Intel Co-CEO Dampens Expectations for First-Gen "Falcon Shores" GPU

[AleksandarK]

Intel's ambitious plan to challenge AMD and NVIDIA in the AI accelerator market may still be a little questionable, according to recent comments from interim co-CEO Michelle Johnston Holthaus at the Barclays 22nd Annual Global Technology Conference. The company's "Falcon Shores" project, which aims to merge Gaudi AI capabilities with Intel's data center GPU technology for HPC workloads, received surprising commentary from Holthaus. "We really need to think about how we go from Gaudi to our first generation of Falcon Shores, which is a GPU," she stated, before acknowledging potential limitations. "And I'll tell you right now, is it going to be wonderful? No, but it is a good first step."

Intel's pragmatic approach to AI hardware development was further highlighted when Holthaus addressed the company's product strategy. Rather than completely overhauling their development pipeline, she emphasized the value of iterative progress: "If you just stop everything and you go back to doing like all new product, products take a really long time to come to market. And so, you know, you're two years to three years out from having something." The co-CEO advocated for a more agile approach, stating, "I'd rather have something that I can do in smaller volume, learn, iterate, and get better so that we can get there." She acknowledged the enduring nature of AI market opportunities, particularly noting the current focus on training while highlighting the potential in other areas: "Obviously, AI is not going away. Obviously training is, you know, the focus today, but there's inference opportunities in other places where there will be different needs from a hardware perspective."



It seems like Falcon Shores won't be the savior that will get Intel at the level of NVIDIA in the GPU market, but rather a step toward a top-notch product. Intel's next iteration after Falcon Shore is called "Jaguar Shores", which will be the late 2025 or early 2025 product for data center AI and HPC. It seems like until Jaguar Shores arrives, a lot of effort goes into software and platform development to get it right. NVIDIA has dominated the game mainly due to its amazing CUDA software, despite competitors like AMD offering comparable hardware performance-wise. Intel needs to enable the software ecosystem a seamless integration of its next-gen accelerators first to ensure its Jaguar Shores GPU can catch up with the rest of the industry.

View at TechPowerUp Main Site | Source

 

[Daven]

...is it going to be wonderful? No, but it is a good first step.

That's very close to what my wife said on our wedding night. I guess that makes sense since following Intel is like a marriage: full of great expectations and sometimes epic let downs. Lol!

 

[Tek-Check]

She was so vague

 

[Daven]

She was so vague

No, I'm pretty sure she was very clear. They have nothing and will have nothing for the foreseeable future. At a time when dozens of players are accelerating their product strategies into new markets forming every 5 years, Intel is going to slow down and iterate? Lolololol. These two nitwits (Michelle and David) are just bidding their time until the golden parachute can be attached.

 

[FreedomEclipse]

and sometimes epic let downs. Lol!

Is that really what she said?

 

[Space Lynx]

Intel is so fucking lucky the USA tax payer keeps them propped up, lol

 

[Daven]

Is that really what she said?

Lol, no. I made it up.

Intel is so fucking lucky the USA tax payer keeps them propped up, lol

The final stage of a failing company is a government bail out.

 

[AcE]

At least honest - that’s rare for an Intel ceo.

 

[bonehead123]

Not really a problem, as they have a back-up plan.....

Just replicate the 14+++++++++++++++++++++++++++++++++ clusterf*ck & everyone will be happy

These two nitwits (Michelle and David) are just bidding their time until the golden parachute can be attached.

Well of course, why else would they even consider taking on a "co-ceo" gig, after seeing what Pattyboi got after only 3 years at the helm, and FAILING miserably !

 

[TechBuyingHavoc]

Not really a problem, as they have a back-up plan.....

Just replicate the 14+++++++++++++++++++++++++++++++++ clusterf*ck & everyone will be happy

Well of course, why else would they even consider taking on a "co-ceo" gig, after seeing what Pattyboi got after only 3 years at the helm, and FAILING miserably !

You forgot a few pluses there.

Also, Intel's naming schemes just suck. Just endless coves and lakes, none of it makes sense, there are way too many architectures or small permutations of architectures for too little actual performance improvements. The same story exists for their nodes, a bunch of marketing fluff and little actual substance there.

The one product that has substantial performance improvements and not a lot of special marketing names and pointless generations? Their Arc GPUs which might not exist in the next few years due to Intel's financial problems.

 

[Vayra86]

Epic facepalm. 'Give me more Agile CPUs'

Its exactly what brought you here idiots.

All I read is fear. They need to iterate so much because they fear they can't control the result. And they fear that because they don't trust their engineers. And they don't trust their engineers because they didn't trust them leading up to this point - OR - they're simply out of good ideas.

This ain't going places, that's for sure now. This is the Intel culture that brought us a post-launch degradation issue that they want to fix with a software update. Super Agile, too bad hardware isn't software - and even in software, Agile is often a curse because if you don't think it through, you'll still create damage along the way in your live environment.

Basically when managers want products and not teams to be Agile you'd better abandon ship ASAP. It'll take a loooong time before I'll be touching an Intel CPU again, with these clowns at the helm. At least Pat was an engineer... I don't know what this is. A team working Agile is fine, if they are professionals. A product that needs to be Agile? That's just a sorry excuse for delivering broken shit. Microsoft is writing that book as we speak, and we knew Intel was already.

 

[eidairaman1]

Lol, no. I made it up.


The final stage of a failing company is a government bail out.

Which that corp should not be bailed out, too many get bailed out and not change their ways for the better.

 

[Wirko]

Super Agile, too bad hardware isn't software

Haha, and Sapphire Rapids obviously was an early victim of that Agile development cycle (or cycles) (many of them).

So far, Intel has released A0, A1, B0, C0, C1, C2, D0, E0, E2, E3, E4 and E5 steppings of Sapphire Rapids processor to fix nearly 500 bugs. Given that modern processors integrate tens of billions of transistors, it is inevitable that have a certain number of bugs. They are called erratas and are mitigated with microcode or even software updates. But 500 erratas seems overwhelming, as does 12 respins considering that a respin costs tens of millions of dollars.

 

[Vayra86]

Haha, and Sapphire Rapids obviously was an early victim of that Agile development cycle (or cycles) (many of them).

View attachment 375580

Yeah so every time you force your teams to launch a product, they have to clean the whole thing up, polish it, erase the critters and you've spent many months just on that and not developing a better CPU. It works in a similar way in software, too. Sure, if you keep your slate clean all the time, regression test everything all the time... you have fewer or nearly no bugs and you can likely keep it cleaner. But then you're also spending a huge amount of resources on regression while you're not delivering, again resources that aren't thinking about better solutions.

 

[marios15]

The timing of his comment sounds like Pat wanted to do a Zen move by going back to the drawing board and build a new arch from scratch or maybe a new process node
Some people may have made fun of Pat gelsinger but the guy was an engineer

Now this guy is taking a shot at that, thinking what he said is a flex!

Because the aim for intel, is to satisfy investors with annual releases of slightly better products

My guess is they scrapped all new products and will start releasing Lakes and 10nm+++++