News Intel GPUs - Intel launches A580

[moinmoin]

Kinda impressive: https://ieeexplore.ieee.org/author/37611561100

But why does AMD need him? Did they hire him just to spite Intel?

He'd been at Silicon Graphics in the 1990's so he and David Wang and the likes likely know each other. The interesting part is how much all these careers deviated since, David Blythe having been at BroadOn, Microsoft and Intel since. Unlike the many people from the SGI/ArtX/ATi lineage he's mainly a software guy though.

 

[eek2121]

They don't need a top tier card. They just need a competitive one (perf/$ and preferably perf/watt). Also, an architecture that is competitive in the enterprise space. If they can pull that off they will be successful.

I am a bit concerned by this development. Hopefully Intel doesn't kill Arc. Arc at launch was a disaster, but it has steadily improved since launch.

 

[Hans Gruber]

You guy need to understand the intel strategy. It's to have 225-250w cards. Why would they kill ARC? Battlemage is supposed to be the 1st high end GPU from Intel. They have an ARC refresh coming later this year and then Battlemage. If Intel has fire sale prices on the A750 and A770. Count me in for the A770 in the $250 or less price range.

That Raja guy was all smoke and mirrors. AMD bought the very best GPU maker in ATI. They have pissed away what ATI once was. AMD never made bad GPU's. They made bad CPU's for a decade before Ryzen.

 

[jpiniero]

You guy need to understand the intel strategy. It's to have 225-250w cards.

The point of Arc was to fill the fabs. Intel's fabs, not TSMC's.

 

[Hans Gruber]

The point of Arc was to fill the fabs. Intel's fabs, not TSMC's.

Intel had 7nm fab problems which ARC GPU's were going to be based on. I heard they wanted to see how their GPU would run on TSMC silicon. It also gave them a closer look as a customer into the competitions silicon. Agreed, not originally part of the Intel plan or roadmap.

 

[mikk]

MILD switches from Arc dGPUs cancelled to "at least" 250mm² die planned" for Battlemage and a 180mm² Celestial die. As for Celestia it's way too early for specific die specifications, this isn't trustworthy at this point. The 250 mm² Battlemage die is a 12GB chip he says. I doubt Intel can release a dGPU which has less VRAM than Alchemist. However he claimed Arc dGPUs had been cancelled which he denies now, what a fail.

 

[Dayman1225]

MILD switches from Arc dGPUs cancelled to "at least" 250mm² die planned" for Battlemage and a 180mm² Celestial die. As for Celestia it's way too early for specific die specifications, this isn't trustworthy at this point. The 250 mm² Battlemage die is a 12GB chip he says. I doubt Intel can release a dGPU which has less VRAM than Alchemist. However he claimed Arc dGPUs had been cancelled which he denies now, what a fail.

If MLID says it’s not cancelled, then it’s cancelled. /s

 

[Exist50]

The point of Arc was to fill the fabs. Intel's fabs, not TSMC's.

I don't think that is really the case. Arc was never intended to be on an Intel process, as far as I'm aware. If anything, I think they isolated it from Intel's fabs to avoid cross-contamination.

 

[jpiniero]

I don't think that is really the case. Arc was never intended to be on an Intel process, as far as I'm aware.

Well in any case, there's pretty obviously now no path to profitability unless they do fab internally. So there's really no point in continuing unless they like losing money. Which they might.

 

[Exist50]

MILD switches from Arc dGPUs cancelled to "at least" 250mm² die planned" for Battlemage and a 180mm² Celestial die. As for Celestia it's way too early for specific die specifications, this isn't trustworthy at this point. The 250 mm² Battlemage die is a 12GB chip he says. I doubt Intel can release a dGPU which has less VRAM than Alchemist. However he claimed Arc dGPUs had been cancelled which he denies now, what a fail.

I'm not going to give MILD's ramblings any real weight, nor do I have any particular insight into what Intel's new GPU roadmap is. However, just for the sake of speculation, if they moved to a single die on a two-year cadence, vs multiple dies on a one-year cadence, that would certainly save a lot of money. Probably would be a more sane place to make the budget cuts than in the IPs themselves.

Well in any case, there's pretty obviously now no path to profitability unless they do fab internally. So there's really no point in continuing unless they like losing money. Which they might.

If they had a competitive architecture, they'd be fine fabbing it externally, but obviously all else equal there would be cost advantaged to doing it on Intel fabs. But that's not a realistic possibility until 18A at best. So maybe Celestial?

 

[Aapje]

Charlie is saying the leaving is amicable. Based on @Exist50 saying leaving was inevitable, maybe he's leaving after demotion back few months ago.

In the tweet, Pat said that Raja leaving is bittersweet. MLID leaked an internal mail where Pat was more explicit, saying that he had mixed emotions. When you say that when someone is leaving, at this level, that is like telling everyone that you are happy they are gone.

That said, an Intel employee said that this is how they often get rid of senior management. Reassign them to a fake job with no actual responsibilities and then they get some months to find a new job while still getting paid.

 

[Aapje]

The point of Arc was to fill the fabs. Intel's fabs, not TSMC's.

I think that the point of Arc is to get all this sweet, sweet AI supercomputer money. GPU's are becoming increasingly useful outside of gaming.

I don't think that it is a coincidence that Arc raytraces really, really good compared to how well it rasterizes. Raytracing uses the same cores that are used for AI calculations.

 

[Hitman928]

Raytracing uses the same cores that are used for AI calculations.

Pretty sure that this is not true. Maybe you are thinking about the Tensor cores in RTX cards? Those are for machine learning but are separate from the RT cores.

 

[GodisanAtheist]

Pretty sure that this is not true. Maybe you are thinking about the Tensor cores in RTX cards? Those are for machine learning but are separate from the RT cores.

