Thursday, August 11th 2022
Intel GPU Business in a $3.5 Billion Hole, Jon Peddie Recommends Sell or Kill
Jon Peddie Research (JPR) provides some of the most authoritative and informative market-research into the PC graphics hardware industry. The firm just published a scathing editorial on the future of Intel AXG (Accelerated Computing Systems and Graphics), the business tasked with development of competitive discrete GPU and HPC compute accelerators for Intel. Founded to much fanfare in 2016 and led by Raja Koduri since 2016; AXG has been in the news for the development of the Xe graphics and compute architecture, particularly with the Xe-HP "Ponte Vecchio" HPC accelerator; and the Arc brand of consumer discrete graphics solutions. JPR reports that Intel has invested several billions of Dollars into AXG, to little avail, with none of its product lines bringing in notable revenues for the company. Xe-LP based iGPUs do not count as they're integrated with client processors, and their revenues are clubbed with CCG (Client Computing Group).
Intel started reporting revenues from the AXG business since Q1-2021, around which time it started selling its first discrete GPUs as the Intel DG1 Xe MAX, based on the same Xe-LP architecture powering its iGPUs. The company's Xe-HPG architecture, designed for high-performance gaming, was marketed as its first definitive answer to NVIDIA GeForce and AMD Radeon. Since Q1-2021, Intel has lost $2.1 billion to AXG, with not much to show for. The JPR article suggests that Intel missed the bus both with its time-to-market and scale.
A sizable launch of Arc "Alchemist" in 2021 or early-2022, in the thick of the GPU supply crisis, would have enabled Intel to cash in on sales to whoever is in the market for a graphics card. With the supply crisis over in the wake of the crypto-currency mining demand for dGPUs, Intel finds Arc "Alchemist" competing with GeForce and Radeon products purely on gaming performance, where its fastest "Alchemist" product matches their mid-range products. Both NVIDIA and AMD are ready to ship their next-generation, which is bound to widen the performance gap with Intel even further. Besides graphics, NVIDIA and AMD are ready with even their next-generation scalar compute products, with NVIDIA's Hopper, and AMD's CDNA3, increasing the performance gap with "Ponte Vecchio."
With the recent axing of the Optane Memory business, which rode on the promise of the pioneering 3D XPoint memory technology that Intel invented, it's open-season on non-performing Intel businesses, especially with CEO Pat Gelsinger seeing favorable outcomes in Washington DC for legislation that makes business favorable for Intel and increases government subsidies for the company. JPR recommends that in light of the losses faced by AXG, Intel should consider selling the entire division off and exiting this market.
The JPR editorial can be read from the source link below.
Source:
Jon Peddie Research
Intel started reporting revenues from the AXG business since Q1-2021, around which time it started selling its first discrete GPUs as the Intel DG1 Xe MAX, based on the same Xe-LP architecture powering its iGPUs. The company's Xe-HPG architecture, designed for high-performance gaming, was marketed as its first definitive answer to NVIDIA GeForce and AMD Radeon. Since Q1-2021, Intel has lost $2.1 billion to AXG, with not much to show for. The JPR article suggests that Intel missed the bus both with its time-to-market and scale.
With the recent axing of the Optane Memory business, which rode on the promise of the pioneering 3D XPoint memory technology that Intel invented, it's open-season on non-performing Intel businesses, especially with CEO Pat Gelsinger seeing favorable outcomes in Washington DC for legislation that makes business favorable for Intel and increases government subsidies for the company. JPR recommends that in light of the losses faced by AXG, Intel should consider selling the entire division off and exiting this market.
The JPR editorial can be read from the source link below.
112 Comments on Intel GPU Business in a $3.5 Billion Hole, Jon Peddie Recommends Sell or Kill
And if Battlemage turns out to be the same disaster as Alchemist, I'd say they will kill off their gaming GPU endeavours and refocus purely on HPC accelerators...
With ATI/AMD and nVidia we had a lot of competition graphics cards launched nearly every 6 months, it's only the last few generations that AMD struggled a bit, but since RDNA AMD is back (ignoring nVidia fanboys especially among reviewers) we will have competition again.
Raja strikes again
Right?
Now this is this a failing of Gelsinger or did he put too much trust in Raja and has learned his lesson.
