hydrogenpi7 wrote:
pnprog wrote:
Krama wrote:
Who is stronger Rayon or Leela?
The situation is not so clear, but apparently, it really depends of the hardware used:
- with a big Nvidia GPU, Ray would be stronger (Ray make good use of Scuda technology)
- Without GPU or without Nvidia GPU, Leela would be stronger (Leela makes good use of CPU)
At that point, I think Ray needs a bigger user base to push its development forward.
Does anyone know how well does the current version of Ray scale?
http://www.yss-aya.com/cgos/19x19/standings.htmlThe listing says the Ray 4.3-4c is only about 100 or so ELO ahead of Leela-0100b3_750Ti.
But since it does better with Cuda than Leela, is it possible that if I play it on something like the below configuration that it might beat an average pro?
http://i.imgur.com/36CaJgm.pngI wonder why Leela is not opens source like ray, and about ray needing a bigger base, if Ray would come out with a GUI, I think since most people these days have a powerful GPU for gaming, that it would quickly get a larger base .
I cannot get Rayon to beat Leela not even a single game.
Unless I'm missing something or did something very wrong, I am confident that Leela is way stronger than Ray...
https://imgur.com/a/TrvPDhttps://gokibitz.com/kifu/B1fNfvwzZhttps://gokibitz.com/kifu/B1twxPDfWhttps://gokibitz.com/kifu/HkdyyDwGbI'm playing on 4790k with a gtx 1050ti mini using Gtp versions of latest leela and ray on sabaki board and leela wins everything though I set same time limits . What is going wrong?
I've tested short games, long games, on my home computer, on Google's cloud with 8x K80's and 24 CPUs.... Turned off pondering for Leela. Made sure I used the GPU version of Rayon and latest Rn4.2.5.... each time Leela GPU GTP beats Ray hands down...