TalkChess.com

My gpu workstation went broken, so I decided to rent gpus via cloud instead of
buying a new system, ofc, this is no solution for long run engine testing,
but suitable for tuning for new architectures or to run some benchmarks.

--
Srdja

Maybe the wrong time to play the smart ass,
but this issue reminds me of some simple rules...

- have always a second (old) system in reverse to be able to surf the web
- have always an working installation media present
- divide always user data from OS (two partitions) on Windows systems

--
Srdja

XBoard sends 'ping' after each 'new', and in tournament mode actually waits with starting the clock until both engines have submitted the corresponding 'pong's. Not sure if it also does that in old-fashioned two-player matches, though. (The tournament mode has another, much more complex NextGame fu...

My Zeta engine needs up to some seconds to init the gpu once, and the gpu memory
for each new game...so what is the best way to signal xboard that clock should
not run?

--
Srdja

Okay, I locked only nodes with depth>1 during ABDADA phase #2, iteration 0, changed to detph>0 and results look better... ################################################################################ # Zeta 099m, startposition, depth 12, best of 4 runs, Nvidia V100: # ### workers #nps #nps speedu...

1. Are you renting a server with only 32-processors on it? NVidia Volta only has 80 SMs. I did rent a 2core cpu with V100. Afaik, the V100 has 80 SMs resp. Compute Units, each with 64 cores divided into four 16 wide SIMD units. But I can run only 2 workers per Compute Unit efficient on Volta and Tu...

Okay, I cleaned up my code, and tried my own + suggestion, it has no effect, at least the scaling seems now stable with up to 32 workers, but I am still clueless why scaling stalls with more than 32 workers... # Zeta v099m, start pos, depth 12, Nvidia V100, tt: 2048 MB, ttabdada: 1536 MB ### workers...

And what are the reasons to believe the opposite? We should believe the opposite because very few algorithms for anything have been found (though those few include some extremely important applications), and those that have are for things that have nothing at all in common with the game of chess, c...

jp wrote: ↑

Tue Jul 02, 2019 11:05 am

smatovic wrote: ↑

Tue Jul 02, 2019 10:59 am

As long as no one proofs that it can not be done,
there is imo prospect that it can be done...

You could claim the same of a classical computer. There is no reason to believe such an algorithm exists, waiting to be found.

And what are the reasons to believe the opposite?

--
Srdja

jp wrote: ↑

Tue Jul 02, 2019 10:54 am

Read my previous post. All the qubits in the world will not help.

As long as no one proofs that it can not be done,
there is imo prospect that it can be done...

--
Srdja