Top Researchers Leave Intel To Build Startup With 'The Biggest, Baddest CPU' (oregonlive.com) 61
An anonymous reader quotes a report from OregonLive: Together, the four founders of Beaverton startup AheadComputing spent nearly a century at Intel. They were among Intel's top chip architects, working years in advance to develop new generations of microprocessors to power the computers of the future. Now they're on their own, flying without a net, building a new class of microprocessor on an entirely different architecture from Intel's. Founded a year ago, AheadComputing is trying to prove there's a better way to design computer chips.
"AheadComputing is doing the biggest, baddest CPU in the world," said Debbie Marr, the company's CEO. [...] AheadComputing is betting on an open architecture called RISC-V -- RISC stands for "reduced instruction set computer." The idea is to craft a streamlined microprocessor that works more efficiently by doing fewer things, and doing them better than conventional processors. For AheadComputing's founders and 80 employees, many of them also Intel alumni, it's a major break from the kind of work they've been doing all their careers. They've left a company with more than 100,000 workers to start a business with fewer than 100.
"Every person in this room," Marr said, looking across a conference table at her colleagues, "we could have stayed at Intel. We could have continued to do very exciting things at Intel." They decided they had a better chance at leading a revolution in semiconductor technology at a startup than at a big, established company like Intel. And AheadComputing could be at the forefront of renewal in Oregon's semiconductor ecosystem. "We see this opportunity, this light," Marr said. "We took our chances." It'll be years before AheadComputing's designs are on the market, but the company "envisions its chips will someday power PCs, laptops and data centers," reports OregonLive. "Possible clients could include Google, Amazon, Samsung or other large computing companies."
"AheadComputing is doing the biggest, baddest CPU in the world," said Debbie Marr, the company's CEO. [...] AheadComputing is betting on an open architecture called RISC-V -- RISC stands for "reduced instruction set computer." The idea is to craft a streamlined microprocessor that works more efficiently by doing fewer things, and doing them better than conventional processors. For AheadComputing's founders and 80 employees, many of them also Intel alumni, it's a major break from the kind of work they've been doing all their careers. They've left a company with more than 100,000 workers to start a business with fewer than 100.
"Every person in this room," Marr said, looking across a conference table at her colleagues, "we could have stayed at Intel. We could have continued to do very exciting things at Intel." They decided they had a better chance at leading a revolution in semiconductor technology at a startup than at a big, established company like Intel. And AheadComputing could be at the forefront of renewal in Oregon's semiconductor ecosystem. "We see this opportunity, this light," Marr said. "We took our chances." It'll be years before AheadComputing's designs are on the market, but the company "envisions its chips will someday power PCs, laptops and data centers," reports OregonLive. "Possible clients could include Google, Amazon, Samsung or other large computing companies."
Good luck (Score:2, Flamebait)
On the other hand maybe you will score a few patents and get a buyout.
With the absolute complete lack of antitrust law enforcement that's really all anyone can hope for anymore.
Re: (Score:2)
I can imagine there could be a shift of some venture capital going from AI to them.
What better opportunity to mold a new CPU architecture to have AI support from the ground up and leave obsolete cruft behind?
Re: (Score:1)
What better opportunity to mold a new CPU architecture to have AI support from the ground up and leave obsolete cruft behind?
Everyone thinks "I'm going to start over and eliminate all the old mistakes". And they just create all new mistakes that are just as bad as the old ones. The "obsolete cruft" has been tested and refined and it works. I hope they are successful, but history says otherwise.
Re: (Score:2)
DEC with Alpha?
Re: (Score:2)
When our best Intel server was a P-166 we also had an 550 mhz Alpha, which IIRC also had multiple pipelines.
Re: (Score:3)
Everyone thinks "I'm going to start over and eliminate all the old mistakes".
But they're not "starting over". They're using RISC-V, which is a mature instruction set with existing toolchains.
Re: (Score:2)
Re: Good luck (Score:2)
If investors thought that had a real chance of happening, they wouldn't invest.
