Using a quad core for testing

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

Jan Brouwer
Posts: 201
Joined: Thu Mar 22, 2007 7:12 pm
Location: Netherlands

Using a quad core for testing

Post by Jan Brouwer »

Hi

I'm considering buying a quad-core PC to improve the testing of my chess program.
I want to run four tournaments in parallel, one on each core. Are there any pitfalls I should be aware of?
It would be nice if the PC can be kept quiet. I wonder how feasible that is with e.g. a Q6600 running under full load for hours without resorting to some form of exotic cooling :?
Perhaps the newer 45 nm quad-cores run cooler?

Thanks
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Using a quad core for testing

Post by bob »

Jan Brouwer wrote:Hi

I'm considering buying a quad-core PC to improve the testing of my chess program.
I want to run four tournaments in parallel, one on each core. Are there any pitfalls I should be aware of?
It would be nice if the PC can be kept quiet. I wonder how feasible that is with e.g. a Q6600 running under full load for hours without resorting to some form of exotic cooling :?
Perhaps the newer 45 nm quad-cores run cooler?

Thanks
One thing... if you play 4 games at a time, there will be random behavior caused by architectural bottlenecks. 4 processors trying to access memory across a single bus is one example, and this will cause random delays. Should not invalidate the testing, but NPS values might be somewhat slower, and might fluctuate more..

You would see significant issues on a dual-processor quad-core box, as 8 cores just stumble all over each other trying to get to memory... And since you are talking Intel and shared L2 cache among 2 or 4 cores depending on what you are talking about, that is additional interference.
krazyken

Re: Using a quad core for testing

Post by krazyken »

Jan Brouwer wrote:Hi

I'm considering buying a quad-core PC to improve the testing of my chess program.
I want to run four tournaments in parallel, one on each core. Are there any pitfalls I should be aware of?
It would be nice if the PC can be kept quiet. I wonder how feasible that is with e.g. a Q6600 running under full load for hours without resorting to some form of exotic cooling :?
Perhaps the newer 45 nm quad-cores run cooler?

Thanks
Since you are probably going to be running an OS on there, you may get more reliable results trying to keep one core idle, so running 3 ponder off matches, will leave one core to take up the slack when background stuff wants to make noise. Also of importance is to keep plenty of RAM in the box for many simultaneous engines, and the OS. If 8 engines are running with 64MB hash thats 512MB for engine hash alone.
jdart
Posts: 4367
Joined: Fri Mar 10, 2006 5:23 am
Location: http://www.arasanchess.org

Re: Using a quad core for testing

Post by jdart »

Cooling in my experience is not really an issue even with the 1st-gen quad chips. I am running a Q6700 overclocked with 4 threads round the clock. It has been rock stable. But it does have a good 3rd-party heat sink/fan installed. The stock Intel cooler is not that good.

-Jon