News Senior Intel CPU architects splinter to develop RISC-V processors — veterans establish AheadComputing

[Admin]

While Intel is busy laying off thousands of employees some of its most experienced CPU architects, with a combined 80+ years at the firm, have left to form a RISC-V startup. AheadComputing was co-founded by Debbie Marr, Mark Dechene, Jonathan Pearce, and Srikanth Srinivasan, with the goal of 'creating compelling open specification core IP.'

Senior Intel CPU architects splinter to develop RISC-V processors — veterans establish AheadComputing : Read more

 

[Alvar "Miles" Udell]

With its pedigree, surely we will hear about AheadComputing again, in the not-too-distant future.

Probably Intel buying them.

 

[bit_user]

I'm getting deja vu from Apple's CPU architects forming Nuvia. That's probably responsible for the 2-3 generations' lull in new core microarchitecture improvements, following the A14/M1. Had Apple not allowed that to happen (i.e. given them cause to leave), it's crazy to think how much further along they might be.

So, I wonder if we can expect a similar drought from Intel, in a few years. Or, maybe these folks were already hard at work on a RISC-V project at Intel and departed after receiving news of its cancellation.

However, they will also meet with people during Happy Hour on Tuesdays, between 4 - 5:30 pm, at Cornelius Pass Roadhouse, Hillsboro…

I hope they get a private room, because I'll bet there will be a few ears closely trained on them.

 

[NinoPino]

For sure this is good news for the IT world.

I think Intel and AMD should quickly jump into the RiscV wagon to prevent loosing that train, I kindly hope they are already designing something now. It is only a matter of years before x86 will be throwing away from the notebook market, than from console market and subsequently from desktops. Server will resist more, but the end is coming.

 

[bit_user]

I think Intel and AMD should quickly jump into the RiscV wagon to prevent loosing that train,

AMD reportedly has ARM cores in the works. I think it's a lot easier to convert an ARM core to a RISC-V core than to build either of them from an x86 design.

It is only a matter of years before x86 will be throwing away from the notebook market, than from console market and subsequently from desktops. Server will resist more, but the end is coming.

Server is where ARM has been making the biggest inroads, lately. If you leave Apple aside, ARM still hasn't taken much of the notebook market. Even a lot of Chromebooks are still Intel.

So, if RISC-V follows that playbook, then they should focus on server, first.

 

[NinoPino]

AMD reportedly has ARM cores in the works. I think it's a lot easier to convert an ARM core to a RISC-V core than to build either of them from an x86 design.

Ohhh, nice news. Also the use of ARM instead of RiscV could be a smart starting move that guarantee immediate access to an already mature software ecosystem.

Server is where ARM has been making the biggest inroads, lately. If you leave Apple aside, ARM still hasn't taken much of the notebook market. Even a lot of Chromebooks are still Intel.

So, if RISC-V follows that playbook, then they should focus on server, fifirst.

Agree with you for cloud and new installations, but legacy x86 server market will be lucrative for many years, at least for hardware and software support reasons.

 

[JRStern]

Good for them.
Intel should be fine, presuming they have fifty others with approximately the same skills, and maybe 20-30 years younger who will now move up. Getting rid of old folks, even who are still doing well, is a problem in big corps.
It's not particularly rocket science any more, taking a RISC core and building up multicore chips with CODECs and NPUs and all that jazz, and maybe they have some good ideas, customers already booked.
Heck, if RISC works, maybe they can branch back to x86 too.

 

[bit_user]

It's not particularly rocket science any more, taking a RISC core and building up multicore chips with CODECs and NPUs and all that jazz, and maybe they have some good ideas, customers already booked.

Huh?? It's not like just anyone can design cores competitive at the cutting edge, you know? Not sure why you're talking about codecs and NPUs, either. Those have a place in the world, but they don't eliminate the need for good general-purpose CPU cores.

That said, I doubt losing four heavyweights will be beyond Intel's ability to cope with. As you say, they should have deep talent pipelines. Assuming those haven't been depleted, I'd assume enough others can probably step up to keep the development pipeline filled.

Then again, it's not as if Intel has been executing flawlessly. Most recently, Sapphire Rapids suffered about 1 year of schedule slip, due to a multitude of chip bugs, some of which (e.g. AMX-related) I'm sure were in the CPU cores designed by the AADG.

