Perpetual benchmark thread... - Page 57

kracker · 2014-01-21, 23:58

Quote:

Originally Posted by henryzz

Only on 1 core 1 thread.
Anything else 27.9 is faster.

Hmm, I see. Wonder why.

chalsall · 2014-01-22, 00:42

Quote:

Originally Posted by kracker

Hmm, I see. Wonder why.

Keep in mind Prime95 (or, at least, mprime) doesn't appear to do any affinity setting during it's benchmark runs (even if AffinityScramble2 is set).

Thus for multi-core (and definitely multi-socket) situations the benchmarks are not fully representative of what one might find in tuned production.

kracker · 2014-01-22, 17:58

Quote:

Originally Posted by chalsall

Keep in mind Prime95 (or, at least, mprime) doesn't appear to do any affinity setting during it's benchmark runs (even if AffinityScramble2 is set).

Thus for multi-core (and definitely multi-socket) situations the benchmarks are not fully representative of what one might find in tuned production.

I know. What I don't know, is why 28.3 is slower in multicore.(Realistic or not it should compare to 27.9)

chalsall · 2014-01-22, 18:32

Quote:

Originally Posted by kracker

(Realistic or not it should compare to 27.9)

Not always.

It will depend on how well (or not) the OS balances the load across the real processors, during the particular run. If threads get thrown on random processors (some of whom may share real cores in the case of a Hyper-Threaded environment) then the results will be sub-optimal.

There is no reason Prime95/mprime shouldn't use the knowledge provided in the AffinityScramble2 setting during bench-marking.

And, yet, it doesn't.

kracker · 2014-01-22, 20:13

Quote:

Originally Posted by chalsall

Not always.

It will depend on how well (or not) the OS balances the load across the real processors, during the particular run. If threads get thrown on random processors (some of whom may share real cores in the case of a Hyper-Threaded environment) then the results will be sub-optimal.

There is no reason Prime95/mprime shouldn't use the knowledge provided in the AffinityScramble2 setting during bench-marking.

And, yet, it doesn't.

I see. Well, I'll be testing more later.

kracker · 2014-01-27, 18:59

4670K at 3.6 GHz.

Code:

