I don't
The bit at 9:20 "the speed at which RISC-V has advanced in the past three years"?
It's not fast enough to bring RISC-V to our desktops within the next few years. I hope I'm wrong but it's just painfully slow compared to past ARM development.
this post was submitted on 20 Apr 2025
80 points (94.4% liked)
Linux
7072 readers
1207 users here now
A community for everything relating to the GNU/Linux operating system
Also check out:
Original icon base courtesy of [email protected] and The GIMP
founded 2 years ago
MODERATORS
What is stopping people from bringing RISC-V to the desktop now? Major distros already support it and you can run x86 programs with box64.
What is not fast enough then?
Two things:
-
Desktop requires mature CPUs (large out-of-order designs with high IPC) and there just aren't really any of those yet. They're starting to arrive (e.g. XiangShan which is even open source!) but as far as I know there isn't a single chip available to buy that's faster than a Raspberry Pi 4.
-
Microcontrollers can get away with only the basic instruction set (add, multiply, load, store etc.) but for high performance you need a ton of extensions that are considered standard. x86 and ARM have had decades to build them up but in RISC-V a lot of them are only recently ratified (e.g. Vector) or still in the process of being defined.
I would say we might see cheap Android phones with RISC-V CPUs in maybe 5 years. Though there's an additional difficulty there in that you need to emulate ARM for games, and I don't think anyone is working on that.
SiFive P670
From what I can tell this might be almost as fast as a RPi 5 (single core). Which is almost as fast as my 12 year old i5-2500K. I guess we'll find out when it is available.
I definitely think we'll get an M1/Zen class RISC-V CPU eventually but I doubt this is it.
It's not even remotely competitive on power vs. X86 processors.
Decent enough processors, AFAICT.
Nothing that matches my i5-8350
Well yes the peformance ceraintly hasn't caught up yet to x86 but the strongest riscv cpu on the market as far as I know has 64 cores on 2ghz. More then enough to run a desktop.
2ghz does not measure it's computing power though, only the cycle speed. Two very different things.
An objective measure is a simple benchmark:
Here's a quad core 1.5ghz RISC-V SoC (noted as VisionFive 2) vs a quad core 1.8ghz ARM chip (noted as Raspberry Pi 400).
It's not even remotely close to usable for all but the most basic of tasks https://www.phoronix.com/review/visionfive2-riscv-benchmarks/6
The best R5 SoC is about as fast as a Pi 4 and better in many ways but also much more expensive.
https://www.eswincomputing.com/en/bocupload/2024/06/19/17187920991529ene8q.pdf
R5 is improving faster than ARM. There are more companies designing R5 chips than ARM. The R5 software ecosystem is essentially ready and waiting.
For many workloads, the GPU or DSP is more important than the CPU. R5 is becoming viable for these use cases.
Automotive, automation, quality control, robotics, aI, are all within reach. The SBC market is just the mainstream version of that. And desktops are just further along the price / performance curve from there.
Yeah, I've been watching. I'm waiting with baited breath when I can start using the machines for day to day. Not what I'd use a SBC for. I guess when there's a ThinkPad that has R-5 I'll really take notice.
Framework sells a laptop with an R5 board. They warn it's intended for developers and slower than a Raspberry Pi though.
Yeah, that's what I said.
Chip designs take time. Then people need to license and manufacture them. We may see marketable performance on servers this year.
For SBCs, the performance has gotten to usable but price / performance sucks. That is a bit of a chicken / egg popularity problem so timing is tough to call. The rift between the US and China is slowing things down. We would have the Milk-V OASIS otherwise.
Desktop is really tough to call timing. The tech could probably be there next year. As ARM is showing though, you need a desktop OS (with market share) to drive that market. It is not going to be Apple. Microsoft cannot even make ARM work. So desktop Linux hardware on RISC-V may be a while.
Some Android phones and tablets could go RISC-V in 2026. If that happens, the same chips could appear on ITX boards for enthusiasts.
Qualcomm could surprise with RISC-V support after what ARM did to them. AheadComputing or somebody else could surprise as well. Mostly likely though, it is just going to take time.
You can run RISC-V on a “desktop” today if you want . Grab a ROMA II or Framework 13. Expect it to be slow.
Eventually, yes. But it's risky to put timelines on things.
CPUs are already not the limiting factor. With enough cores, mobile CPUs are fine for most situations. It's GPU speed that affects most people, and that's because of games.
There are 4 main bottlenecks in computers, and they generally take turns being the most relevant. CPU, GPU, RAM, and storage. Bus speed can also be a bottleneck, but that is generally factored in and we know how to make faster buses for the most part, using parallelization if nothing else.
Right now, for home computer use, GPU is the biggest factor. Good thing, too, because CPUs are plateauing, and will probably require a fundamental change in architecture or programming techniques to get past it.
You still need base CPU speed for a system to be usable. Try running a modern GPU on a 10 year old CPU. It's even worse for some, where the GPU driver needs a relatively fast CPU for the GPU to run at full speed. Mostly Intel GPUs have this issue, which is sad cause they are the most affordable, but can't be paired with an just an affordable CPU (or an older one).
And we're very far away with RISC-V from the kind of performance your need to run modern games, or even decade old games. Let alone fully utilizing a high end GPU.
"Most people" do not play video games at all.
And most people don't complain about computers being slow anymore. And when they do, it's usually because of memory, disc, or network speeds. It's almost never because of CPU cycles. The people complaining about performance that's related to cycles are usually complaining about GPU processing.
It's almost never a CPU power issue, anymore. Unless you're a developer or scientist, and you're actually trying to compute something. I have two beefy computers in my house - my desktop, for coding, and my media server, because Jellyfin insists on transcoding everything. The rest are all ARM, and mostly old ARM, and they're all perfectly capable of doing their jobs. RISCV would be, too.
Looking forward to playing Dosbox on it
Fingers-Crossed