Heck, if RISC works, maybe they can branch back to x86 too.

That requires an x86 architecture license, which only Intel, AMD, and Via Technologies hold (for legacy reasons). I don't foresee Intel issuing them to anyone else.

 

[DavidC1]

Good for them.
Intel should be fine, presuming they have fifty others with approximately the same skills, and maybe 20-30 years younger who will now move up. Getting rid of old folks, even who are still doing well, is a problem in big corps.

What? Are you serious? You know you need deep knowledge and experience that only comes with time right? Things older people have? We're not talking 70s and 80s age people here.

AADG is the group that's responsible for the "Royal Core" project that's been surfacing for a while. While some reading says the project itself hasn't been doing well, believe it when the first A in AADG is "Advanced". It was supposed to have radical concepts. Of course the performance increase was supposed to be radical as well with some saying Royal Core 1 aimed for 2x over Golden Cove(Alderlake/Raptorlake) per clock and Royal Core 2 aiming for 3x.

Perhaps it was overly ambitious. But looking at the pedigree of engineers it's a not a small loss for Intel. They won't die overnight, but they are being picked apart piece by piece.

I fear for the future of this company. Massive debt, degradation issue that affects 2 years worth of mainstream products which will break the bubble that's been keeping AMD at 20% client share, uncompetitive architectures.

 

[DS426]

RISC-V or not, IMO, AMD should take their lovely cash coffers and nab up some of these Intel veterans. Maybe Intel isn't even wrong from letting many of their six-figure folks go given their recent circumstances, but if so, AMD might as well capitalize on this. Moreover, if RISC is supposedly going to be the future of CPU computing, these seem like the folks to have on one's team to move that capability forward. Darnit, FFS, when someone has experiencing going back to 286/386/486, are you really going to let that walk away?

No wonder startups with less than 1/10 the resources and free cash as the big boyz are able to do so much!

 

[Marco_il_bello]

Sorry, I read everything and it's very interesting but then I open their site and, I was incredulous that the text in the pages comes out and it's all crooked, if you press on the careers page the text comes out all out of the graphics, ok it's the content that matters but it seems strange to me that in 2024 this happens to people who are in the IT sector anyway.
It also says that the company has been listed but I can't find anything at all on the stock exchange sites.
I also think that, it can be interesting but I also think that intel has fired many people for its now inefficiency, maybe it fired those who made less money first? Or did he fire those who understood more than the others?

How strange

 

[bit_user]

Sorry, I read everything and it's very interesting but then I open their site and, I was incredulous that the text in the pages comes out and it's all crooked, if you press on the careers page the text comes out all out of the graphics, ok it's the content that matters but it seems strange to me that in 2024 this happens to people who are in the IT sector anyway.

The font indeed looks weird on my browser (Firefox). It's like the opposite of bold. Also, low contrast vs. some of the backgrounds, to the extent that readability is significantly compromised. Didn't try other browsers, but it reminds me of the bad old days when a website would only work right on Internet Explorer.

It also says that the company has been listed but I can't find anything at all on the stock exchange sites.

It's a startup only founded a couple months ago! They probably have only angel investors, at this point.

If we take the example of Nuvia, it was founded in "early 2019", completed their series-A round of venture capital funding ($53M) in mid November 2019, completed series-B round of VC funding in Sept. 2020 ($240M), and was acquired (for $1.4B) by Qualcomm in mid-Jan 2021 (completed mid-March). As we all know, they didn't ship cores as part of a product until mid-2024 (the chips were ready at the end of 2023, when they started showing demos and leaking benchmarks). Now, part of the reason it took them so long to produce silicon is probably due to their initial targeting of server cores, which wasn't the market Qualcomm bought them for.

If Nuvia went the IPO route, then they would've first had to establish a revenue stream. That would mean shipping at least one generation of products and having another pretty far along in the pipeline. So, about 5 years is perhaps a good target for that scenario. Their investors would definitely know this.

You could look at some other players in the RISC-V market, like SiFive and (more recently) Tenstorrent. Both are private, with founding dates of 2015 and 2016, respectively. So, it could easily turn out to be a longer endeavor than the 5-year timeframe. If I had to guess, I'd say they're hoping to follow the Nuvia path and get acquired.

