Code: Select all
Nodes/second CPU / Memory Cores/Threads Extension Member OS
510.152.600 2x Intel Xeon 6 Sierra Forest 1TB DDR5 384threads avx512 TataneSan LIpman has new bench leader!
Moderator: Ras
-
Jouni
- Posts: 3529
- Joined: Wed Mar 08, 2006 8:15 pm
- Full name: Jouni Uski
Ipman has new bench leader!
Jouni
-
Werewolf
- Posts: 1961
- Joined: Thu Sep 18, 2008 10:24 pm
Re: Ipman has new bench leader!
You know what's around the corner, right?
-
jpqy
- Posts: 553
- Joined: Thu Apr 24, 2008 9:31 am
- Location: Belgium
Re: Ipman has new bench leader!
311.761.149 AMD EPYC™ 9754 DDR5 6400 256threads vnni512 denuit L
This is single cpu..when Ipman can get 2x Epyc 9754 it will pass Intel again.
For sure amazing nodes/sec. this 2x Xeon 6 Sierra Forest 384threads getting!
This is single cpu..when Ipman can get 2x Epyc 9754 it will pass Intel again.
For sure amazing nodes/sec. this 2x Xeon 6 Sierra Forest 384threads getting!
-
Jouni
- Posts: 3529
- Joined: Wed Mar 08, 2006 8:15 pm
- Full name: Jouni Uski
Re: Ipman has new bench leader!
New addition: 492.900.873 2x AMD EPYC 9754 ddr5 512threads
In endgames this system has shows 1,2-1,5 Gnps many times
.
In endgames this system has shows 1,2-1,5 Gnps many times
.Jouni
-
Hai
- Posts: 639
- Joined: Sun Aug 04, 2013 1:19 pm
Re: Ipman has new bench leader!
https://ipmanchess.yolasite.com/amd--in ... ckfish.php
It's not that special.
Let's say the result would be even better = 500000 kn/s using Linux.
We will lose probably 10% to 20% when using Windows (ChessBase or Fritz or...) = Between 400000 kn/s to 450000 kn/s.
The Hyperthreading/Multithreading and elo gain problem was also discussed in another topic.
The kn/s increase from Hyperthreading doesn't translate in much Elo gain, because the number of search threads also doubles, leading to redundant searches that increase NPS, but don't actually contribute anything useful.
See: viewtopic.php?p=955593&hilit=hyperthreading#p955593
Intel CPU with Hyperthreading 60000 kn/s and newest Intel CPU which has no Hyperthreading 20000 kn/s. -66%.
A 45000 kn/s Hyperthreading CPU is probably +5 Elo stronger than a 15000 kn/s Hyperthreading CPU.
A 45000 kn/s Hyperthreading CPU is probably +1 Elo stronger than a 15000 kn/s no Hyperthreading CPU.
The AMD with 450000 kn/s Hyperthreading should be still stronger than a 150000 kn/s no Hyperthreading CPU but probably only +0.1 Elo.
Stockfish 14.1 with the ARM speed up code improvements from Stockfish developers should have between 50000 kn/s to 65000 kn/s on Apple M4 MAX MacBook.
So we can say 150000 kn/s are enough to have a fair fight(nearly same strength) vs the 450000 kn/s from AMD and 150000 kn/s is only 2.3 to 3 times faster than Apples M4 MAX MacBook.
It's not that special.
Let's say the result would be even better = 500000 kn/s using Linux.
We will lose probably 10% to 20% when using Windows (ChessBase or Fritz or...) = Between 400000 kn/s to 450000 kn/s.
The Hyperthreading/Multithreading and elo gain problem was also discussed in another topic.
The kn/s increase from Hyperthreading doesn't translate in much Elo gain, because the number of search threads also doubles, leading to redundant searches that increase NPS, but don't actually contribute anything useful.
See: viewtopic.php?p=955593&hilit=hyperthreading#p955593
Intel CPU with Hyperthreading 60000 kn/s and newest Intel CPU which has no Hyperthreading 20000 kn/s. -66%.
A 45000 kn/s Hyperthreading CPU is probably +5 Elo stronger than a 15000 kn/s Hyperthreading CPU.
A 45000 kn/s Hyperthreading CPU is probably +1 Elo stronger than a 15000 kn/s no Hyperthreading CPU.
The AMD with 450000 kn/s Hyperthreading should be still stronger than a 150000 kn/s no Hyperthreading CPU but probably only +0.1 Elo.
Stockfish 14.1 with the ARM speed up code improvements from Stockfish developers should have between 50000 kn/s to 65000 kn/s on Apple M4 MAX MacBook.
So we can say 150000 kn/s are enough to have a fair fight(nearly same strength) vs the 450000 kn/s from AMD and 150000 kn/s is only 2.3 to 3 times faster than Apples M4 MAX MacBook.