The performance here is less interesting than the apparently quasi imminent availability of real RVA23 devices. If/when that occurs it will set a baseline and things should not warrant from scratch rebuilds constantly for at least a while.
Anyone that believes semiconductor performance claims before device availability is in for a rough time.
Manufacturers don't lie about how many clock cycles Dhrystone or Coremark or SPEC takes on their design.
The first two can be checked at the verilator (or commercial equivalent) stage, and SPEC in FPGA, both long before physical chips.
The only things in doubt are:
- whether they hit the hoped-for MHz
- whether the code you want to run has the same characteristics as their benchmark
For example SG2044 GeekBench results have been appearing on the site from time to time for more than half a year now. GB as a whole is pretty much worthless for what I care about, but the Clang test within it aligns closely with what I do, including the rqtio between multi-thread and single-thread.
You are right, but I think this is different than say performance claims for the next generation by mainstream GPU or CPU manufacturers.
This is a chicken and the egg problem. You need interest in the architecture and belief in performance for device availability to start to pop, after all who would invest a lot of money in something they don't think will have a market, but you also need device availability to support performance claims.
So I think for newer products, performance claims will have to do.
> You'll be able to buy RISC-V at that level sometime next year.
Will you? What exactly are Tenstorrent's plans for Ascalon? They are chasing the AI market, and I haven't seen any clear indication that they intend to sell their RISC-V CPUs as unbundled SKUs, or on "main computer" PCBs (ATX/mATX/ITX/etc.) I'm not sure they even care about that market.
I hope they'll do this, but that's all I have right now. If you know something in particular, please share. If I could buy a M1 or better class Ascalon desktop system I'd be happy to pay an Apple-esq premium, and I believe many others would as well.
Tenstorrent have clearly said, I think in a presentation at a recent RISC-V Summit, that they want to see wide distribution of Ascalon to as many users and developers as possible, including laptops, to help kick-start the mainstream RISC-V ecosystem.
Not just expensive servers, unlike some of the other high performance CPUs coming soon.
This might be what you have in mind, or similar[1]: RISC-V Taipei Day May 21, 2025. There are some hard numbers on Ascalon performance in there, anyhow.
One figure from that: "36 SPECINT2K17 Rate" for the "8 core Athena Chiplet." That is on par[2] with a Intel Core i5-9600K (circa 2018 CPU.) That's enough for useful workloads. I'm actually still using one for Linux work on a daily basis.
That's about what I would expect: it's in the ballpark of plausibility, given the givens. It's not reasonable to expect parity with Zen 5 or whatever. They'll need some years yet to ramp up their designs.
I excited about Tenstorrent. Keller as the boss: one can hope for great things.
Versus the competition? Modern x86 and ARM chipsets are not only proprietary from the ISA-up, they have the same proprietary extensions.
"RISC chipset adds in proprietary CISC features" has been in my headlines so often that I assumed nobody cared. Helluva time to go up-in-arms, in Q3 2025.
Are you sure about the "slower" part? Because I have both a VisionFive 2 and the OrangePi RV2, and the OrangePi RV2 is way faster in my different tests. Even in single core scenarios.
So is P550 ... 8.65 is the claim I think. It doesn't have the MHz though. My i7-4790K (which is still in the back room somewhere) had a base of 4.0 GHz and could turbo to 4.2 or 4.4. My Megrez is doing just 1.8 GHz.
Main issue with P550 is no V. Huge handicap, can't easily make up for that.
RVA23 requires V. These incoming CPUs are the real deal, a perfectly usable performance level in the present day.
(unless they cache-starve them or some other such fuck up)
My 4790k was still my main PC until I recently (weeks) built a new PC with 9800x3d, 96GB ECC, rx7900gre.
4790k had 4 cores. If these new chips clock reasonably, say, 3+ GHz, and have 8 or 16 cores, they'll easily generally outdo this old -but very usable still- Intel chip.
No V instruction set is going to be a big handicap though, as that's pretty much RISC-V's equivalent of AVX and missing those instructions is going to suck big time.
Pretty much my issue is that the one that is the closest to shipping is also the one missing V instructions. I am far more skeptical over potential processors especially after Horse Creek just vanished into thin air.
Anyone that believes semiconductor performance claims before device availability is in for a rough time.