News for hackers. Stuff that matters.
Re:Why Gelsinger was wrong for Intel by roenxi on Sunday December 08, 2024 @09:25PM (#42362559)
This isn't a space I pretend to understand in detail, but Cantrill's objections to Larrabee here seem confusing. With hindsight, Intel abandoning Larrabee has been a disaster and they should have sustained the effort to develop a great GPGPU chip. Without taking any position at all on Gelsinger's execution his strategic vision seems to have be supported by what happened in practice after his attempt.

If Intel had been practising making discrete GPU from 2009 to 2024 instead of giving up then it does seem quite reasonable that Nvidia would be a quarter of its size. One of the obvious mis-steps Intel made over the last 20 years was failing to keep trying Larrabee-style chips until they figured out how to make them work. Ok version 1 might be a disaster and version 2 might be bad. But we're looking back over almost 20 years! There was a lot of room to figure out how to be good and tap in to the huge new market that was opening up.

Re:Why Gelsinger was wrong for Intel by ekeane on Sunday December 08, 2024 @09:38PM (#42362618)
Intel has a habit of coming up with great ideas, then deciding that they aren't immediately successful or immediately 60+% gross margins, so it clearly isn't worth following up. Intel's R&D department is a grave yard of great products murdered at the alter of Gross Margin.

Larrabee is one such product. It wasn't immediately great, and didn't immediately promise great returns, and didn't quite have a market yet, so it was killed. I can't imagine where Intel would be in the AI world if they had the foresight to stick with Larrabee.

Re:Why Gelsinger was wrong for Intel by brookst on Sunday December 08, 2024 @10:58PM (#42362966)
Yep. Intel has no patience. They have high margin businesses throwing off lots of cash, so the finance-driven culture says “why waste money on risky new things that can’t possibly equal that level of profitability in the short term?”

Somehow they forgot that all of their highly profitably businesses were risky bets at one time.

Re:Why Gelsinger was wrong for Intel by josephg on Monday December 09, 2024 @02:32AM (#42363787)
Right. Google has the same problem. Any time they create a new product, it’s a coin flip on whether or not it’ll still be around in a year or two. I’ve steered multiple companies away from Google’s offerings because I just don’t think you should trust that any of their newer products will still be around in 5-10 years.

Good on them for sticking with their promises for firebase. But it’s a pity about stadia. Don’t bet your business on a flip of the Google coin.

Re:Why Gelsinger was wrong for Intel by bjornsing on Monday December 09, 2024 @06:08AM (#42365018)
> Somehow they forgot that all of their highly profitably businesses were risky bets at one time.

And that all of their highly profitable businesses are risky bets long term.

Re:Why Gelsinger was wrong for Intel by bcantrill on Sunday December 08, 2024 @10:17PM (#42362787)
To be clear, I think that they should have kept iterating on Larrabee (though I would have liked to see it with an open software architecture). With respect to Larrabee, my shock was that Gelsinger had such confidence that iterating on Larrabee would have resulted in an NVIDIA of one-quarter its current size (or one-quarter the size when he recorded the interview in 2019). That's the bit that's bonkers to me: yes, they should have kept iterating (though some acknowledgement from Gelsinger that the first iteration was plagued with problems, not all of them technical[0]) -- but success was emphatically not assured just because it was Intel.

[0] https://brightsideofnews.com/blog/an-inconvenient-truth-inte...

Re:Why Gelsinger was wrong for Intel by MBCook on Sunday December 08, 2024 @10:36PM (#42362872)
If they had kept going, and really got it working, their fab advantage might have let them really go after Nvidia.

In an absolute best case scenario, he may be right.

But that seems incredibly unlikely.

Still, they should’ve had a much MUCH better GPU story than they did.

Re:Why Gelsinger was wrong for Intel by nemothekid on Sunday December 08, 2024 @11:12PM (#42363024)
>their fab advantage might have let them really go after Nvidia.

I think Cantrill spelled it out in the article, and I think people are overlooking (and for some reason really continue to overlook, or just have no idea) that nvidia's advantage is Software.

I remember Larabee as a teenager - I just graduated high school, the forums were talking about it, and I had finally saved enough money to buy an 8800GTX. The 8800GTX was also coincidentally, the first consumer GPU to support CUDA. Almost 2 decades later and it feels like nvidia has an insurmountable moat in AI due to CUDA. This wasn't an accident, nvidia spent a ton of time, with developers, getting it right. FTA:

>that NVIDIA had an indisputable lead, even in 2009; that for Intel to dominate NVIDIA it would have required conjuring software expertise and focus with which Intel has famously struggled

Ignoring this piece is huge in claming that nvidia would be a quarter of the size. Can you even imagine Intel investing in a non-x86 developer platform for 15 years while they were in the lead?