Compare your results to other computers at http://www.mersenne.org/report_benchmarks
Intel(R) Core(TM) i5-4670K CPU @ 3.40GHz
CPU speed: 3588.46 MHz, 4 cores
CPU features: Prefetch, SSE, SSE2, SSE4, AVX, AVX2, FMA
L1 cache size: 32 KB
L2 cache size: 256 KB, L3 cache size: 6 MB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
TLBS: 64
Prime95 64-bit version 28.3, RdtscTiming=1
Best time for 768K FFT length: 2.896 ms., avg: 3.018 ms.
Best time for 896K FFT length: 3.590 ms., avg: 3.671 ms.
Best time for 1024K FFT length: 4.154 ms., avg: 4.206 ms.
Best time for 1280K FFT length: 5.138 ms., avg: 5.213 ms.
Best time for 1536K FFT length: 6.396 ms., avg: 6.629 ms.
Best time for 1792K FFT length: 7.508 ms., avg: 7.585 ms.
Best time for 2048K FFT length: 8.524 ms., avg: 8.790 ms.
Best time for 2560K FFT length: 11.151 ms., avg: 11.234 ms.
Best time for 3072K FFT length: 13.080 ms., avg: 13.182 ms.
Best time for 3584K FFT length: 15.742 ms., avg: 15.864 ms.
Best time for 4096K FFT length: 17.958 ms., avg: 18.117 ms.
Best time for 5120K FFT length: 22.582 ms., avg: 22.864 ms.
Best time for 6144K FFT length: 28.228 ms., avg: 28.465 ms.
Best time for 7168K FFT length: 32.562 ms., avg: 32.855 ms.
Best time for 8192K FFT length: 37.678 ms., avg: 38.298 ms.
Timing FFTs using 2 threads.
Best time for 768K FFT length: 1.546 ms., avg: 1.583 ms.
Best time for 896K FFT length: 1.947 ms., avg: 1.989 ms.
Best time for 1024K FFT length: 2.229 ms., avg: 2.362 ms.
Best time for 1280K FFT length: 2.732 ms., avg: 2.778 ms.
Best time for 1536K FFT length: 3.360 ms., avg: 3.389 ms.
Best time for 1792K FFT length: 3.972 ms., avg: 4.042 ms.
Best time for 2048K FFT length: 4.584 ms., avg: 4.698 ms.
Best time for 2560K FFT length: 5.824 ms., avg: 5.992 ms.
Best time for 3072K FFT length: 6.880 ms., avg: 6.948 ms.
Best time for 3584K FFT length: 8.215 ms., avg: 8.287 ms.
Best time for 4096K FFT length: 9.419 ms., avg: 9.488 ms.
Best time for 5120K FFT length: 11.884 ms., avg: 12.013 ms.
Best time for 6144K FFT length: 14.816 ms., avg: 15.005 ms.
Best time for 7168K FFT length: 17.121 ms., avg: 17.579 ms.
Best time for 8192K FFT length: 19.816 ms., avg: 19.966 ms.
Timing FFTs using 3 threads.
Best time for 768K FFT length: 1.101 ms., avg: 1.152 ms.
Best time for 896K FFT length: 1.286 ms., avg: 1.320 ms.
Best time for 1024K FFT length: 1.607 ms., avg: 1.649 ms.
Best time for 1280K FFT length: 1.948 ms., avg: 2.056 ms.
Best time for 1536K FFT length: 2.369 ms., avg: 2.398 ms.
Best time for 1792K FFT length: 2.824 ms., avg: 2.918 ms.
Best time for 2048K FFT length: 3.304 ms., avg: 3.334 ms.
Best time for 2560K FFT length: 4.129 ms., avg: 4.162 ms.
Best time for 3072K FFT length: 4.911 ms., avg: 5.136 ms.
Best time for 3584K FFT length: 5.875 ms., avg: 6.001 ms.
Best time for 4096K FFT length: 6.688 ms., avg: 7.459 ms.
Best time for 5120K FFT length: 8.515 ms., avg: 8.614 ms.
Best time for 6144K FFT length: 10.506 ms., avg: 10.641 ms.
Best time for 7168K FFT length: 12.202 ms., avg: 12.354 ms.
Best time for 8192K FFT length: 14.144 ms., avg: 14.314 ms.
Timing FFTs using 4 threads.
Best time for 768K FFT length: 0.846 ms., avg: 0.901 ms.
Best time for 896K FFT length: 1.104 ms., avg: 1.121 ms.
Best time for 1024K FFT length: 1.260 ms., avg: 1.293 ms.
Best time for 1280K FFT length: 1.612 ms., avg: 1.648 ms.
Best time for 1536K FFT length: 1.979 ms., avg: 2.001 ms.
Best time for 1792K FFT length: 2.371 ms., avg: 2.404 ms.
Best time for 2048K FFT length: 2.882 ms., avg: 2.921 ms.
Best time for 2560K FFT length: 3.467 ms., avg: 3.505 ms.
Best time for 3072K FFT length: 4.268 ms., avg: 4.321 ms.
Best time for 3584K FFT length: 5.042 ms., avg: 5.164 ms.
Best time for 4096K FFT length: 5.741 ms., avg: 5.851 ms.
Best time for 5120K FFT length: 7.370 ms., avg: 7.417 ms.
Best time for 6144K FFT length: 8.927 ms., avg: 9.048 ms.
Best time for 7168K FFT length: 10.501 ms., avg: 10.859 ms.
Best time for 8192K FFT length: 12.150 ms., avg: 12.262 ms.
Best time for 61 bit trial factors: 1.909 ms.
Best time for 62 bit trial factors: 1.974 ms.
Best time for 63 bit trial factors: 2.168 ms.
Best time for 64 bit trial factors: 2.079 ms.
Best time for 65 bit trial factors: 2.556 ms.
Best time for 66 bit trial factors: 2.966 ms.
Best time for 67 bit trial factors: 2.950 ms.
Best time for 75 bit trial factors: 2.860 ms.
Best time for 76 bit trial factors: 2.869 ms.
Best time for 77 bit trial factors: 2.850 ms.

kracker · 2014-02-26, 23:17

Reposting from my Haswell i5-4430, this time with FMA3(28.4) and full turbo boost in BIOS.

