CPU: 2x AMD EPYC 9754 (256 cores/512 threads)
GPU (für Lc0): 8x NVIDIA GeForce RTX 5090
Of course this don't change anything. Stockfish wins about 57 - 43 like previously.
New TCEC hardware for season 28
Moderator: Ras
-
- Posts: 3611
- Joined: Wed Mar 08, 2006 8:15 pm
- Full name: Jouni Uski
-
- Posts: 57
- Joined: Mon Feb 20, 2017 8:29 am
- Location: Rialto, Venice
Re: New TCEC hardware for season 28
Forgive me, but are we moving from premises to conclusions without clear evidence? Was the theoretical speed difference between the old and new hardware factored in, considering the tests currently underway before the championship?
I feel your conclusions might be a bit rushed, based on empirical analysis of the actual improvements.
Maybe we can revisit this later? I ask because I've performed simulations on the real nodes involved, both pre - and post-upgrade.
This approach is serious, only empirical approach on node numbers
Not yours. My opinion on real data is that Leela scales much better than Stockfish and tends to level the gap. New hardware performs better 3x as regards CPU but 4/5 x as regards GPU.
Now is up to you.
I feel your conclusions might be a bit rushed, based on empirical analysis of the actual improvements.
Maybe we can revisit this later? I ask because I've performed simulations on the real nodes involved, both pre - and post-upgrade.
This approach is serious, only empirical approach on node numbers
Not yours. My opinion on real data is that Leela scales much better than Stockfish and tends to level the gap. New hardware performs better 3x as regards CPU but 4/5 x as regards GPU.
Now is up to you.
-
- Posts: 1099
- Joined: Fri Sep 16, 2016 6:55 pm
- Location: USA/Minnesota
- Full name: Leo Anger
Re: New TCEC hardware for season 28
You are right. The Komodo people have shown me/us data that CPU chess engines don't scale well past 30 cores. Its great though that the tournament is upgrading its hardware again.
Advanced Micro Devices fan.