Besides, intel likely doesn't see this as any more of a threat than other architectures that had the same goals, like PowerPC.
No, the world still hasn't ended. Go back to your fidget spinner, I'll let you know when you can stop.
Re: (Score:3)
Intel *do* have an interest in Risc-V - at one point they were investing in Horse Creek and exploring plans to acquire SiFive.
But given their 'core focus' announcements of the past few months, no side project on another architecture is going to pass muster, so it may well be that Intel management encouraged the spin off startup.
Going back to something akin to Horse Creek, what would be a likely scenario is if Ahead used their contacts within Intel to fab their designs on a compute-module form factor that co
Re: Good luck (Score:1)
Isn't this how half of Silicon Valley 1.0 started? Researchers/designers/scientists leave one company to start their own? Why is this so noteworthy?
Re: (Score:2)
Indeed. Intel's founders included many of the Traitorous Eight [wikipedia.org].
Re: (Score:2)
Probably because of the people involved, which makes them more likely to be successful than other startups.
Re: (Score:2)
prevent Intel from using their Monopoly to crush you.
What monopoly?
Lately, Intel has been a crushee, not a crusher.
Wrong name, bad vibe (Score:3)
Big Beautiful Chip would have done it.
Cerebras laughs (Score:2)
I have never been very impressed with the level of hype in Cerebras's self-promotion, but their ridiculously large [cerebras.ai] chips currently hold the title for "biggest, baddest CPU" and will not be easy to displace.
Re: (Score:3)
With blackjack and hookers! (Score:3)
On second thought, forget the CPU.
RISC-V is going to win. (Score:2)
First, RISC-V being open source will result in widespread and extraordinary development. Much of it will be successful in enhancing performance, losing power consumption, and introducing new capabilities.
Second, this team might be thinking of the next big leap forward, something entirely new, leapfrogging RISC-V, maybe keeping the open source advantage.
But RISC-V will win, it will advance behind its only true competition, ARM, and leaving the other major legacy players behind.
Re: (Score:1)
2025 = Year of the RISC-V desktop!!! ... O wait that still hasn't happened for Linux so yea not going to happen.
Re: (Score:2)
Desktop is so 90s...
Year of the Linux Desktop ... (Score:3)
2025 = Year of the RISC-V desktop!!! ... O wait that still hasn't happened for Linux so yea not going to happen.
We have already had the Year of the Linux Desktop. Surprisingly, it was the Year of Linux on the Windows Desktop. As Microsoft allowed Windows users to run various Graphical Linux distributions on the Windows Desktop using the Windows Subsystem for Linux (WSL2). Linux Distros install by the Microsoft App Store.
However 2025 is The Year of the RISC-V microcontroller. $5 on the Raspberry Pi Pico 2 microcontroller. ARM and RISC-V cores.
https://www.raspberrypi.com/pr... [raspberrypi.com]
Re: (Score:2)
No we didn't. Permission != Adoption. 99% of computer users in the world don't know what Linux is, much less use it in Windows, in a VM, or natively.
the ecosystem (Score:2)
Re: the ecosystem (Score:2)
ARM and/or its licensees already use these elements, they license or co-license them. The premium is paid for the CPU. Of course, in mobile, the whole ecosystem needs to be considered. That's almost the easy part, unless the legacy partners (ARM, Intel, Samsung, etc) choose to leverage their positions and refuse to sell into the RISC-V market...
Re: (Score:2)
I'm waiting for RISC-VII
or maybe RISC-VIII
(Of course RISC-IX would be taken out by the Butlerian Jihad)
Re: RISC-V is going to win. (Score:2)
Way to spice things up.
Sell (Score:2)
Re: (Score:2)
I gambled on bit of their stock after their big drop last year, but if their senior technical folks are bailing, time to get out.
Any time you say "I see a signal so it's time to sell/buy", the actual time was weeks ago. Unless you are an electronic trader responding to freshly published information.
That's not saying you shouldn't sell. But this news could already be built into the price, and it could be time to buy. It's really hard to tell.