Code:

Compare your results to other computers at http://www.mersenne.org/report_benchmarks
Intel(R) Core(TM) i5-4430 CPU @ 3.00GHz
CPU speed: 2977.43 MHz, 4 cores
CPU features: Prefetch, SSE, SSE2, SSE4, AVX, AVX2, FMA
L1 cache size: 32 KB
L2 cache size: 256 KB, L3 cache size: 6 MB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
TLBS: 64
Prime95 64-bit version 28.4, RdtscTiming=1
Best time for 768K FFT length: 3.265 ms., avg: 3.358 ms.
Best time for 896K FFT length: 4.061 ms., avg: 4.287 ms.
Best time for 1024K FFT length: 4.598 ms., avg: 5.087 ms.
Best time for 1280K FFT length: 5.966 ms., avg: 6.712 ms.
Best time for 1536K FFT length: 7.227 ms., avg: 7.656 ms.
Best time for 1792K FFT length: 8.665 ms., avg: 8.741 ms.
Best time for 2048K FFT length: 9.953 ms., avg: 10.052 ms.
Best time for 2560K FFT length: 12.723 ms., avg: 13.237 ms.
Best time for 3072K FFT length: 15.264 ms., avg: 15.604 ms.
Best time for 3584K FFT length: 18.170 ms., avg: 18.831 ms.
Best time for 4096K FFT length: 20.786 ms., avg: 21.983 ms.
Best time for 5120K FFT length: 26.395 ms., avg: 26.989 ms.
Best time for 6144K FFT length: 31.787 ms., avg: 32.680 ms.
Best time for 7168K FFT length: 37.845 ms., avg: 38.714 ms.
Best time for 8192K FFT length: 43.805 ms., avg: 44.922 ms.
Timing FFTs using 2 threads.
Best time for 768K FFT length: 1.740 ms., avg: 1.760 ms.
Best time for 896K FFT length: 2.164 ms., avg: 2.190 ms.
Best time for 1024K FFT length: 2.581 ms., avg: 2.611 ms.
Best time for 1280K FFT length: 3.281 ms., avg: 3.323 ms.
Best time for 1536K FFT length: 3.929 ms., avg: 4.039 ms.
Best time for 1792K FFT length: 4.728 ms., avg: 4.789 ms.
Best time for 2048K FFT length: 5.495 ms., avg: 5.546 ms.
Best time for 2560K FFT length: 6.979 ms., avg: 7.403 ms.
Best time for 3072K FFT length: 8.434 ms., avg: 8.607 ms.
Best time for 3584K FFT length: 10.027 ms., avg: 10.148 ms.
Best time for 4096K FFT length: 11.616 ms., avg: 12.202 ms.
Best time for 5120K FFT length: 14.601 ms., avg: 15.142 ms.
Best time for 6144K FFT length: 17.339 ms., avg: 18.018 ms.
Best time for 7168K FFT length: 20.976 ms., avg: 21.495 ms.
Best time for 8192K FFT length: 24.259 ms., avg: 24.567 ms.
Timing FFTs using 3 threads.
Best time for 768K FFT length: 1.221 ms., avg: 1.263 ms.
Best time for 896K FFT length: 1.586 ms., avg: 1.625 ms.
Best time for 1024K FFT length: 1.893 ms., avg: 1.926 ms.
Best time for 1280K FFT length: 2.519 ms., avg: 2.626 ms.
Best time for 1536K FFT length: 3.135 ms., avg: 3.392 ms.
Best time for 1792K FFT length: 3.780 ms., avg: 3.847 ms.
Best time for 2048K FFT length: 4.513 ms., avg: 4.707 ms.
Best time for 2560K FFT length: 5.689 ms., avg: 5.901 ms.
Best time for 3072K FFT length: 6.827 ms., avg: 6.880 ms.
Best time for 3584K FFT length: 8.115 ms., avg: 8.249 ms.
Best time for 4096K FFT length: 9.490 ms., avg: 9.529 ms.
Best time for 5120K FFT length: 11.845 ms., avg: 12.117 ms.
Best time for 6144K FFT length: 14.344 ms., avg: 14.784 ms.
Best time for 7168K FFT length: 16.957 ms., avg: 17.177 ms.
Best time for 8192K FFT length: 19.912 ms., avg: 20.086 ms.
Timing FFTs using 4 threads.
Best time for 768K FFT length: 0.978 ms., avg: 0.995 ms.
Best time for 896K FFT length: 1.335 ms., avg: 1.543 ms.
Best time for 1024K FFT length: 1.718 ms., avg: 1.795 ms.
Best time for 1280K FFT length: 2.352 ms., avg: 2.459 ms.
Best time for 1536K FFT length: 2.966 ms., avg: 3.035 ms.
Best time for 1792K FFT length: 3.608 ms., avg: 3.655 ms.
Best time for 2048K FFT length: 4.236 ms., avg: 4.286 ms.
Best time for 2560K FFT length: 5.341 ms., avg: 5.440 ms.
Best time for 3072K FFT length: 6.503 ms., avg: 6.881 ms.
Best time for 3584K FFT length: 7.591 ms., avg: 7.765 ms.
Best time for 4096K FFT length: 8.827 ms., avg: 8.903 ms.
Best time for 5120K FFT length: 11.076 ms., avg: 11.509 ms.
Best time for 6144K FFT length: 13.330 ms., avg: 13.970 ms.
Best time for 7168K FFT length: 15.703 ms., avg: 16.276 ms.
Best time for 8192K FFT length: 18.561 ms., avg: 18.653 ms.
Best time for 61 bit trial factors: 2.146 ms.
Best time for 62 bit trial factors: 2.215 ms.
Best time for 63 bit trial factors: 2.437 ms.
Best time for 64 bit trial factors: 2.333 ms.
Best time for 65 bit trial factors: 2.863 ms.
Best time for 66 bit trial factors: 3.329 ms.
Best time for 67 bit trial factors: 3.296 ms.
Best time for 75 bit trial factors: 3.214 ms.
Best time for 76 bit trial factors: 3.213 ms.
Best time for 77 bit trial factors: 3.204 ms.

