Search found 70 matches

by ankan
Thu Nov 15, 2018 6:42 am
Forum: Computer Chess Club: General Topics
Topic: How good is the RTX 2080 Ti for Leela?
Replies: 48
Views: 12613

Re: How good is the RTX 2080 Ti for Leela?

Another thing to notice in this table is that fp16 numbers tend to be >> 2x fp32 numbers. This makes me a bit suspicious. That is because we use Tensor Cores for FP16 path. The speedup is only about 3X (and not 8X as raw TFlops nos would suggest) because fp32 path uses Winograd algorithm for convol...
by ankan
Sun Oct 28, 2018 4:27 am
Forum: Computer Chess Club: General Topics
Topic: How good is the RTX 2080 Ti for Leela?
Replies: 48
Views: 12613

Re: How good is the RTX 2080 Ti for Leela?

oh, that’s intereresting. what does minibatch do? I guess evaluates 512 positions as a batch?! but who provides the positions, I mean if everytime I want an evaluation, I have to wait for somebody else to want another 511 positions before i get get mine, I’m going to be doing a lot of stalling, wai...
by ankan
Sat Oct 27, 2018 6:11 am
Forum: Computer Chess Club: General Topics
Topic: How good is the RTX 2080 Ti for Leela?
Replies: 48
Views: 12613

Re: How good is the RTX 2080 Ti for Leela?

I think I might have used different settings (just to maximize nps). I also might have the GPU overclocked when I posted that result. Generally letting it run for longer (like for 5m nodes) with bigger nncache setting results in higher nps. I just updated the sheet with benchmark at two different se...
by ankan
Mon Sep 17, 2018 3:29 pm
Forum: Computer Chess Club: General Topics
Topic: How good is the RTX 2080 Ti for Leela?
Replies: 48
Views: 12613

Re: How good is the RTX 2080 Ti for Leela?

However, there do seem to be some differences with the CUDA cores between Quadro and Geforce. Forget that comment - I see why it's wrong now. What's FP16 accumulate? Tensor cores perform small matrix multiplies and accumulate. See https://devblogs.nvidia.com/programming-tensor-cores-cuda-9/ for mor...
by ankan
Mon Sep 17, 2018 4:25 am
Forum: Computer Chess Club: General Topics
Topic: How good is the RTX 2080 Ti for Leela?
Replies: 48
Views: 12613

Re: How good is the RTX 2080 Ti for Leela?

I’m not accusing you of lying but why would Nvidia cripple the CUDA cores on the 2080 Ti for FP16 (presumably to protect Quadro) and then allow the tensor cores to run full speed? In a week or two Lc0’s speed on this card will finally be revealed- I hope you’re right I don't know from where people ...
by ankan
Sun Sep 16, 2018 4:09 am
Forum: Computer Chess Club: General Topics
Topic: 2080 Ti
Replies: 10
Views: 3134

Re: 2080 Ti

Nvidia are claiming tensor can process FP16, which is quite a big deal. If it can’t be used it’s useless. Different sites are claiming very different numbers. I just want to find out how fast the card really is for Lc0. First you have to understand what tensor cores are. They process with mixed pre...
by ankan
Sun Sep 16, 2018 3:57 am
Forum: Computer Chess Club: General Topics
Topic: If you could buy any single CPU system for chess . . .
Replies: 32
Views: 5140

Re: If you could buy any single CPU system for chess . . .

Threadripper 2990wx (32 core/64 thread) or 2950x (16 core/32 thread) are looking very good - at least in term of perf/$:
E.g: see stockfish benchmarks at the bottom of the page:
https://www.phoronix.com/scan.php?page= ... 90wx&num=5
by ankan
Sun Sep 16, 2018 3:46 am
Forum: Computer Chess Club: General Topics
Topic: How good is the RTX 2080 Ti for Leela?
Replies: 48
Views: 12613

Re: How good is the RTX 2080 Ti for Leela?

It should be very similar to a Titan V for lc0. It has tensor cores enabled, and it's peak fp16 tensor math throughput is almost exactly same as a Titan V (114 Tflops vs 110 Tflops): https://www.anandtech.com/show/13282/nvidia-turing-architecture-deep-dive/6 I have one, but I can't post any benchmar...
by ankan
Thu Jul 19, 2018 12:25 pm
Forum: Computer Chess Club: General Topics
Topic: Something goes wrong with lc0 since yesterday?
Replies: 272
Views: 43265

Re: Something goes wrong with lc0 since yesterday?

ID9155 was after many drops in the learning rate (4 or 5?). Test10 had only a single LR drop till now. Bigger LR can help the network learn fast initially and avoid local minima, but it also prevents it converging into a minima (causing relatively large fluctuations in performance after every traini...
by ankan
Thu Jun 07, 2018 2:10 pm
Forum: Computer Chess Club: General Topics
Topic: First Win by Leela Chess Zero against Stockfish dev
Replies: 30
Views: 5900

First Win by Leela Chess Zero against Stockfish dev

2 wins in a match of 50 games: https://lichess.org/L6Vpgi3w https://lichess.org/8ulwmKUl Time control: 5 min + 10 seconds per move. 4GB hash size, No opening book for lc0, Arena\Books\olympiad.abk for Stockfish, no tablebases lc0-win-20180604-cuda92-cudnn714-experimental running on Titan V (network ...