> I wonder how a truly open project like that could be affected by geopolitics interests, and also, if it would really be effective.
Or in this case, the other way around: how open ISA may affect geopolitics.
RISC-V being free-for-all is a given at this point. But politicians-being-stupid also is. So who knows what they come up with.
Export stops on specific high-perf implementations won't do anything. China would just license from elsewhere, or make their own (which they are perfectly capable of doing).
Not to mention it works both ways: what China contributes to the RISC-V ecosystem, also benefits US interests. Frustrating that with export bans, just creates a lot of hostility for near-0 gain. Read: pointless.
Don't forget the Sophon SG2380 using 16 SiFive P670 cores, with test chips due in probably September or so. That's equivalent to Arm A78, so a step or two ahead of the RK3588 and Pi 5, as well as have more cores.
No Chinese chip using the SiFive P870 (currently their highest end core, with a SPECInt2k6 of around 18/GHz) has been announced, but that doesn't mean a license agreement doesn't already exist. That's up in Arm Cortex X3 territory.
For a v3 tapeout there is certainly still a lot of work to do.
But the latest simpoint SPECint2006 rtl simulation [0], reported 44.08@3GHz (14.7/GHz), which is already faster than the P670 (12/GHz). Most of the v3 work seems to now focuses to the rvv implementation, which wasn't used in the SPEC run.
The HPCA'24 tutorial reported a similar number, but also a SPECint2006 peak score of 49.96@3GHz. [1]
It seems like if the US government had jumped on this particular ball maybe 18 months to two years ago, they might have had a decent shot of making this work.
As things stand internationally now though, it's probably a pretty long shot.
Who knows... with the right pressures in the right places, they could make it happen I guess.
The foundation moved their headquarters to Switzerland exactly to prevent this kind of actions, the only thing US can prevent is import of any China made product, and there is plenty to chose from, not only CPUs.
12 comments
[ 3.5 ms ] story [ 42.7 ms ] threadI wonder how a truly open project like that could be affected by geopolitics interests, and also, if it would really be effective.
Or in this case, the other way around: how open ISA may affect geopolitics.
RISC-V being free-for-all is a given at this point. But politicians-being-stupid also is. So who knows what they come up with.
Export stops on specific high-perf implementations won't do anything. China would just license from elsewhere, or make their own (which they are perfectly capable of doing).
Not to mention it works both ways: what China contributes to the RISC-V ecosystem, also benefits US interests. Frustrating that with export bans, just creates a lot of hostility for near-0 gain. Read: pointless.
All Chinese RISC-V core designs have been proprietary designs thus far.
Because RISC-V ISA is open standard, there's nothing else the US government can control.
Maybe you mean Chinese-made, rather than proprietary?
As many open hardware RISC-V core designs come from China.
The JH7110 SoC from the Chinese firm Starfive uses SiFive's U74 core. Eswin, also Chinese uses SiFive's P550 core in their upcoming EIC7700 SoC.
> All Chinese RISC-V core designs have been proprietary designs thus far.
There is the OpenC910 [1] and OpenXiangShan [2].
[1] https://github.com/T-head-Semi/openc910
[2] https://github.com/OpenXiangShan/XiangShan
No Chinese chip using the SiFive P870 (currently their highest end core, with a SPECInt2k6 of around 18/GHz) has been announced, but that doesn't mean a license agreement doesn't already exist. That's up in Arm Cortex X3 territory.
The 2nd gen Nanhu, taped out in November so maybe sampling about now, is supposed to be in 2.0 GHz A76 / P550 range.
Maybe 3rd gen might be around P670, but that's some years away from being in the market.
The HPCA'24 tutorial reported a similar number, but also a SPECint2006 peak score of 49.96@3GHz. [1]
[0] https://www.zhihu.com/people/openxiangshan?utm_id=0 (click on the first post and scroll down till you see the tabel)
[1] https://raw.githubusercontent.com/OpenXiangShan/XiangShan-do...
As things stand internationally now though, it's probably a pretty long shot.
Who knows... with the right pressures in the right places, they could make it happen I guess.
The PRC joined the WTO in 2001. That was the ballgame.