing market, how come we have gone from 2 companies (we'll ignore niche space stuff) to ~10 in the last 20 years? Not many IC fields where there is a growth in manufacturers instead of consolidation...
I think Xilinx did a fine job with their AI Engines and AMD decided to integrate a machine learning focused variant on their laptops as a result. The design of the intel NPU is nowhere near as good as AMD's. I have to say that AMD is not a software company though and while the hardware is interesting, their software support is nonexistent.
Also, if you're worried about FPGAs that doesn't really make much sense, since Effinix is killing it.
https://www.synflow.com/
ETA they also paid out almost $10 in dividends.
Ouch - your work is so good we will pay 10x what it is worth, because we are not good enough to do it.
But you are not good enough for us. Maybe they couldn't a binary tree.
That was before I learnt about the many and varied ways in which Intel sabotages itself, and released that Intel's underperformance has little to do with a lack of good technical ideas or talent.
I.e. I was young and naive. I am now considerably less young, and at least a little less naive.
Intel soon discovered the obvious, which is that customers with applications well-suited to FPGAs already use FPGAs.
https://www.doc.ic.ac.uk/~wl/teachlocal/arch/papers/cacm19go...
And maybe this is/was a pipe dream - maybe there aren't enough people with the skills to have a "golden age of architecture". But MSFT was deploying FPGAs in the data center and there were certainly hopes and dreams this would become a big thing.
That's in no small part because the industry & tools seem to be stuck decades in the past. They never had their "GCC moment". But there's also inherent complexity in working at a very low level, having to pay attention to all sorts of details all the time that can't easily be abstracted away.
There's the added constraint that FPGA code is also not portable without a lot of extra effort. You have to pick some specific FPGA you want to target, and it can be highly non-trivial to port it to a different one.
And if you do go through all that trouble, you find out that running your code on a cloud FPGAs turns out to be pretty damn expensive.
So in terms of perf per dollar invested, adding SIMD to your hot loop, or using a GPU as an accelerator may have a lower ceiling, but it's much much more bang for the buck and involves a whole lot less pain along the way.
You can certainly pencil out FPGA or ASIC systems that which would attain high levels of efficient parallelism if there wasn't memory bandwidth or latency limits but there are. If you want to do math that GPUs are good at, you use GPUs. Historically some FPGAs have let you allocate bits in smaller slices so if you only need 6 bit math you can have 6 bit math but GPUs are muscling in on that for AI applications.
FPGAs really are good at bitwise operations used in cryptography. They beat CPUs at code cracking and bitcoin mining but in turn they get beat by ASICs. However there is some number of units (say N=10,000) where the economics of the ASIC plus the higher performance will drive you to ASIC -- for Bitcoin mining or for the NSA's codebreaking cluster. You might prototype this system on an FPGA before you get masks made for an ASIC though.
For something like the F-35 where you have N=1000 or so, could care less about costs, and might need to reconfigure it for tomorrow's threats, the FPGA looks good.
One strange low N case is that of display controllers for retrocomputers. Like it or not a display controller has one heck of a parts count to make out of discrete parts and ASIC display controllers were key to the third generation of home computers which were made with N=100,000 or so. Things like
https://www.commanderx16.com/
and are already expensive compared to the Raspberry Pi so they tend to use either a microcontroller or FPGA, the microcontroller tends to win because an ESP32 which costs a few buck is, amazingly, fast enough to drive a A/D converter at VGA rates or push enough bits for HDMI!
Rapid product development. Got a project that needs to ship in 6-9 months and will be on the market for less than two years in small volume? Thats where FPGAs go. Medical, test and measurement, military, video effects, telepresence, etc.
The problem (for Intel) is that you don't sell billions of dollars of FPGAs into a mass market this way.
First off, clock rates on an FPGA run at about a tenth that of CPUs, which means you need a 10× parallelism speedup just to break-even, which can be a pretty tall order, even for a lot of embarrassingly parallel problems.