RISC-V's heritage is US and Swiss ... (Score:5, Informative)
RISC-V is open source, not based in the US, and has the potential to completely change the economics of the processor market. This is why the U.S. government considers the technology a threat to national security and may try to suppress its adoption.
Nope. RISC-V is an open standard, originating from taxpayer funded research at UC Berkeley. Currently overseen by a Swiss non-profit.
There is a lot of fear around China having a first class processor option to produce.
The fear is around "backdoors" in China-based implementations of RISC-V. These may very well be prohibited. However EU based implementations will likely be just fine, and of course there will be US based implementations as well.
Noncompete? (Score:2)
The article doesn't say if these people were bound by some kind of noncompete agreement. If they were, that could be a big hurdle for them.
Noncompete- trade secrets & customer relations (Score:3)
The article doesn't say if these people were bound by some kind of noncompete agreement. If they were, that could be a big hurdle for them.
In Oregon the noncompetes seem to be limited, focused on trade secrets and customer relationships. So general CPU architecture design and development experience is not prohibited. The timespan is also limited to 12-months, which of course does not end an obligation to protect a former employer's trade secrets.
Dade said it in 1995 (Score:2)
"Risc is good".
RISC won over CISC ... x86 just an API (Score:3)
All modern CISC CPUs are internally RISC-like architectures but have the added benefit of essentially instruction set compression...
I think that is better described as hardware based instruction set translation, each x86 instruction generating a series of RISC microops that will actually be scheduled and executed.
Also, given that CISC has been downgraded to a software front end - a software API for the RISC hardware - I'd say RISC won. It is this hardware translation to RISC that allowed the x86 API to persist. Successfully thwarting Alpha, PowerPC and MIPS. All CPU architectures that Microsoft had running on the retail version of Wi
Re: (Score:1)
I think that is better described as hardware based instruction set translation, each x86 instruction generating a series of RISC microops that will actually be scheduled and executed.
Also, given that CISC has been downgraded to a software front end - a software API for the RISC hardware - I'd say RISC won. It is this hardware translation to RISC that allowed the x86 API to persist. Successfully thwarting Alpha, PowerPC and MIPS. All CPU architectures that Microsoft had running on the retail version of Windows NT 4. Ready to go whichever way the public chose. Obviously the public chose compatibility and low cost, the x86 API.
Well, for the exact reason you are using to claim that RISC won I think are saying the opposite. X86 CISC is the actual ISA in use. No one is compiling directly to internal x86 microops...
Internally, yes, RISC is a much better architecture to optimize for in hardware--so in that sense it absolutely "wins", but the added benefit of an outward-facing CISC ISA which reduces instruction cache pressure and makes for much, much smaller binaries was a huge benefit in the early days. Most of the RISC architectures
Re: (Score:2)
Compaq and Microsoft had a Redmond lab which was porting Windows 2000 to the Alpha chip. Unfortunately Capellas was more interested in the upcoming merger with HP stopped production of the Alpha. At that time my company's fastest Intel server ran a P-166 while 64-bit Alpha chips that ran at 550 mhz were available.
Incumbent economies of scale won, not RISC or CISC (Score:3)
RISC killed it when a small team could crank out a more than competitive RISC processor on a gate array.
By the late 90's cpu chips had become much much larger in terms of gate count. This created enormous opportunities for microarchitecture optimisations but exploiting this opportunities required huge expensive teams and enough demand for the finished product to pay for it all. Only the PC market was big enough and that was owned by x86. Other architectures, RISC and CISC alike, starved. ARM survived by
Please Intel... (Score:2)
Start focusing on clock speed and core count again. Stop trying to make chips that go faster by predicting the future in ways that are highly exploitable by malware.
Re: (Score:3)
Start focusing on clock speed and core count again. Stop trying to make chips that go faster by predicting the future in ways that are highly exploitable by malware.
They would if they could, but Physics has something to say about faster clock speeds. Intel has been telling devs for decades that future gains will mainly be through additional cores. I suppose you can add dedicated coprocessors too, like GPUs.