Finally, it's certainly worth mentioning that probably about 90% of venture-backed startups fail, and these are the ones who complete at least series-A funding. I don't have good or recent data on this, but I'm pretty sure it's (still) true.

 

[KraakBal]

And there it is. The real stab to Intel's heart.

 

[JRStern]

That requires an x86 architecture license, which only Intel, AMD, and Via Technologies hold (for legacy reasons). I don't foresee Intel issuing them to anyone else.

They can ask, old friends and all, they might choose to cover some niche that Intel doesn't, or just do it all so well that Intel allows them under cross-license and then buys them out.

 

[JRStern]

What? Are you serious? You know you need deep knowledge and experience that only comes with time right? Things older people have? We're not talking 70s and 80s age people here.

Nothing lasts forever. It's in everyone's interest to support the Silicon Valley ecosystem.
Most companies these days like to keep things turning over.
I don't know who has been at fault for the many failures over the last ten years or so, but let's just say it gets exhausting to everyone involved and maybe it pays everyone to try new things.

 

[DavidC1]

Nothing lasts forever. It's in everyone's interest to support the Silicon Valley ecosystem.

Intel had bad management, all over the company and in nearly all layers, that's what's killing them. It's not a simple 1+1 equation, it never is with such complex systems.

Internal politics, brain-dead decisions, short-term profits, lazy vision, bureaucracy all contributed to this. And I watched this company for 25 years. It existed for at least that long.

This isn't AMD pre-Zen situation, where they had much lower overhead and a focused product line. Intel is all over the place.

 

[purpleduggy]

RISCV and ARM will replace X86

 

[bit_user]

RISCV and ARM will replace X86

Seems likely, but there's quite a lot of uncertainty about when and how.

Also, I think there's sometimes a mistaken assumption that this is a zero sum situation with AMD and Intel. However, AMD is reportedly designing a new ARM-based SoC for launch next year. Intel has been dabbling in RISC-V for years and might just skip past ARM and go straight into making their own RISC-V cores.

The only way it's bad is if Intel and AMD face more competition in their core markets, which could hurt their market share and margins. We'll see if they (AMD, in particular) can grow fast enough to outpace any such challenges to their profits. If they can mount meaningful challenges to Nvidia's AI dominance, then maybe.

 

[purpleduggy]

Seems likely, but there's quite a lot of uncertainty about when and how.

Also, I think there's sometimes a mistaken assumption that this is a zero sum situation with AMD and Intel. However, AMD is reportedly designing a new ARM-based SoC for launch next year. Intel has been dabbling in RISC-V for years and might just skip past ARM and go straight into making their own RISC-V cores.

The only way it's bad is if Intel and AMD face more competition in their core markets, which could hurt their market share and margins. We'll see if they (AMD, in particular) can grow fast enough to outpace any such challenges to their profits. If they can mount meaningful challenges to Nvidia's AI dominance, then maybe.

AMD and Intel have the best optimizations, so if they were to design a new competitive RISCV chip as a flagship it would be the best on the market. However I think we would need massive investment into making ARM and RISCV work well with existing software. We need serious paid developers to build on RISCV. We need something like the Prism ARM x86 emulator for Windows 11 systems on existing software, and massive investment from industry to make RISCV software. We need steam support on RISCV. AI can go far into making this happen. Anything with an open codebase without crazy dependencies can be made to work on RISCV. we need a sort of dev bounty platform for RISCV ports where we can crowdfund this. I'd happily put forward $10 each month towards paying devs to build RISCV apps, focusing on the most important, especially the hardest, the drivers. I bet we can find a few thousand people who would be just as interested to crowdfund this. Please help do this. My money is ready. The biggest RISCV dev teams so far is the debian and ubuntu teams I think.

 

[bit_user]

I think we would need massive investment into making ARM and RISCV work well with existing software.

ARM is basically there. Amazon has been pushing it for like 5 years, with nearly all of their internal services switched over to it. Google and Microsoft have also joined the party.

RISC-V will take a while to reach the same point, but I'm sure it'll have an easier time as the third major ISA than ARM did achieving the second major ISA status. Year, POWER preceded them and is still kicking around, but it never even reached the same penetration that ARM now has and it never will.

The biggest RISCV dev teams so far is the debian and ubuntu teams I think.

These days, it's certainly the Chinese.