Europe bets on RISC-V for homegrown supercomputing platform
-
Give me something like Talos2 with a full OSS firmware and a performant CPU... and hell, a half-competitive open source graphics core too. It doesn't need to be peak performance, it needs to be good enough.
I've been trying to work with SBC's for a while for video decoding platforms and just wound up getting stuck on x86 because the ARM situation with weirdo custom kernels for anything useful is just... annoying.
-
it's complicated. afaik asml has agreements with the us govt, and cross licensing with american companies. also, asml only makes lithography tools, there's a LOT more to making semiconductors than just exposing patterns. and a few of the biggest vendors like kla and amat are american. kla in particular is essentially a monopoly in the metrology space.
-
ARM was bought by the Japanese, it's no longer European. RISC-V is the future.
-
Considering that you can buy some Raspberry Pi micro computers (these are ARM architecture computers) for less than €100 that are performance competitive with a lot of existing hardware; this idea would make a ton of sense for Europe to implement. I think Europe could probably start designing and manufacturing chips locally within 2 to 5 years on the low end 5 to 10 years on the high end.
ARM and RISC are not equal. The fastest current RISC CPU is an absolute potato. Then you've got ARM-based chips way faster than a Pi. Then there's silicone like the M4. It's a big uphill for RISC, which is why this, and the investments from the Chinese, are good but longer-term plays.
-
ARM and RISC are not equal. The fastest current RISC CPU is an absolute potato. Then you've got ARM-based chips way faster than a Pi. Then there's silicone like the M4. It's a big uphill for RISC, which is why this, and the investments from the Chinese, are good but longer-term plays.
The question should be then what ARM CPU compares to current RISC-V best CPU and see the gap in years.
-
so, I don't know if the shit hole made anything WORTH copying, but why respect american intellectual property? you know americans don't respect yours. copy NVIDIA's CUDA shit, if that's efficient. fuck em.
The efficiency is not on the API it is on the microarchitecture. The value of copying the API is just to run unmodified software made for CUDA.
-
The efficiency is not on the API it is on the microarchitecture. The value of copying the API is just to run unmodified software made for CUDA.
There's also a lot of efficiency in hardware-specific kernels. A generic rocm build vs. one with hand-written kernels (not even for the proper card just a close enough one to have the same instructions) is like a 10x performance drop. That's on the matrix multiply up to convolve these tensors level, on the layer above that you then have things like smart memory management and scheduling as well as minimising how much work needs to be done in the first place (re-ordering operations so tensors stay small) and stuff.
You can port cuda code to vulkan or opencl -- but you're going to have to reimplement all of that. Just getting the BLAS layer to not suck is a challenge.
-
ARM and RISC are not equal. The fastest current RISC CPU is an absolute potato. Then you've got ARM-based chips way faster than a Pi. Then there's silicone like the M4. It's a big uphill for RISC, which is why this, and the investments from the Chinese, are good but longer-term plays.
Not a matter of instruction set, though. Current RISC-V designs are built from scratch by companies pretty much doing their first chip and/or design studios out of the microcontroller space, if say AMD would spend a year slapping a RISC-V insn decoder onto their existing designs that shit would fly.
I guess of the big performance vendors Quallcomm will be first, they have a bone to pick regarding ARM licensing.
-
The question should be then what ARM CPU compares to current RISC-V best CPU and see the gap in years.
This would be hard to quantify. A year of work one year ago would take significantly less time now since the knowledge exists.
-
This would be hard to quantify. A year of work one year ago would take significantly less time now since the knowledge exists.
I'm talking about "X OSS software in version Y needed Z second to do this precise job". Not "compare MFLOPS".
-
System shared this topic onSystem shared this topic on