(This one is probably a little bit garbled) My understanding is that the design of FPGAs is such that they're intrinsically worse at getting you FLOP/memory bandwidth number than other designs, which also gives you a cap on expected perf boosts.
The programming model is also famously bad. FPGAs are notorious for taking forever to compile--and the end result of waiting half an hour or more might simply be "oops, your kernel is too large." Also, to a degree, a lot of the benefits of FPGA are in being able to, say, do a 4-bit computation instead of having to waste the logic on a full 8-bits, which means your code also needs to be tailored quite heavily for an FPGA, which makes it less accessible for most programmers.
Intel tried to get around this problem by having a common framework. So one compiler (based on clang) with multiple backends for their CPUs, FPGAs, and GPUs. But in practice it doesn't work. The architectures are too different.
We run such oscillators as dummy payloads for thermal tests while we are waiting for the real firmware to be written.
When a CMOS switches, it essentially creates a very brief short circuit between VCC and GND. That's part of normal dynamic power consumption, it's expected and entirely accounted for.
But I don't know how these cloud FPGAs could enforce that you don't violate setup and hold times all over the place. When you screw up your crossings and accidentally have a little bit of metastability, that CMOS will switch back and forth a little bit, burn some power, and settle one way or the other.
Now if you intentionally go out of your way to keep one cell metastable as long as possible while the neighbors are cold, that's going to be one hell of a localized hotspot. I wouldn't be surprised if thermal protection can't kick in fast enough.
It's just kibitzing though, I'm not particularly inclined to try with my own hardware
So selling FPGA's was a bad move? Or was the purchase price just wildly out-of-line with the--checking...$9.8B annual market that's expected to rise to $23.3B by 2030?
Yes, but pairing an FPGA somewhat tightly integrated with an actually powerful x86 CPU would have made an interesting alternative to the usual FPGA+some low end ARM combo that's common these days.
And maybe it would have lead somewhere. Perhaps. But they didn't.
Intel: Welcome, Altera. We'd like you to integrate your FPGA fabric onto our CPUs.
Altera: Sure thing, boss! Loads of our FPGAs get plugged into PCIe slots, or have hard or soft CPU cores, so we know what we're doing.
Intel: Great! Oh, by the way, we'll need the ability to run multiple FPGA 'programs' independently, at the same time.
Altera: Ummmm
Intel: The programs might belong to different users, they'll need an impenetrable security barrier between them. It needs full OS integration, so multi-user systems can let different users FPGA at the same time. Windows and Linux, naturally. And virtual machine support too, otherwise how will cloud vendors be able to use it?
Altera: Uh
Intel: We'll need run-time scaling, so large chips get fully utilised, but smaller chips still work. And it'll need to be dynamic, so a user can go from using the whole chip for one program to sharing it between two.
Intel: And of course indefinite backwards compatibility, that's the x86 promise. Don't do anything you can't support for at least 20 years.
Intel: Your toolchain must support protecting licensed IP blocks, but also be 100% acceptable to the open source community.
Intel: Also your current toolchain kinda sucks. It needs to be much easier to use. And stop charging for it.
Intel: You'll need a college outreach program. And a Coursera course. Of course students might not have our hardware, so we'll need a cloud offering of some sort, so they can actually do the exercises in the course.
Altera: I guess to start with we
Intel: Are you profitable yet? Why aren't you contributing to our bottom line?
As to why it didn't work, well, I'm not plugged into this space to have a high degree of certainty, but my best guess is "FPGAs just aren't that useful for that many things."
But they didn't actually sell it. At least not in any form anybody could buy. So, yeah, we get the OP claiming it was an obvious technological dead-end.
And if they included it on lower-end chips (the ones they sold just a few years after they brought Altera), we could have basically what the RasPI 2040 is nowadays. Just a decade earlier and controlled by them... On a second thought, maybe this was for the best.
FPGAs are not ideal for raw parallel number crunching like in AI/LLMs. They are more appropriate for predictable real-time/ultra-low-latency parallel things like the the modulation and demodulation of signals in 5G base state stations.
