I won't be able to do any test on 64bit version anymore as yesterday I have uninstalled WinXP x64... As for the rest, I see what I can do, but that may take me some time, so probably I will do it over the weekend.

Core2 Duo self built, gigabyte DQ6 motherboard, 965P chipset, Geil PC6400 DDR2@5-5-5-15 (as per printed spec on DIMMs), non-overclocked. Note: These scores are prior to doing a prime95 or memtest burn-in.

CPU Affinity to default:
Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
CPU speed: 2399.91 MHz
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 32 KB
L2 cache size: unknown
L1 cache line size: 64 bytes
L2 cache line size: unknown
Prime95 32-bit version 24.14, RdtscTiming=1
Best time for 512K FFT length: 10.550 ms.
Best time for 640K FFT length: 14.199 ms.
Best time for 768K FFT length: 17.478 ms.
Best time for 896K FFT length: 20.887 ms.
Best time for 1024K FFT length: 23.114 ms.
Best time for 1280K FFT length: 29.357 ms.
Best time for 1536K FFT length: 35.838 ms.
Best time for 1792K FFT length: 42.579 ms.
Best time for 2048K FFT length: 47.315 ms.
Best time for 2560K FFT length: 62.526 ms.
Best time for 3072K FFT length: 76.194 ms.
Best time for 3584K FFT length: 91.854 ms.
Best time for 4096K FFT length: 103.021 ms.
Best time for 58 bit trial factors: 4.660 ms.
Best time for 59 bit trial factors: 4.629 ms.
Best time for 60 bit trial factors: 4.631 ms.
Best time for 61 bit trial factors: 4.658 ms.
Best time for 62 bit trial factors: 7.427 ms.
Best time for 63 bit trial factors: 7.435 ms.
Best time for 64 bit trial factors: 7.122 ms.
Best time for 65 bit trial factors: 7.071 ms.
Best time for 66 bit trial factors: 7.080 ms.
Best time for 67 bit trial factors: 7.046 ms.



When setting CPU Affinity to 1:

Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
CPU speed: 2399.97 MHz
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 32 KB
L2 cache size: unknown
L1 cache line size: 64 bytes
L2 cache line size: unknown
Prime95 32-bit version 24.14, RdtscTiming=1
Best time for 512K FFT length: 10.531 ms.
Best time for 640K FFT length: 14.223 ms.
Best time for 768K FFT length: 17.459 ms.
Best time for 896K FFT length: 20.825 ms.
Best time for 1024K FFT length: 23.091 ms.
Best time for 1280K FFT length: 29.345 ms.
Best time for 1536K FFT length: 35.832 ms.
Best time for 1792K FFT length: 42.574 ms.
Best time for 2048K FFT length: 47.300 ms.
Best time for 2560K FFT length: 62.474 ms.
Best time for 3072K FFT length: 76.201 ms.
Best time for 3584K FFT length: 91.847 ms.
Best time for 4096K FFT length: 102.781 ms.
Best time for 58 bit trial factors: 4.665 ms.
Best time for 59 bit trial factors: 4.671 ms.
Best time for 60 bit trial factors: 4.671 ms.
Best time for 61 bit trial factors: 4.656 ms.
Best time for 62 bit trial factors: 7.430 ms.
Best time for 63 bit trial factors: 7.435 ms.
Best time for 64 bit trial factors: 7.121 ms.
Best time for 65 bit trial factors: 7.071 ms.
Best time for 66 bit trial factors: 7.079 ms.
Best time for 67 bit trial factors: 7.045 ms.

Interesting to see CPU affinity made some slight changes.

-- Craig

Hey Cruelty,

Did you happen to get those benchmarks this weekend?

George, do you have an idea why this happens:

"Something interesting happens when disabling all instructions except SSE2: for FFT>=2048 there is a speed increase..."

William

Sorry, I forgot about that :blush:
Give me another 12 hours and I will post those benchmarks :smile:

This is a Pentium D 915 2.8 Ghz dual-core oc'ed 10% to 3.08 Ghz

Intel(R) Pentium(R) D CPU 2.80GHz
CPU speed: 3080.98 MHz
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 16 KB
L2 cache size: 2048 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
TLBS: 64
Prime95 32-bit version 24.14, RdtscTiming=1
Best time for 512K FFT length: 14.791 ms.
Best time for 640K FFT length: 18.889 ms.
Best time for 768K FFT length: 23.014 ms.
Best time for 896K FFT length: 27.947 ms.
Best time for 1024K FFT length: 32.135 ms.
Best time for 1280K FFT length: 39.525 ms.
Best time for 1536K FFT length: 47.790 ms.
Best time for 1792K FFT length: 57.256 ms.
Best time for 2048K FFT length: 63.606 ms.
Best time for 2560K FFT length: 84.823 ms.
Best time for 3072K FFT length: 102.261 ms.
Best time for 3584K FFT length: 122.331 ms.
Best time for 4096K FFT length: 136.389 ms.
Best time for 58 bit trial factors: 8.934 ms.
Best time for 59 bit trial factors: 8.943 ms.
Best time for 60 bit trial factors: 8.977 ms.
Best time for 61 bit trial factors: 8.985 ms.
Best time for 62 bit trial factors: 12.275 ms.
Best time for 63 bit trial factors: 12.311 ms.
Best time for 64 bit trial factors: 14.398 ms.
Best time for 65 bit trial factors: 14.302 ms.
Best time for 66 bit trial factors: 14.339 ms.
Best time for 67 bit trial factors: 14.269 ms.

Voila :smile:

Perfect, thanx Cruelty; appreciate it.

I'll private message you soon the benchmarks for some old software (old LLR or PRP) when I get a chance. They aren't important and won't need to be posted in the forum.

Thanx again.

Cruelty,

Did you mean to say in the above post:

"Something interesting happens when disabling all instructions except SSE2: in the 64-bit Prime95 for FFT>4096; there is a speed increase..."

?
(probably happens to for 32 bit version)

Probably makes since because of the 4 Meg cache...; since new architecture perhaps code has to be rewritten...


P.S.--Xyzzy, why does the edit button not come up sometimes to let you edit your message?

[QUOTE=wfgarnett3;86302]"Something interesting happens when disabling all instructions except SSE2: in the 64-bit Prime95 for FFT>4096; there is a speed increase..."[/QUOTE]

Probably just a normal variation in benchmark timings.

[QUOTE=Prime95;86306]Probably just a normal variation in benchmark timings.[/QUOTE]
Hmmm, the difference between "RDTSC, CMOV, Prefetch" enabled benchmark and benchmark with those features disabled varies between 3 and 7% in favour of "SSE2 + MMX" only benchmark. This doesn't look like some standard variation to me.
I will rerun benchmarks one more time today to verify this.

I have verified it once more, however this time I was not able to reproduce the difference I have observed before - the differences are less than 1% actually. This makes me think what could actually happened before :question: