Windows performance on 8-node machine about asmfish

[tthsqe12]

there are a myriad of hypotheses

1. march stockfish has decreased nps from jan stockfish
2. cfish on your machine may use lzcnt instead of bsf
3. lzcnt may have a higher throughput but similar latency
4. asmfish is tuned for single-threaded performance and might take greater advantage of the OoOE capabilities of your processor
5. bulldozer's OoOE might suffer when saturated with threads
6. asmfish's funny way of updating hash entries might not play nice on your machine with many threads
7. cfish may have a magical chakra alignment

I don't really see this as addressable by me as I only have a lowly 4700 and no need to upgrade.
Once you have addressed 1, you could see if 2+3 or 4+5 is the issue by running both engines with 32 threads and 4 threads on each node pair. As far as I know, Cfish has this capability. In asmfish, it is
setoption name nodeaffinity value 0 2 4 6
setoption name threads value 32
isready
if nodes 0 and 1 are the fused bulldozer cores, 2 and 3, ect.
Also, if you send me the executable you are using, I can check if 2 is true.

[CoffeeOne]

Thx for answering, I wrote a response to your private email address, maybe we can arrange a teamviewer session.