Intel was an early player to so many massive industries (e.g. XScale, GPGPU, hybrid FPGA SoCs). Intel abandoned all of them prematurely and has been left playing catch-up every time. We might be having a very different discussion if literally any of them had succeeded.
Did AMD massively overpay, or has the FPGA market fundamentally shifted? Curious to see how this new benchmark ripples into AMD’s stock valuation.
My anecdotal example would be high end broadcast audio processors. These do quite a bit beyond the actual processing of audio, in particular, into baseband or even RF signal generation.
In any case these devices used to be fully analog, then when they first went digital were a combination of DSPs for processing and FPGAs for signal output. Later generations dropped the DSP and did everything in larger FPGAs as the larger FPGAs became available. Later generations dropped the whole stack and just run on an 8 core Intel processor using real time linux and some specialized real time signal processing software with custom designed signal generators.
The high core and high frequency CPUs became good enough and getting custom made chips became exceptionally cheap as well. FPGAs became rather pointless in this pipline.
The US military, for a time, had a next generation radio specification that specifically called for the use of FPGAs, as that would allow them to make manufacturer agnostic radios and custom software for them. That never panned out but it shows the peak use of FPGAs to manage the constraints of this time period.
I’ve said it before, Intel is where technology companies go to die. Fortunately while Altera is probably a mess of useless Intel drone MBAs, there’s a decent core that can be salvaged. Best of luck to them.
Selling now also makes sense. There was only one serious competitor in 2015. Now you got Tariffs both ways to the main place where everything is build, and said place has own homegrown vendors like GOWIN, Sipeed, Efinix. But the biggest reason is amount of stuff designed in the West/Taiwan is falling with China taking over actual product design.
https://itif.org/publications/2024/08/19/how-innovative-is-c...
>In 2015, China released its “Made in China 2025” (MIC 2025) strategy, which refined some of these targets, setting a goal of achieving 40 percent self-sufficiency in semiconductors by 2020 and 70 percent by 2025.
https://en.wikipedia.org/wiki/Made_in_China_2025
>In 2024, the majority of MIC 2025's goals were considered to be achieved, despite U.S. efforts to curb the program.
Products coming out of China no longer use STM microcontrollers, Vishay/Analog mosfets/diodes and Altera/Xilinx FPGAs. Its all Chinese semiconductor brands you never heard about. Good example is this teardown of Deye SUN-5K-SG04LP1 5kW hybrid solar inverter https://www.youtube.com/watch?v=n0_cTg36A2Q
This was the heyday at Intel. I left within a year because I noticed that the talent that was respected, compensated and influential at Intel was the sales engineers. I can't pretend to have known that would lead to the decline of the company, but I knew that as an engineer uninterested in sales, that it wasn't the place for me.
Should have stuffed it under the bed instead…
Intel paid $16.7 billion in 2015 and sold it for $8.75 billion?! What about all the money dumped into Altera from 2015 to 2025? How much was that? Is Intel just handing over the FPGA market to AMD?
https://download.intel.com/newsroom/2021/archive/2015-12-28-...
Will we see an AMD-esque fab spin-off?
When AMD spun off their fabs into what became Global Foundries, it was difficult for many to see the upside. However, today, it seems not being tied to any particular fab/tech is one of AMD's biggest advantages.
>Intel flogs off majority stake in Altera to private equity for $4B
>Buy high, sell low: FPGA biz cost x86 giant $16B decade ago
for those not up on this stuff
Selling out to PE is a signal this company is about to get gutted and loaded to the tits with debt and management fees from PE.
https://subnautica.fandom.com/wiki/Alterra_Corporation
I mean it's a pipe dream, but why not.
The SEC should investigate them, see whether there was any inside trading to benefit from this horrible value loss.
This criminal lack of performance needs to be brought up during the upcoming shareholders meeting. Responsible must pay the price.
Would you hire again the Intel CEOs, head of Intel Capital, any members of Intel’s board of directors after such abysmal performance?