Former Intel engineers form AheadComputing to break CPU performance limits with RISC-V design

[Skye Jacobs]

What just happened? In the heart of Oregon's Silicon Forest, a group of veteran chip architects has traded the security of a tech giant for the high-stakes world of startups, aiming to redefine the future of microprocessors. The founders of AheadComputing, who collectively spent nearly a century at Intel, are now channeling their expertise into building a new class of CPU, one they believe could upend the industry's established order.

Rather than building on the legacy of x86, the team is betting on RISC-V, which they believe is better suited to the demands of modern workloads, particularly those driven by artificial intelligence and cloud computing.

AheadComputing's founders are focused on a technical challenge that has stymied the industry: the bottleneck in per-core performance as AI and machine learning workloads become ubiquitous.

Even as specialized accelerators like GPUs grab headlines, today's CPUs remain essential for many stages of data processing in AI workflows. Yet most organizations report that their AI applications are hindered by the limitations of current general-purpose processors, particularly in terms of bandwidth and data throughput.

AheadComputing's executive team includes Mark Dechene, Alon Mahl, Jonathan Pearce, Srikanth Srinivasan and Debbie Marr.

The company has developed 64-bit RISC-V CPUs designed from the ground up for enhanced single-thread and multi-core performance, with a particular emphasis on power efficiency, thermal density, and scalability across multiprocessor systems. This approach is a direct response to the needs of AI workloads, which often involve single- or low-parallelism tasks that general-purpose CPUs must handle efficiently.

The startup's ambitions are supported by $21.5 million in seed funding and a rapidly growing team, which has expanded from four to forty employees in just five months. The technical roadmap includes not only high-performance cores for servers and data centers but also scalable solutions for mobile and edge applications, reflecting the increasingly fragmented ecosystem of modern computing. The company is also actively seeking strategic partners to integrate its technology into a wide array of platforms, from cloud infrastructure to specialized logic chips.

Jim Keller

"AheadComputing is doing the biggest, baddest CPU in the world," Debbie Marr, the company's CEO, told The Oregonian. The leadership team's background at Intel, where they tackled similar challenges at scale, provides them with a unique perspective on how to break through the performance ceilings of traditional designs.

Their move to a startup environment, where they manage everything from technical design to office logistics, has enabled them to operate with speed and focus that were previously impossible in a corporate setting.

The choice of RISC-V is central to this technical strategy. Unlike proprietary architectures, RISC-V's open standard allows for rapid innovation and customization, free from the licensing constraints that have long limited new entrants in the CPU market. This flexibility is critical as the industry shifts toward modular chiplet designs and away from monolithic, one-size-fits-all processors.

While AheadComputing's chips are still in development, the company's technical vision and rapid progress have already drawn industry attention, including the addition of renowned chip architect Jim Keller to its board.

The founders are clear-eyed about the risks, but equally convinced of the potential rewards. With a focus on technical innovation and a willingness to break from tradition, AheadComputing is staking its future on the belief that the next great leap in computing will come from openness, specialization, and speed.

Image credit: The Oregonian

Permalink to story:

Former Intel engineers form AheadComputing to break CPU performance limits with RISC-V design

 

[WhiteLeaff]

Not even ARM after all this time is able to compete with the dominance and broad compatibility of x86 CPUs

 

[noblody]

RISCV first needs to find its niche and then expand it. With process scaling providing diminishing returns minimizing wasted silicon and software translation layers is good for a lot of products. Not every device needs to be compatible with old software.

 

[Cocojambo13]

If Jim Keller is involved, it's gonna be a winner.

 

[bviktor]

Jim Keller is one of those guys who single-handedly progressed human technological advancement by years, if not decades. If you're a self-aware person, it must be a scary position to be in.

He was the lead architect of the AMD K8 microarchitecture[3][4][5] (including the original Athlon 64)[3][6][7] and was involved in designing the Athlon (K7)[5] and Apple A4/A5 processors.[3][8][9][10] He was also the coauthor of the specifications for the x86-64 instruction set[8][11] and HyperTransport interconnect.[3][11][12] From 2012 to 2015 he returned to AMD to work on the AMD K12[13] and Zen microarchitectures.[14][15]

And that's just the intro. Every few years he comes forward and just happens to throw together a revolutionary new architecture. Like wtf man.

If he's working on RISC-V now, that's a pretty good sign that we can hope for great advancements in the coming years. Fingers crossed.

 

[AnilD]

Kind of wild that we’ve reached a point where general-purpose CPUs are the bottleneck in AI workflows again. Veteran Intel engineers betting on RISC-V and ditching corporate inertia is exactly the kind of disruption the CPU market needs. If they can deliver even half of what they're aiming for, it could really shake things up.

 

[GeoffreyA]

If the industry must abandon x86, which remains to be seen, it's best to jump to the open-source RISC-V than a half-hearted step to proprietary ARM, itself carrying legacy weight. Unfortunately, the defining factor will be Windows, and Microsoft has invested a lot of effort in ARM: in its toolchains, ARM is on the same footing as x86, and it takes a mere recompile to make an ARM binary.

 

[Loadedaxe]

The addition of Jim Keller to their board is a significant endorsement.

Keller, as we all know, is known for his work on groundbreaking processors at AMD, Apple, and Intel, and that brings a wealth of expertise that could accelerate AheadComputing’s development.

His involvement alone brings a promising future for RISC.

This is exactly what is needed to shake things up!

 

[SRB]

Between the article and the comments, I though I was reading a press release for a new chip. Turns out it's a group of engineers from Intel think they can make RISC-V the latest universal processor that can do anything and fits any workload imaginable. And for the cherry on top, one of the most brilliant CPU engineers of recent memory has just joined the company. Not as the lead engineer, but.....wait for it....a member of the board of directors.

You'll pardon me if I don't see this as the next Nvidia rocketing to the top of the stock market.

The list of chips falling on their face (SGI-RISC, DEC Alpha, Motorola 68k, Transmetta, etc) far exceed the number of chips that have hung on through the decades (ARM, AMD, Intel, IBM).

Look at the bright side, there is never a shortage of companies with the next big idea. Maybe this, or another one, will actually succeed.

 

[daffy duck]

Jim Keller is one of those guys who single-handedly progressed human technological advancement by years, if not decades. If you're a self-aware person, it must be a scary position to be in.



And that's just the intro. Every few years he comes forward and just happens to throw together a revolutionary new architecture. Like wtf man.

If he's working on RISC-V now, that's a pretty good sign that we can hope for great advancements in the coming years. Fingers crossed.

And good old Intel has dumped Keller's "Rentable Units" concept which would have eliminated the idea of P and E cores and had a single core that could be run in any manner deemed fit for the purpose essentially.