BOINC 5.8 Benchmark
21 Dec 2006 11:42:19 UTC
Topic 192219
(moderation:
I have been testing almost every function of the BOINC Client 5.8. The only issue, I have assumed so far. Is that the application has a possibly different benchmark system. My OP\\s stats for the two different versions.
BOINC 5.4.11
Measured floating point speed: 403 million ops/sec
Measured integer speed: 702 million ops/sec
(403 and 702 were max I could attain on either)
BOINC 5.8
Measured floating point speed: 399 million ops/sec
Measured integer speed: 686 million ops/sec
(399 and 686 were max I could attain on either)
I know these are merely benchmark calculations. And similar, being small difference... Some computers may vary largely on benchmark results. But, when I specifically prepare my computer for a benchmark. My benchmark numbers, are very slightly changed. Only within about 3 points. So, if anyone has a idea on the reason for this, would be much helpful. THANKS!
d3xt3r.net
BOINC 5.8 Benchmark
)
Best to ask on the BOINC forums, since Einstein does not have anything to do with the programming of BOINC.
IMHO this is normal and
)
IMHO this is normal and within Tolerance.
The new BOINC Code likely has non-Benchmark related parts (GUI, I/O etc.) that have slightly reduced performance of BOINC itself. I also assume V5.80 is still for Alpha-Testers and not recommended for production-level usage.
Also, let's not forget that the BOINC benchmark code has have much more significant quirrels since years now (Linux code still benches 40% (!) slower). Anything inconsistent around the Benchmark would not surprise me at all.
PS.
Have you tried H.Oda's WCPUl2 Utility ?
It has the ability to change the Pentium II's L2 Cache Latency on the fly (back in the old days, I was able to safely reduce L2 Cache Latency on a PII 400 from default 7 all the way downto 2, which gained an overall 10%)
You should be able to gain a similar amount on your Pentium II 450MHz (keep in mind some Prime95 torture self-testing is recommended to assure correct operation)