Re:Why Gelsinger was wrong for Intel by MBCook on Sunday December 08, 2024 @11:27PM (#42363074)
Not the Intel we have. But if the graphics card stuff had really started to show promise, maybe they would have become an Intel that would.

You’re dead right that Nvidia is where they are because of their software story and all the work they put into it.

I don’t know if Intel could have learned that lesson.

Re:Why Gelsinger was wrong for Intel by baq on Monday December 09, 2024 @12:29AM (#42363282)
> I don’t know if Intel could have learned that lesson.

Intel invented this model with x86.

Re:Why Gelsinger was wrong for Intel by therealcamino on Monday December 09, 2024 @11:39AM (#42367742)
Really? It doesn't seem to me like much Intel-written software has been key to x86's success. The things I can remember from Intel are compilers, performance analysis tools, and threading and math libraries. But none of them were must-haves, or differentiated x86 chips from competing CPU architectures. Most of the software that made x86 successful came from third parties.

And more recently, the reviews I read of the Archmage graphics drivers didn't make it sound like Intel had understood the importance of the software that was essential to a pretty big hardware release.

Re:Why Gelsinger was wrong for Intel by baq on Monday December 09, 2024 @03:46PM (#42370388)
Your examples are simply too recent. They've been doing it since the 70s: https://en.wikipedia.org/wiki/Intel_system_development_kit - that they forgot what made them the success they are is unfortunate, but also wholly self-inflicted.
Re:Why Gelsinger was wrong for Intel by therealcamino on Thursday December 19, 2024 @11:26AM (#42463018)
Those are dev boards -- hardware, not software.
Re:Why Gelsinger was wrong for Intel by markhahn on Monday December 09, 2024 @12:33AM (#42363300)
not sure why people give so much credit to Cuda, at least now.

AI hasn't cared about Cuda for years - sure, Torch on NV will use it, but AI is all about Torch/TF, not details below that.

Re:Why Gelsinger was wrong for Intel by hnaccount_rng on Monday December 09, 2024 @02:04AM (#42363669)
Because Torch only exists on CUDA-platforms. At least in any useful form.

This was, if anything, the main theme at Supercomputing: “there is not even a point in talking about Nvidia’s new benchmarks. You all are going to buy it anyways” coupled with very few “we are betting on buying cheap hardware (Intel/AMD GPUs) and hope that we can build the relevant parts of CUDA ourselves” and the latter is pretty much a desperation move of labs/sites that simply cannot get NVIDIA GPUs (either price or availability).

And yes that is probably the seed of the end of Nvidia’s dominance. But it will take 20 years and multiple fuck ups. Just as it did with Intel

Re:Why Gelsinger was wrong for Intel by rhaps0dy on Monday December 09, 2024 @01:50AM (#42363605)
What? This is really wrong. The name of the game these days is optimizing memory and FLOPs usage on these GPUs. For example, Hopper (H100) cards introduced the WGMMA (Warp-group Matrix-Multiply Accumulate) instruction, and anyone who does LLM training jumped to use it ASAP because without it you can't fully utilize the FLOPs. Anyone training AIs at the cutting age cares a lot about CUDA details. https://hazyresearch.stanford.edu/blog/2024-05-12-tk
Re:Why Gelsinger was wrong for Intel by hnaccount_rng on Monday December 09, 2024 @02:07AM (#42363680)
That’s not what OP meant though. That’s caring about _hardware_ capability. You could do that yourself for different hardware. And (OP’s point) any alternative hardware provider could do that for you. It just happens to be really, really hard
Re:Why Gelsinger was wrong for Intel by rhaps0dy on Monday December 09, 2024 @01:36PM (#42369013)
I think "AI is all about Torch/TF, not details below that" directly contradicts the fact that ML people very much care about the details, to squeeze performance out of the hardware.
Re:Why Gelsinger was wrong for Intel by nemothekid on Monday December 09, 2024 @02:09PM (#42369363)
>AI is all about Torch/TF, not details below that

This could be 100% true, and the best ML engineers in the world could be completely oblivious to CUDA and it wouldn't change why nvidia is leading right now.