I'm no Intel fan, but we need them to be competitive and a 3rd gpu player is most needed. I hope they don't bail on discrete gpus, but Arc has been an unmitigated disaster for them.
Combine that with rising geopolitical tensions in East Asia where most of the manufacturing is done ....
I'm sure Intel had a crystal ball yrs ago & could EASILY see this mess! :laugh:
I was expecting their 1st product to have flaws here and there, and I could reason with that. But killing this before it even started? Damn.
Just my thoughts, hey who am I anyway? I'm just a random guy over the internet after all.
They're just throwing out some guesses that all the rest of us can in forums.
Yes, Intel would perhaps be better of by selling or axing their discrete GPU branch, like they did Optane. At least in the short run. But I don't think they will. ARC, as late and underpowered and undeveloped it is, at least shows that they can enter the discrete GPU business, competively maybe with next generation. And even if they fudge that one too and it takes too long to compete with 2022 Nvidia and AMD generations, I think they are in for the long run, to learn how to do GPU computing, not just gaming.
The gaming-portion is basically to help fund the research into the high-end stuff. Much like how NVidia uses lower-end 3060 GPUs to fund the research of $10,000 A100 server GPUs, Intel needs to get the video-game market hooked onto some product that supports the stack. That's the only way this will be economically feasible.
Now maybe the PC market is too difficult to break into. Then Intel should instead try with the console market first, or something else along those lines. Or maybe Intel makes a compute-only chip and finds a new market of buyers (but unlikely, video gamers have shown to be wanting good chips and funding this kind of SIMD-compute research). The only other niche is maybe the AI / Deep Learning fellows. But casting a wide net and getting as many customers as possible is just the most obvious strategy here.
They are in a free fall right now, last quarter was 22% down compared to last year, but so is the rest of the market. We are in a recession, whatever the definitions say.
I'm sure many shareholders would welcome axing such an expensive project for momentary gain, no matter the potential. And that's what might actually happen.
ATI/AMD : from a Mach64 LT, a few Rage, a X1650 Pro passing by a X1950 Pro a HD3650 3670 4870 a 5XXX i don't remember :laugh: 7870 R9 270 290 and then RX 6700 XT (and pobably a few more i forgot)
Nvidia: from Tnt1 , Geforce 256, Geforce 2 GTS, :oops: MX400 :oops: 460 480, 560 580 (even a 580 SLI of Matrix Platinum) 670, 770, 980 and lastly 1070.
i spent enough time with them since then :laugh: nonetheless i's down to personal point of view and luck, in my case, i guess :ohwell:
edit: as i said, i really loved the Real3d Starfighter i740, even tho it was sh!t performance wise in 2D and utterly weak in 3D (i had a 8mb version ) (although drivers were fine with that one :oops: )
and i will not talk about the S3, Permedia, PowerVR and 3Dfx i owned.
As for that link, it is looks more like cypto hangover than anything else.
Now, how bad is being 13% behind RDNA1? Not great, to put it softly, but not an unrecoverable catastrophe either. Especially given Mr Koduri's involvement...
I think we need to look into the big picture - long term. With introductory ARC the performance is not that bad tbh as every new product ( with a new arch.) needs time to flourish and become stable. I think intel can create an impact in the dGPU market in the coming years.
Slowly it will get there!
AMD was a financially starving company with shrinking market share.
At some point (and it certainly wasn't 7 years) they bet on Zen architecture and then it delivered.
Once GPU division got proper funding IN A MERE YEAR AMD was able to hit with RDNA1, which was a major leap forward, essentially closing perf/mm2 gap and perf/watt gaps vs team green. One more year, and RDNA2 disrupts NV lineup, forcing Huang to drop a tier (that is why we got 10GB 3080 and 12 GB on lower end cards). The only reason we haven't seen that havoc in NV financials is the crypto bubble.
So, bottom line is, it doesn't take 7 years for a competent GPU manufacturer to roll out serious competition.
Intel, of course, is rather lacking respective experience, but... perhaps 13% worse than RDNA1 for the first product is not that bad.
in short what did happen for both enterprise? Raja happened :laugh: (kinda sad tho ... )
Look at the star wars prequels, that is what happens when you think its all 1 person responsible.....
But in terms of embarassing statements made - it is absolutely solely Raja. It stopped the second he left AMD at AMD and it started to pop out of Intel once got active there.