Ubuntu is one of the leading Linux distributions for RISC-V hardware thanks to Canonical’s strategic partnerships with companies like DeepComputing – who just announced a powerful new RISC-V AI PC running Ubuntu 24.04 LTS.

The DC-ROMA RISC-V AI PC—apologies for the caps, it’s how it’s stylised—is built around the company’s new RISC-V Mainboard II, which is designed for use in the Framework 13″ and 14.2″ laptops.

Though designed for Framework laptops, owning one isn’t a requirement. A nifty enclosure allows this mainboard to be used as a regular PC you connect to a monitor, keyboard and mouse.

The board itself uses ESWIN’s advanced RISC-V AI SoC EIC7702X—sorry for even more caps, it’s how it’s stylised—which is a 64-bit RISC-V processor with 8 SiFive P550 ‘out-of-order’ CPU cores.

What is RISC-V? It’s an open-source instruction set architecture (“processor”) that anyone is free to build, improve, share and sell without having to pay licensing fees or royalties. RISC-V is to ARM and x86 platforms what Linux is to Windows and macOS.

There’s an integrated Imagination GPU with support for up to 8K @ 50FPS video encoding, an NVMe SSD slot, and up to 64GB of RAM (not user upgradeable). Framework’s swappable expansion modules make it trivial to add whatever ports and connections you need.

Ubuntu + RISC-V = TOPS Team

Canonical is understandably upbeat about the arrival of more RISC-V hardware it can support:

Canonical is proud to be part of this important step for the RISC-V ecosystem. Together with DeepComputing, Framework, ESWIN, and SiFive, we’re enabling developers to build next-generation AI solutions on RISC-V leveraging Ubuntu and open source software components.

Director of Silicon Alliances at Canonical, Gordan Markuš

Nirav Patel, Founder and CEO of Framework, is equally pleased, noting that he’s happy to see DeepComputing “bringing a substantially more powerful RISC-V processor to developers” via its modular and upgradeable Framework device ecosystem.

DeepComputing launched the first-generation RISC-V mainboard for the Framework laptop last year, costing from $199.

While capable, RISC-V performance remains behind that of mature processor architectures, including ARM, and software support in general remains formative. Developers, enthusiasts and tinkerers may enjoy the challenges and exploit the potential. End users less so.

A Game of TOPS Trumps

Of its AI PC, DeepComputing say is is “built for developers pioneering edge and AI-native applications, it delivers over 40 TOPS of local AI compute, enabling complex AI models—such as large language models (LLMs)—to run entirely on-device, without relying on the cloud.”

“40 TOPS”?

DeepComputing and Canonical do lead on this delivering up to “50 TOPS” of AI-assistive number-crunching, so presumably it can reach that.

Btw, TOPS (or Trillion Operations Per Second) is the number of calculations a chip can perform for AI and machine-learning related jobs. The higher the number, the better it is (in theory1) for running things like LLMs, handling inferencing, etc locally.

The spec sheet, however, does say the NPU tops out (heh) at 40 TOPS — which is more (theoretical) power than the neural engine in Apple Silicon M4 chips (38 TOPS), but 10 short of the headline claim.

The other compute likely comes from a combination of GPU (listed as offering 2 TOPS), the CPU and/or the 512-bit-wide vector processor.

It’s a shame the AI-assisted press release and promo pages for this don’t spell things out more clearly (for a brain like mine). They do promise ‘seamless’ this and ‘effortless’ that so, y’know, why worry about hardware specs when you’re spending hundreds of dollars! </snark>

Where to buy the DC-ROMA RISC-V AI PC

Pricing starts at $349 for the mainboard with 32GB RAM, enclosure and various modules, through to $1,099 for a mainboard with 64GB RAM, 1TB SSD and Framework 14.2″ laptop to put it in.

You can pre-order the—here come the All Caps—DC-ROMA RISC-V AI PC right now, from the official DeepComputing store. Shipping begins “Q3 2025” – but beware of tariffs or taxes as it comes from China.

Anyone heading to the RISC-V Summit Europe or Computex Taipei 2025 will be able to visit the DeepComputing booth to get a hands-on experience with the AI PC running Ubuntu 24.04.

  1. Getting decent performance from an NPU relies on more than just the NPU being present. Solid, well-optimised software, APIs, drivers and other ‘boring’ bits are necessary. ↩︎