There is a laptop that has neither Intel nor AMD nor Snapdragon inside. DeepComputing's DC-ROMA runs a RISC-V processor on Framework's modular platform and promises 40 trillion AI operations per second without leaving the device, according to the spec sheet published by the manufacturer. The machine ships with Ubuntu 24.04 and targets developers who want to tinker with the open architecture without giving up a real computer.
The date matters and it is worth making it clear from the start. The device launched in 2025 and the outlets that covered it, such as DailySynapse in October of that year, reported immediate availability on the same base that is sold today. What changes now is the software, because Ubuntu 26.10 comes out tomorrow, October 15, with official desktop images for RISC-V, as Canonical announced on its forum on October 9. The hardware was waiting for its moment and the system has finally arrived.
That combination changes the calculation for anyone who programs. Until now, using RISC-V on the desktop meant settling for server editions or unofficial builds maintained by the community, whereas from tomorrow there is an image with GNOME and a lightweight one with Xfce aimed at virtual machines. The DC-ROMA stops being a collector's whim and becomes a viable workstation for local AI and open development.
A Framework laptop with a RISC-V heart
DeepComputing builds the DC-ROMA on Framework's modular platform, the brand that popularized repairable laptops with interchangeable parts. The buyer gets a RISC-V mainboard that fits into the Framework Laptop chassis.
That board inherits the screen and keyboard along with the chassis battery. The decision reduces the risk of the experiment, because if the board falls short the chassis accepts other boards in the future. The device is presented as the first RISC-V-based AI PC designed for developers.
The brain is the ESWIN EIC7702X, a system on chip with eight SiFive P550 cores for general workloads. The P550 are 64-bit cores with performance per watt that already proved its worth on earlier development boards.
Alongside them travels an eight-core vector cluster to accelerate parallel computing and an engine that encodes 8K video at 50 frames per second. Memory supports up to 64 GB of LPDDR5 and storage reaches up to 1 TB in NVMe, figures for a serious laptop.
The bet has a meaning that goes beyond the spec sheet. RISC-V is an open architecture, with no licenses that pay a toll on every chip, and that attracts those who research or teach processor design.
Programming on open silicon allows inspecting, modifying and redistributing with a freedom that proprietary architectures do not offer. The DC-ROMA puts that silicon inside a chassis that anyone can open with a screwdriver.
40 TOPS of AI without leaving the device
The device's other argument is its 40 TOPS neural unit, that is, 40 trillion operations per second dedicated to inference. That power makes it possible to run language and vision models directly on the local machine, without sending data to the cloud or paying for every query.
For a developer it means prototyping assistants and pipelines on the train, at a client's home or in a lab without a connection. Privacy stops being a slogan and becomes architecture, with the latency that comes from not depending on any remote server.
CEO Yuning Liang sums it up as a bet on local AI, sustainable computing and openness. The company describes the device as a foundational step toward a future built on privacy.
The pitch sounds ambitious, though it fits a real niche of programmers who distrust depending on remote services for every experiment. That audience prefers to pay once for hardware rather than subscribe for life to an API.
The figure should be put in its October 2026 context. The 40 TOPS of the original model fall below the accelerators of the latest x86 and ARM generations, and DeepComputing itself already showed a 50 TOPS variant in May 2025.
Nobody buys this device to win a benchmark. It is bought to have usable AI on an open architecture, and in that field it has hardly any direct competition among the laptops sold today.
Ubuntu 26.10 debuts an official desktop for RISC-V
The twist this week comes from Canonical. Ubuntu 26.10, codenamed Stonking Stingray, comes out tomorrow, October 15, with desktop images for RISC-V for the first time in the distribution's history.
Until now there were only desktop images for 64-bit Intel and ARM, while RISC-V made do with server editions. The announcement was published by the Foundations team on the Ubuntu forum on October 9.
There are two experimental images available to try right now. The first is the classic Ubuntu Desktop with GNOME, which works well on the SpacemiT K3 Pico-ITX board that Canonical uses as a reference.
The second is a minimal Xubuntu with Xfce, aimed at those testing in virtual machines, because the GNOME desktop takes more than half an hour to boot under QEMU emulation. Both still lack official support and carry known bugs.