1728K FFT: (with only dual 1600 memory :verysadface: )
1 thread =9 ms
2 threads=10 ms
3 threads=12 ms
4 threads=15 ms

axn · 2014-02-27, 03:25

Quote:

Originally Posted by kracker

1728K FFT: (with only dual 1600 memory :verysadface: )
1 thread =9 ms
2 threads=10 ms
3 threads=12 ms
4 threads=15 ms

?

kracker · 2014-02-27, 20:12

Quote:

Originally Posted by axn

?

The nice memory bottleneck

axn · 2014-02-28, 04:02

Quote:

Originally Posted by kracker

The nice memory bottleneck

No, I mean the 1728K FFT. Is that a supported length? If so, 1792K appears to be faster.

kracker · 2014-02-28, 23:06

Quote:

Originally Posted by axn

No, I mean the 1728K FFT. Is that a supported length? If so, 1792K appears to be faster.

Hmm. Might be my typo, will check it out.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Perpetual "interesting video" thread...	Xyzzy	Lounge	43	2021-07-17 00:00
LLR benchmark thread	Oddball	Riesel Prime Search	5	2010-08-02 00:11
Perpetual I'm pi**ed off thread	rogue	Soap Box	19	2009-10-28 19:17
Perpetual autostereogram thread...	Xyzzy	Lounge	10	2006-09-28 00:36
Perpetual ECM factoring challenge thread...	Xyzzy	Factoring	65	2005-09-05 08:16