10:46:03
sech1:
looks like 256-bit vector width is kind of a standard for RISC-V CPUs with vector support: https://cdn.mos.cms.futurecdn.net/waPNfjysh8LBEhHraZZiTh.png
10:46:13
sech1:
So I'll target 256-bit width in all my RandomX code
10:48:39
sech1:
Btw that CPU is designed by Jim Keller (AMD Zen, Apple M series)
10:51:53
sech1:
wow, even monsters like this exist: https://p2pool.io/u/37d40fdffe1ab072/image.png
10:51:57
sech1:
512 bit wide vector unit
10:52:04
sech1:
and 4 MB L3 cache per core
10:52:11
sech1:
will be a monster for RandomX
14:31:21
hyc:
fun stuff
20:51:23
jpk68:
sech1: Which RISC-V board are you using? Orange Pi?
21:03:36
DataHoarder:
jpk68: 15:23:59 <@sech1> Got a new toy to play with https://p2pool.io/u/8f4ac95d0af11879/20251120_152219.jpg
21:03:40
DataHoarder:
Orange Pi RV2
21:08:14
jpk68:
Thanks. Missed that photo somehow
21:23:11
jpk68:
Unfortunately it seems pretty hard to find RISC-V boards with reliable hardware AES. IIRC the performance gain from the Raspberry Pi 4 to 5 (which added hardware AES) was fairly substantial.
21:24:46
sech1:
I'm working on writing vectorized software AES for RISC-V, it should be much faster than the regular code.
21:25:25
jpk68:
Oh, cool
21:31:44
jpk68:
Is AES used for anything other than filling/hashing the dataset and scratchpad? I'm not too knowledgeable about what takes the most time per hash
21:34:03
jpk68:
I know this is a terrible comparison, but the Pi 5 is around 5x faster than my RISC-V board, which also has 4 cores.
21:34:09
hyc:
fAES is used in multiple parts of the algorithm
21:52:24
sech1:
soft aes is around 30% of RandomX hash time
21:52:39
sech1:
so if I make it 2x faster, hashrate will increase by ~15%