Anyone who thinks it's "trivial" to use AMD cards should help geohot and his tinygrad boxes. Despite emailing the CEO several times, it was a clear uphill battle to get something working, and even now the sales page lists driver quality as "mediocre".

If you open a GitHub issue or a forum post about your weird CUDA application that isn't working right, you will have an engineer help you. This is in nvidia's DNA and they did this before AI started printing money. On the flipside, you can email Lisa Su today and still not have problems fixed in AMDs drivers. I'm sure AMD is trying, this isn't to say they aren't, but having decent software culture at these large companies is hard.

If AMD can't do it, I feel even less confident about Intel. If the CUDA moat truly didn't matter, we would see AMD cards in demand. However it doesn't and it isn't for lack of trying.

Re:Why Gelsinger was wrong for Intel by hnaccount_rng on Tuesday December 10, 2024 @03:59AM (#42375095)
I literally sat at a table where the argument was made: The Intel accelerator is faster. It provides more FLOPs. But you can’t get actual users to use it outside of our LINPack runs, since it just doesn’t provide bindings to all the software actually intended to run. Granted that was in the HPC space (ie FP64 performance), but that is literally the argument.

That doesn’t mean, that ML people don’t care about some hardware specifics btw. Of course they do! Because CUDA is very, very good at exploiting the specific features of the newer cards (I’ve forgotten, what exactly this was though). But take away the simple integration into their favorite AI-SDK and it would be completely irrelevant.

The point is just that, if Intel were to build strictly better hardware, but not provide the interface to torch, then they would still loose to Nvidia. Yes, yes, this is not completely true. Given enough advantage someone else would build the IntelCUDA. But this is literally the bet that those academics make. And it’s so risky, that only academics are doing that

Re:Why Gelsinger was wrong for Intel by MBCook on Monday December 09, 2024 @09:54AM (#42366734)
Maybe not now (I’m not in a position to know), but wasn’t it a HUGE factor in them becoming the player they are in non-graphics stuff? Because they had great APIs/libraries/documentation to play in that space?
Re:Why Gelsinger was wrong for Intel by wmf on Sunday December 08, 2024 @09:55PM (#42362687)
What if they kept trying on Larrabee and it never worked?
Re:Why Gelsinger was wrong for Intel by lowbloodsugar on Sunday December 08, 2024 @10:35PM (#42362868)
Which was a reasonable possibility. Larabee was a daft idea. It was never going to work, let alone at consumer prices and power.
Re:Why Gelsinger was wrong for Intel by Brian_K_White on Sunday December 08, 2024 @10:19PM (#42362797)
What if aliens came and made everything obsolete?

In an infinite universe, anything can happen, and so "it might not work" is is always true for everything. No matter what you do, it might not work, and some of the most important and valuable things ever aren't attainable without a lot of investment before you can start to reap any return.

Not all ideas are good ideas, so of course there are reasons not to do some things, or to abandon something you started, but merely "what if it never works" is not one.

Asking that is about equal to asking what if a piano falls on you tomorrow. Sure it could happen, but so what? What does that change? If you don't walk down the street for fear of pianos, you don't get to the grocery store or your job and then you die from starving instead of from a piano.

Re:Why Gelsinger was wrong for Intel by boulos on Monday December 09, 2024 @12:16PM (#42368058)
It was 15 years ago today that it was cancelled (https://web.archive.org/web/20100106162815/http://news.cnet....).

You need to remember that was during the financial crisis, and the first chips were going to be laughably bad and uncompetitive. From the quote in that article:

> Justin Rattner (Intel Senior Fellow) demonstrated Larrabee hitting one teraflop, which is great but you could walk across the street and buy an ATI graphics board for a few hundred dollars that would do five teraflops.

Intel had no chance for v0 to do anything but lose money. They were way off track, and spent most of the time trying to get to a software rasterizer plus hardware that could work for some games.

They did pivot Larrabee into the various Knight chips (Knights Corner, Landing and whatever else) since the "it's just x86" was a pretty compelling story in the 2009 era. But ultimately, HPC wasn't that great a market in the early 2010s for accelerators yet and NVIDIA ended up totally eating their lunch over time.

I think Intel could have stuck with the HPC market and gotten lucky once ML exploded, but realistically until about 5 years ago there wasn't that much ML spending yet either. 10 years of last place discrete GPU work would have been difficult to invest in.

tl;dr: the business case for Larrabee hinged on having a successful gaming business and it wasn't even close.