The lack of vector support really shows, as most geekbench benchmarks use SIMD, but MT performance is quite good.
brucehoult · 1d ago
> The lack of vector support really shows, as most geekbench benchmarks use SIMD
Despite that it's beating the newer gen 1.8 GHz Pi 4 (my Pi 4 is 1.5 GHz) single-threaded on everything except Image Compression and Machine Learning ... nice.
If the SG2380 had happened (US sanctions prevented it, for those who don't know) we should have a RISC-V board beating the Pi 5 by now, and be less than two years behind Arm. (or three years for RK3588 Rock 5)
I think we'll have something competitive with Radxa Orion O6 in the next 12 months, only 1 year behind -- and ARMv9 and RVA23 are *really* comparable in features.
evanjrowley · 1d ago
With an adapter that splits the M.2 into multiple SATA connections, this might be a great Mini-ITX motherboard for some NAS hardware that I've been hoping to revive.
vs 1.8GHz SiFive P550: https://browser.geekbench.com/v5/cpu/compare/23667112?baseli...
Is about 30% faster ST, but a lot faster MT. It's 400% faster on the MT clang compilation benchmark, so this may be a good build server.
vs 1.8GHz Pi4: https://browser.geekbench.com/v5/cpu/compare/23667112?baseli...
vs 2.4GHz Pi5: https://browser.geekbench.com/v5/cpu/compare/23667112?baseli...
The lack of vector support really shows, as most geekbench benchmarks use SIMD, but MT performance is quite good.
Despite that it's beating the newer gen 1.8 GHz Pi 4 (my Pi 4 is 1.5 GHz) single-threaded on everything except Image Compression and Machine Learning ... nice.
If the SG2380 had happened (US sanctions prevented it, for those who don't know) we should have a RISC-V board beating the Pi 5 by now, and be less than two years behind Arm. (or three years for RK3588 Rock 5)
I think we'll have something competitive with Radxa Orion O6 in the next 12 months, only 1 year behind -- and ARMv9 and RVA23 are *really* comparable in features.