TalkChess.com

Leela Chess Zero 42518 vs Stockfish 100619

Hardware 2950x, RTX 2080 ti
Stockfish 32 Thread, 4 Gb HT
Leela Chess Zero 2 threads, 9900000 NN CS.
6 man Tablebase
12 move opening book
TC= 1m/40
Ponder = On

System Specs Ponder On:
Power Draw = 700 Watts
All Cores 4.225 Ghz
CPU Temp 65 c
GPU Temp 68 c
Dimm Temp 56 c
VRM Temp 54 c
Motherboard Temp 41 c
CPU Usage for Stream 0.2% (hardware encoder)
970 Evo Pro Hash Table read speed 72,913 MB/s

Live Steam: https://www.youtube.com/watch?v=pieHydYLOzI

Thanks. Do you notice any performance decrease during the game?

Nordlandia wrote: ↑Tue Jun 11, 2019 9:07 pm Thanks. Do you notice any performance decrease during the game?

Nope.

And both engines run like a champ as you can see in the video. And very comparable results. Running on ponder is the strongest setting on my system. Since for both engines the performance hit is nothing, and the engines can ponder on each others moves.

And it is on video for anyone to inspect.
I have thousands of games on video for you to compare.

The funny thing is people who test with their background programs still enabled are causing a much bigger hit to Stockfish... My idle cpu % is less then 1%.

The biggest issue would be both programs hitting the hash tables at the same time. But with my M.2 drive. This is not a bottle neck, and I have no slow down running engines in ponder mode.

And memory band width is not a issue with 8 sticks of quad channel memory.

And I use a hardware encoder to stream. So no performance hit when streaming games.

It probably do not make much sense. But it is possible to play with ponder over the internet so one can fully utilize RAM without any interference. As long you got playchess subscription. Lending engine as private user is possible. Move overehead parameter have to be adjusted to compensate for the transmission delay or ping. We can play if you got time one day.

https://en.chessbase.com/post/how-to-se ... oud-engine