[camel-cdr]

The SpacemiT K3 with 8 SpacemiT X100 RVA23 cores, which are faster than Pi4 but slower than Pi5, should be available in a couple of months:

geekbench: https://browser.geekbench.com/v6/cpu/16145076

rvv-bench: https://camel-cdr.github.io/rvv-bench-results/spacemit_x100/...

There are also 8 additional SpacemiT-A100 cores with 1024-bit wide vectors, which are more like an additional accelerator for number crunshing.

The Milk-V Titan has slightly faster scalar performance, than the K3.

[LeFantome]

> faster than Pi4 but slower than Pi5

It may actually be faster than a Pi5.

The benchmark is well tuned for ARM64 but not so well adapted to RISC-V, especially the vector extensions.

You may still be right of course. The SpaceMIT K3 is exciting because it may still be the first RVA23 hardware but it is not exectly going to launch a RISC-V laptop industry.

[camel-cdr]

There isn't much to tune in some, e.g. the clang benchmark. We know that many of the benchmarks already have RVV support (compare BPI-F3 results between versions) and three are still missing RVV support. I think the optimized score would be in the 500s, but that's still a lot lower than Pi5.

[kombine]

> The Milk-V Titan has slightly faster scalar performance, than the K3.

So the main difference between this Milk-V Titan and the upcoming SpacemiT K3 is that the latter has better vector performance?

[camel-cdr]

The Titan has no SIMD/Vector support at all, so it doesn't support RVA23.

[snvzz]

The K3 is able to run RVA23 code, the Titan is not; it lacks V.

It matters, as the ecosystem settled on RVA23 as the baseline for application processors.