Re: (Score:2)
I think "dedicated coprocessors" are going to have longer traces, and thus be slower, unless, perhaps, you connect them with an optical link. (That, of course, has its own problems.)
Re: (Score:2)
Stop trying to make chips that go faster by predicting the future in ways that are highly exploitable by malware.
That seems like short-term thinking. A better approach would be to figure out how to do the future-prediction optimizations in ways that malware cannot exploit, so we can reap the benefits of the optimizations.
FWIW, RISC-V available on Raspberry Pi Pico 2 - $5 (Score:3)
"The unique dual-core, dual-architecture capability of RP2350 allows users to choose between a pair of industry-standard Arm Cortex-M33 cores, and a pair of open-hardware Hazard3 cores, giving developers a chance to experiment with the RISCV architecture in a stable, well-supported environment."
https://www.raspberrypi.com/pr... [raspberrypi.com]
Re: (Score:3)
Re: (Score:2)
and history repeats itself (Score:2)
"call me Zilog" . . .
perhaps they could call this processor the "Z80,000" :)
Nearly a century at Intel (Score:2)
which the use case makes them money? (Score:2)
These people sound very skilled and ambitious and may come up with an extra-snazzy RISC-V implementation, but there is already a lot of that out there. Maybe they can design devices that run Linux fast enough and cheap enough to sell as a commodity product. More power to them.
The real money at this point is in Nvidia-scale hardware and the associated software. As the article stated, the "professor who teaches microprocessor design and computer architecture" is "skeptical that it can prevail as the industry
Heard this tune before (Score:3)
I've heard this before: Transmeta, IDT, Cyrix, NexGen, Chips & Tech, TI, UMC, etc.. They all announced the next big CPU too but most either didn't live up to their claim or were bought out by the big usual suspects.
Re:Heard this tune before (Score:5, Interesting)
Interesting name you put first. Transmeta's failure was one of timing. The thought of a processor focused on low thermal and power envelope was simply not understood and their product fobbed off as irrelevant. If they had started a decade later they may very well have become the ARM of today, a name I chose to write to demonstrate that yes it is actually possible for an outside company to enter the computing CPU market and in some cases even dominate (see Apple)
Re: (Score:2)
It is my understanding that the VLIW model hinged on the compiler doing the critical work, but no one ever managed to write such a compiler.
Re: (Score:2)
I actually gave Transmeta a try years ago. I purchased a Compaq TC1000 tablet PC and the product was very well thought out. The Transmeta CPU though was hugely under powered for the task. A coworker of mine purchased the TC1100 a year later that contained a Pentium M CPU and it made the tablet amazing. His battery life was right on par with mine as well even though the Transmeta CPU was supposed to be superior in this regard. I also used a few of those Cyrix upgrade CPUs back in the 386/486 days. They were
Whatever they are doing TFA doesn't explain it (Score:2)
RISC hasn't been new for decades.
RISC V isn't new either. There are many players already.
Neither can be the foundation for a major new player unless the founders are idiots and I doubt they are.
So, whatever wrinkle will set them apart is something not disclosed in TFA or any other source I have found.
Ex-Beast Lake Engineers (Score:4, Interesting)
Re: (Score:2)
Plain old SMT does that, just not to a very large degree. It runs two threads, on shared resources where a resource can be "rented" by one thread or the other.
Steve Jobs quote (Score:2)
The article summary (Score:2)
reads like an 80s ad.... Risc is the future Risc is so much faster, we are going to build a risc design...
There has always been a better way (Score:1)
There has always been a better way, but stubborn fascination with x86 has held the chip industry back for decades. It is definitely time to move on and I'm glad someone has taken it up.
What's the niche? (Score:2)
Maybe there is still some room for more massively parallel processing for internet services with tiny cores or massive SMT? SPARC tried the latter but failed, but you never know.
Just making traditional CPUs, but with different ISA, for servers is unlikely to do much. Because of all the speculation the ISA is only an abstraction, which combined with the ocean of cache makes the ISA mostly irrelevant.