mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Hardware (https://www.mersenneforum.org/forumdisplay.php?f=9)
-   -   Problem trial factoring + 64 bit (https://www.mersenneforum.org/showthread.php?t=4253)

EPF 2005-06-26 02:58

Problem trial factoring + 64 bit
 
I use a new AMD64 X2 4600+ system + Windows XP Professional x64
Prime 95 v. 24.12
The Benchmark-results for trial factoring are ok:

AMD Athlon(tm) 64 X2 Dual Core Processor 4600+
CPU speed: 2412.25 MHz
CPU features: RDTSC, CMOV, Prefetch, 3DNow!, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 64-bit version 24.12, RdtscTiming=1
Best time for 512K FFT length: 19.417 ms.
Best time for 640K FFT length: 25.089 ms.
Best time for 768K FFT length: 30.459 ms.
Best time for 896K FFT length: 36.589 ms.
Best time for 1024K FFT length: 40.672 ms.
Best time for 1280K FFT length: 51.727 ms.
Best time for 1536K FFT length: 63.289 ms.
Best time for 1792K FFT length: 76.679 ms.
Best time for 2048K FFT length: 85.756 ms.
Best time for 2560K FFT length: 113.498 ms.
Best time for 3072K FFT length: 138.862 ms.
Best time for 3584K FFT length: 167.647 ms.
Best time for 4096K FFT length: 187.319 ms.
Best time for 58 bit trial factors: 3.148 ms.
Best time for 59 bit trial factors: 3.143 ms.
Best time for 60 bit trial factors: 3.345 ms.
Best time for 61 bit trial factors: 3.450 ms.
Best time for 62 bit trial factors: 4.091 ms.
Best time for 63 bit trial factors: 4.405 ms.
Best time for 64 bit trial factors: 5.676 ms.
Best time for 65 bit trial factors: 6.724 ms.
Best time for 66 bit trial factors: 6.684 ms.
Best time for 67 bit trial factors: 6.664 ms.

But when running trial factor the times for 1000 iterations are very different from this: CPU0 needs about 19 s, CPU1 about 16 s for 65 bit (no other programs running, each CPU gets the full 50% for prime95.exe)

Any ideas what the problem may be???? How can these values be so different? Where's the time lost?

Prime95 2005-06-26 03:26

I don't think there is a problem. An "iteration" in 64-bit TF does 3 times as much work as an "iteration" in 32-bit prime95. The TF benchmark compensates for this by normalizing 64-bit timings (dividing by 3).

EPF 2005-06-26 04:12

Problem solved!
 
Thanks for the quick answering! Yes, the %-steps are 3-times bigger, now i can sleep in peace again.


All times are UTC. The time now is 21:21.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, Jelsoft Enterprises Ltd.