- Yeah I recall initially NV was saying tensor cores were used in denoising, but that never actually panned out and they've been repurposed hard into DLSS applications.

Also, reading the early years of this thread were pretty wild. People were all over the place. By 2020/2021 though it looked like opinion had pretty strongly turned against Intel as they kept missing launch targets and dates.

 

[IntelUser2000]

And now they've been cutting the former-AXG budget hard, and laying off a good number of people. I don't think they'll kill Arc, but it's tough to see a path to them legitimately competing with Nvidia, or even AMD.

Unless Battlemage is indeed much more competitive, then in spirit, MLID's former(not the recent one) claims of ARC is more or less cancelled is correct is it not?

He did get few things right such as XeSS name and Redwood Cove core. Someone said he has really good sources, but he misinterprets them in his way.

They are also cutting prices like mad. Newegg was doing $225 A750 promotion. How do they make money at all? Also why are they so aggressively pricing it? Is the refresh really coming and are they preparing stock for that?

 

[igor_kavinski]

Is the refresh really coming and are they preparing stock for that?

I hope it doesn't mean the opposite.

What if, Intel opensources the drivers and takes a few years of hiatus from the dGPU market, keeping their technology alive in mobile iGPUs? I'm sure Pat doesn't feel good that ARC didn't register enough on Nvidia's threat radar to make them mock it. One good thing about opensource drivers is that Intel may be able to find better GPU driver writers more easily once they decide to get serious again.

 

[A///]

Come on Nvidia, hire Raja! ... You know you want to! Redefine the GPU market in a profoundly seismic way!


( Leeea may be an AMD fan person )

I'm sure jh has threatened death to any executive considering hiring on raja.

 

[A///]

The point of Arc was to fill the fabs. Intel's fabs, not TSMC's.

Not quite. Even if intel figures out their lower nodes it's still very good to go ahead with tsmc. It'll allow them to focus on cpus of their own to sell and until IFS comes roaring online they've got big potential customers to use and not themselves. there's many advantages to using tsmc as a partner foundry rather than do it yourself.

 

[A///]

what a fail.

as a leaker, youtuber, and mouth breathing idiot. IDK why people even watch his bs videos.

 

[Exist50]

Unless Battlemage is indeed much more competitive, then in spirit, MLID's former(not the recent one) claims of ARC is more or less cancelled is correct is it not?

I don't think so. I've been unable to get real numbers on their budget cuts, but I didn't get the impression they were severe enough to herald Arc's cancelation. I think Battlemage will certainly be better, but competitive with Nvidia? Nah.

Though against my better judgement, I did very briefly skim the MLID video. The one thing that caught my attention was the "uncompetitive zombie" remark. If they could coast for a few years, putting in enough investment not to fall further behind, and cutting their financial losses dramatically, that might just work. Around 2026 is when Gelsinger's promised budget cuts should end, and we'll have much more clarity on the future of their server biz and IFS by then. It would probably be Intel's first real opportunity to expand on the design side.

They are also cutting prices like mad. Newegg was doing $225 A750 promotion. How do they make money at all? Also why are they so aggressively pricing it? Is the refresh really coming and are they preparing stock for that?

I doubt they're making any money on these, but the PC market is pretty crazy right now. From their financial reports, sounds like they have a ton of excess inventory that they need to get through one way or another. They can either do a fire sale, or write it all off, and it seems they chose the former.

 

[Terzo]

Unless Battlemage is indeed much more competitive, then in spirit, MLID's former(not the recent one) claims of ARC is more or less cancelled is correct is it not?

He did get few things right such as XeSS name and Redwood Cove core. Someone said he has really good sources, but he misinterprets them in his way.

They are also cutting prices like mad. Newegg was doing $225 A750 promotion. How do they make money at all? Also why are they so aggressively pricing it? Is the refresh really coming and are they preparing stock for that?

This is subjective, but the DC area Microcenters used to have 20+ units in stock of A770 all the time. Out of curiosity I just looked and the Fairfax/Rockville Locations have 3 New in box (plus another 3 open box) between the two of them. I don't remember Microcenter ever putting them on a discount, so either they finally sold off their initial stock and haven't ordered more or more people are starting to buy them.

 

[DAPUNISHER]

Homie getting fired has invigorated this discussion.

The software for ARC has progressed shockingly fast IMO. If I were a betting man I'd have lost my ass wagering on that happening. At the present pace, by the time battlemage, or as I shall call it, Gandalf, could be breakout success. Instead of Gandalf the Grey, we could get, Gandalf the White.

I can envision day one reviews where instead of having to talk about all the things that are broken or buggy, the review is about how install, setup, and testing were very run of the mill; a good thing. And how it outperforms everything else at the respective price points.

AMD and Nvidia are judged on flagships, but Intel would get a hero's welcome for simply providing plug and play, stable and solid, bang for buck mid range and entry level champs. I can dream

 

[VirtualLarry]

AMD and Nvidia are judged on flagships, but Intel would get a hero's welcome for simply providing plug and play, stable and solid, bang for buck mid range and entry level champs. I can dream

Intel has a chance to make themselves the Volkswagen of GPUs. That is, if they can avoid being the Trabant of GPUs.

 

[Aapje]

The software for ARC has progressed shockingly fast IMO. If I were a betting man I'd have lost my ass wagering on that happening. At the present pace, by the time battlemage, or as I shall call it, Gandalf, could be breakout success.

Yes and no.

The stability and game support has improved dramatically, but the performance is still very far off from what you'd expect given the die size (and thus cost). So it seems to be a flaw in the design of the hardware.

Without a big improvement on that front, Intel can only be competitive at a price where they make a loss, which is not sustainable.

 

[Thunder 57]

I don't get why he made so much money being such a failure.