8GB..

I can easily hit a wall with 4GB if I want to - run two 64-bit chess engines, each with 2GB of hash memory, and, well, you need more than 4GB to do that.... it's not possible with just a 4GB machine... I currently run each with 1GB hash memory to fit within a 4GB machine. However I have an 8GB upgrade sitting here waiting to go in at some point.

Chess is one of the few things that can use a lot of memory and multiple CPUs. Many engines support up to 8CPUs and some a great deal more. The program Zappa Mexico II x64 can use up to 1TB of memory (although I doubt that this has actually been tested) and supports up to 512 CPUs (on which it has been tested). All this technology for just £50.. Not that anyone on this forum is remotely interested in the game of chess... But there you go, it's one justification for 8GB anyway
 
I can easily hit a wall with 4GB if I want to - run two 64-bit chess engines, each with 2GB of hash memory, and, well, you need more than 4GB to do that.... it's not possible with just a 4GB machine... I currently run each with 1GB hash memory to fit within a 4GB machine. However I have an 8GB upgrade sitting here waiting to go in at some point.

Chess is one of the few things that can use a lot of memory and multiple CPUs. Many engines support up to 8CPUs and some a great deal more. The program Zappa Mexico II x64 can use up to 1TB of memory (although I doubt that this has actually been tested) and supports up to 512 CPUs (on which it has been tested). All this technology for just £50.. Not that anyone on this forum is remotely interested in the game of chess... But there you go, it's one justification for 8GB anyway

are you some chess grand master or something?? its all well and good that it support it, does it actually need it?
 
OK - It's fairly pointless unless you need to use it for computer chess.;)

:) :)

Quite staggering though, that if I had a 512 CPU machine with 1TB of RAM I could use it ALL. In my lifetime I doubt I'll ever see such a machine at an affordable price
 
Back to the topic.. if for some reason it wouldn't do 425x8, I'd just change it to 378x9. With the amount of cache on a Q6600, I doubt there's a noticeable performance difference.
 
I can easily hit a wall with 4GB if I want to - run two 64-bit chess engines, each with 2GB of hash memory, and, well, you need more than 4GB to do that.... it's not possible with just a 4GB machine... I currently run each with 1GB hash memory to fit within a 4GB machine. However I have an 8GB upgrade sitting here waiting to go in at some point.

Chess is one of the few things that can use a lot of memory and multiple CPUs. Many engines support up to 8CPUs and some a great deal more. The program Zappa Mexico II x64 can use up to 1TB of memory (although I doubt that this has actually been tested) and supports up to 512 CPUs (on which it has been tested). All this technology for just £50.. Not that anyone on this forum is remotely interested in the game of chess... But there you go, it's one justification for 8GB anyway

Chess engines are quite interesting :) - and yes, "embarassingly" (as the terminology goes) parallelisable. I guess the Zappa engine uses MPI or something on a cluster for the huge CPU number support.

Want to write my own engine at some point - difficult to even write a bad one though!
 
Chess engines are quite interesting :) - and yes, "embarassingly" (as the terminology goes) parallelisable. I guess the Zappa engine uses MPI or something on a cluster for the huge CPU number support.

I don't think so, the author says

"Zappa Mexico can be used on Windows or Linux computers with up to 512 CPU cores. Unfortunately, the parallel implementation relies on shared memory, so you cannot hook two computers together with an ethernet cable and use them together, unless you are willing to give SGI several hundred thousand dollars. "
 
I don't think so, the author says

"Zappa Mexico can be used on Windows or Linux computers with up to 512 CPU cores. Unfortunately, the parallel implementation relies on shared memory, so you cannot hook two computers together with an ethernet cable and use them together, unless you are willing to give SGI several hundred thousand dollars. "

Oh right, interesting.
For a commercial chess engine, shared memory makes much more sense anyway -- the average person doesn't have a personal distributed memory cluster! :p
 
Back
Top Bottom