Author: enrico carrisco
Date: 22:05:53 01/16/06
Well, George Sobala and a few others knew I was working on this comparison since Paderborn. I'd first like to thank George Sobala and Alfred Andrejcic for their time to provide me with some benchmarks on their very impressive machines. I still have a puddle of drool on my keyboard.. ;) I was not able to get test numbers from a Quad Opteron 2.4GHz system, but I scaled the numbers up from a Quad Opteron 2.2GHz system -- too close to argue... ;) Some of the information below is repetitive, but it makes it easier to get the result you want from the aspect you want. Plus, it makes sure I don't get a reply five seconds later from Uri where he found a discrepancy... j/k Uri. :) The systems: ------------ Apple Quad 2.5GHz (Dual, dual-core) PowerPC G5 using Deep Shredder 9.11 AMD Quad 2.2GHz (Dual, dual-core) Opteron 275 using Deep Shredder 9.12 64-bit Conditions: ----------- Both systems used a 512MB hash and the start position in all tests. Results: -------- Quad G5 2.5GHz - Shredder 9.11 ------------------------------ 1 CPU = 392.865kn/s / x.xx 2 CPUs = 781.735kn/s / 1.989 4 CPUs = 1488.089kn/s / 3.787 Quad Opteron 2.2GHz - Shredder 9.12 64 bit ------------------------------------------ 1 CPU = 542.582kn/s / x.xx 2 CPUs = 994.213kn/s / 1.832 4 CPUs = 1852.996kn/s / 3.415 [SCALED] Quad Opteron 2.4GHz - Shredder 9.12 64 bit --------------------------------------------------- 1 CPU = 591.907kn/s / x.xx 2 CPUs = 1084.596kn/s / 1.832 4 CPUs = 2021.450kn/s / 3.415 Conclusions: ------------ The Quad Opteron 2.4GHz was ~35.8% faster than the Quad G5 2.5GHz for Shredder. The Quad Opteron 2.2GHz was ~24.5% faster than the Quad G5 2.5GHz for Shredder. The Quad G5 2.5GHz was 26.4% slower than the Quad Opteron 2.4GHz for Shredder. The Quad G5 2.5GHz was 19.7% slower than the Quad Opteron 2.2GHz for Shredder. The Quad G5 2.5GHz was ~10.9% more SMP efficient with 4 threads for Shredder. The Quad G5 2.5GHz was ~8.6% more SMP efficient with 2 threads for Shredder. Perhaps Deep Shredder 9.12's SMP efficiency could be improved with some optimizations. With the Opteron's bidirectional hypertransport bus and super low latency, I am surprised at the SMP efficiency Deep Shredder achieves -- I will have to test how it does in 32-bit mode on Opterons and AMD MPs. Regards, -elc. efficiency potential.
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.