Author: Robert Hyatt
Date: 07:11:38 03/18/03
Go up one level in this thread
On March 18, 2003 at 04:46:33, enrico carrisco wrote: >On March 18, 2003 at 00:01:44, Robert Hyatt wrote: > >>On March 17, 2003 at 22:59:30, Aaron Gordon wrote: >> >>>On March 17, 2003 at 18:47:27, Eugene Nalimov wrote: >>> >>>>I just run the experiment. I used 2 otherwise identical 64-bit systems, one with >>>>3Mb of L3 cache, other with 1.5Mb. Machine with bigger cache run Crafty's >>>>"bench" comman 12% faster (1 CPU). >>>> >>>>That means that >>>>(1) Crafty's working set don't fit into 1.5Mb, >>>>(2) For systems with cache 1.5Mb or less (i.e. for almost all x86 systems) for >>>>Crafty memory speed matter. >>>> >>>>Thanks, >>>>Eugene >>> >>>Those types of systems aren't what people normally use. Most people here have a >>>Pentium 3, Athlon, Pentium 4, etc. Here is something I found with Crafty. >>> >>>Using the Nforce2 chipset I'm able to run the ram at speeds from 50% up to 200% >>>(100% being synchronous) of the fsb speed. I tested 200MHz FSB (400DDR) with >>>200MHz memory (400DDR) and 200fsb with 100MHz memory (200DDR). >>>The difference between ~1.6gb/s memory and ~3.2gb/s memory with craftys 'bench' >>>command was 0.14%. Yes, about one seventh of one percent. >> >>That might well suggest _another_ bottleneck in that particular machine.... > >What he didn't mention is that this _supposed_ bottlenecked machine is running >crafty faster than any single P4/XEON available, both at 1.6gb/s and 3.2gb/s >memory bandwidth. (Well, I guess there is that guy with the freon cooled P4 at >4GHz that is a hair faster...) > >Point being, there is no bottleneck here. Memory speeds have a fairly inelastic >effect on overall chess engine performance (with regards to the abovementioned >platorms.) > >-elc. I am not sure what you mean by "fairly inelastic effect." The effect of memory speed is significant. The best data I used to get came from Crays. They had multiple types of memory systems, from big/cheap to big/expensive. big/cheap would use large banks (and few of them) while the expensive version would use 32-way interleaving. Crafty ran _signficantly_ faster on the 32-way interleaved system than it did on something with lower overall memory performance. At one point I had two dual machines at the same clock rate (I think they were 300's but it was a few years ago.) One by Dell with memory interleaving, one by a "local stop-and-rob store" that did not have interleaving. The performance was distinctly better on the interleaved dell... If you could somehow double the memory bandwidth and halve the memory latency, you would see a large performance boost in Crafty. Otherwise it would not run faster on processors with bigger cache, everything else being equal. I didn't save the data, but I did test this on my older quad 700. The BIOS it had recognized up to 550mhz processors, and when I upgraded to 700, the main difference was that the 700's had 1mb L2 cache. Crafty ran significantly faster even though the 700's were running at 550mhz until I upgraded the Intel BIOS so that it would recognize 700mhz and ramp the clock speed up to where it was supposed to be. I seem to recall a 10-15% improvement, but I didn't save the data as it wasn't particularly interesting at the time.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.