AMD Zambezi news, info, fans !

[dess]

It does have a buffer so they are not stuck by the write speed.

It depends on the BW and size of this buffer...

And the l1 write is correct because you expect it to be in the facinity of the l2cache.

I think SW like AIDA64 could make a distinction of "short" (fitting in the WCC) and "long" (not fitting in the WCC) writes here. (I've noticed the authors about it.)

These results seem more like a read/write through up to the memory. So either a hardware or an Aida64 software issue.

Isn't it like to the L3, instead? Memory BW is roughly ~15 GB/s.

Anyway, the authors don't think it's AIDA64's fault. It could be a HW problem, but possibly they din't have had the latest AGESA when they've done tests on actual HW.