The following table summarizes the DC-ROMA hardware alongside the software that accompanies it this week, with the source backing each data point so you can check it before deciding on any purchase.
| Part |
Detail |
Source |
| Processor |
ESWIN EIC7702X with 8 SiFive P550 cores |
DeepComputing |
| AI acceleration |
40 TOPS NPU on the local machine |
DeepComputing |
| Memory and storage |
Up to 64 GB LPDDR5 and 1 TB NVMe |
DeepComputing |
| Installed system |
Ubuntu Desktop 24.04 LTS |
DailySynapse |
| October 15 novelty |
RISC-V desktop images in Ubuntu 26.10 |
Canonical |
| Reference board |
SpacemiT K3 Pico-ITX with official support |
Canonical |
The calendar could not be rounder for DeepComputing. Its device ships with Ubuntu 24.04 LTS, a stable base with years of support ahead, and right now a new version appears with a native desktop for its architecture.
Users will be able to choose between the solidity of the LTS and the novelty of 26.10 depending on their risk tolerance. That is exactly the flexibility this device's audience demands, accustomed to compiling and testing.
Who this device makes sense for
The DC-ROMA does not compete with the office laptop and should be judged by its own yardstick. Its natural audience is the developer who compiles for RISC-V and needs to test on real silicon instead of settling for slow emulators.
Having the chip on the desk speeds up the trial-and-error cycle, and that is worth more than a few extra points in any comparison. The laptop format also adds the possibility of taking demos to talks and trade shows.
The second profile is the local AI researcher. Anyone working with sensitive data, such as medical records or prototypes under a confidentiality agreement, cannot upload anything to an external API or trust deletion promises.
A 40 TOPS NPU makes it possible to run mid-sized models with acceptable latency and without a single byte leaving the disk. Memory of up to 64 GB helps load quantized models of a respectable size.
The third group looks to the classroom and the lab. Universities that teach computer architecture find in RISC-V a platform without industrial secrets, and a complete laptop is more attractive than a bare board with cables everywhere.
Students compile, run and measure on a device that also serves for everyday use. That versatility justifies the investment better than any isolated development board that ends up in a drawer.
- Developers who compile and test software on real RISC-V silicon in a laptop format.
- Researchers who need local inference for privacy, with up to 64 GB for quantized models.
- Teachers and students of computer architecture who want an open platform without loose parts.
- Enthusiasts of repairable hardware who value Framework's modular chassis and its interchangeable parts.
The price of being a pioneer comes with fine print
No frontier comes free and neither does this one. The application ecosystem for RISC-V is still young, and although Debian and Ubuntu are moving fast, proprietary software pieces that are taken for granted on other architectures are missing.
The buyer must accept that some tools will arrive late or require compiling from source code with their own patches. That friction is part of the deal and should be known before paying for the device.
The price also calls for a second thought. The second revision of the board started at 349 dollars in the manufacturer's store in May 2025, a reasonable figure for a pioneering board but high compared with a conventional laptop of similar power.
Add to that the Framework chassis if one is not already owned. The whole comes to a range that is only justified by the value of the experiment and by the taste for breaking new ground.
DeepComputing itself acknowledges the product's early character with its focus on developer communities. The first units went to that loyal audience, with the promise of documenting and improving along the way.
Anyone looking for a device to work on tomorrow without surprises has had more mature options in x86 and ARM for years. Anyone who wants to push open architecture will find here their best laptop ally.
RISC-V stops being a promise on the desktop
The DC-ROMA story sums up where RISC-V stands in October 2026. Open hardware now fits in a repairable laptop, local AI now runs without the cloud and the leading operating system now publishes a desktop for the architecture.
Each piece arrived separately and this week they fit together for the first time. The 2025 launch was the announcement and Ubuntu 26.10 is the piece that was missing to round out the proposal.
The road ahead is measured in software. Polished drivers are missing, native applications are missing and the big studios still need to optimize their engines for these chips. But the direction is clear, because every cycle brings more compiled packages and more powerful boards. The DC-ROMA is nobody's final destination, though it is the most fun vehicle for traveling the road.