mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware

Reply
 
Thread Tools
Old 2020-03-25, 16:59   #793
kladner
 
kladner's Avatar
 
"Kieren, ktony"
Jul 2011

3×52×127 Posts
Default

Quote:
Originally Posted by Viliam Furik View Post
Could you do it on range of FFTs (e.g. 2048 to 8192), and without testing all implementations?
It will take a little while, but under lockdown I have time on my hands.
EDIT: Here ya go.
Attached Files
File Type: txt CPU clock bench 02.txt (9.6 KB, 14 views)

Last fiddled with by kladner on 2020-03-25 at 17:49
kladner is offline   Reply With Quote
Old 2020-03-25, 21:18   #794
Viliam Furik
 
Jul 2018
Martin, Slovakia

616 Posts
Default

Quote:
Originally Posted by kladner View Post
It will take a little while, but under lockdown I have time on my hands.
EDIT: Here ya go.
Thank you. I've looked at the results, and it seems like there is little advantage from RAM speed with those speeds and that CPU. And for some reason, the 4000 turns out fastest on some FFT lengths, but it is not that much so it may be a measurement error (some background tasks).

Is it possible that when RAM is faster than CPU clock, it will not be used the same as with faster CPU? I'm thinking about this exact thing, because I have 3200 MHz RAM now, and I want to know whether I should buy 4000 MHz or 4400 MHz RAM because my CPU is manually overclocked to 4,1 GHz (Ryzen 9 3900X).
Viliam Furik is offline   Reply With Quote
Old 2020-03-25, 22:02   #795
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

1A5A16 Posts
Default

Quote:
Originally Posted by Viliam Furik View Post
Thank you. I've looked at the results, and it seems like there is little advantage from RAM speed with those speeds and that CPU.
Note the massive L3 cache on that chip. It is not surprising that RAM speed is not a factor.
Prime95 is offline   Reply With Quote
Old 2020-03-26, 01:51   #796
axn
 
axn's Avatar
 
Jun 2003

11B616 Posts
Default

Quote:
Originally Posted by Prime95 View Post
Note the massive L3 cache on that chip. It is not surprising that RAM speed is not a factor.
L3 is 8MB on a 6700K, which is mediocre.
axn is offline   Reply With Quote
Old 2020-03-26, 02:25   #797
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

76378 Posts
Default

I think he was referring to the Ryzen?
Edit: 70MB L3, if my search result is to be believed.

Last fiddled with by VBCurtis on 2020-03-26 at 02:26
VBCurtis is offline   Reply With Quote
Old 2020-03-26, 03:04   #798
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

2·3,373 Posts
Default

Quote:
Originally Posted by axn View Post
L3 is 8MB on a 6700K, which is mediocre.
Sorry, confused the benchmark with JCoveiro's Ryzen.
Prime95 is offline   Reply With Quote
Old 2020-03-26, 03:40   #799
axn
 
axn's Avatar
 
Jun 2003

2×2,267 Posts
Default

Quote:
Originally Posted by Prime95 View Post
Sorry, confused the benchmark with JCoveiro's Ryzen.
Ok, makes sense. That still leaves the question of why mem speed is not affecting performance. I suppose, with just 4 cores, all of the tested RAM speeds are sufficient to feed the CPU.

Quote:
Originally Posted by VBCurtis View Post
I think he was referring to the Ryzen?
Edit: 70MB L3, if my search result is to be believed.
Ryzen L3 are all power-of-2, but since it is a victim cache, AMD specs list L2+L3 as a single "cache" number. So if you see 70MB for a processor, that is 64MB L3 + 6MB (12*512KB) L2.
axn is offline   Reply With Quote
Old 2020-03-26, 16:15   #800
kladner
 
kladner's Avatar
 
"Kieren, ktony"
Jul 2011

3·52·127 Posts
Default

I have to note that the RAM speed was the same in all these runs. Only the CPU clock was different. I took the very similar results as an indication that this system is memory-bound.
EDIT: Also, these tests were not run under strict lab-like conditions. I did shut down obvious cycle-stealing apps like performance monitors and browsers. In line with normal operation on the machine I deliberately left mfaktc running on the GPU as part of the environment. Allowance has to be made for margin-of-error.

Last fiddled with by kladner on 2020-03-26 at 16:21
kladner is offline   Reply With Quote
Old 2020-03-30, 04:12   #801
Rodrigo
 
Rodrigo's Avatar
 
Jun 2010
Pennsylvania

13·71 Posts
Default

I didn't see any benchmarks here for the i5-7500, so here goes:

Code:
[Sun Mar 29 19:08:40 2020]
Compare your results to other computers at http://www.mersenne.org/report_benchmarks
Intel(R) Core(TM) i5-7500 CPU @ 3.40GHz
CPU speed: 3371.20 MHz, 4 cores
CPU features: Prefetchw, SSE, SSE2, SSE4, AVX, AVX2, FMA
L1 cache size: 4x32 KB, L2 cache size: 4x256 KB, L3 cache size: 6 MB
L1 cache line size: 64 bytes, L2 cache line size: 64 bytes
Machine topology as determined by hwloc library:
 Machine#0 (total=28406668KB, Backend=Windows, hwlocVersion=2.0.4, ProcessName=prime95.exe)
  Package (total=28406668KB, CPUVendor=GenuineIntel, CPUFamilyNumber=6, CPUModelNumber=158, CPUModel="Intel(R) Core(TM) i5-7500 CPU @ 3.40GHz", CPUStepping=9)
    L3 (size=6144KB, linesize=64, ways=12, Inclusive=1)
      L2 (size=256KB, linesize=64, ways=4, Inclusive=0)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x00000001)
            PU#0 (cpuset: 0x00000001)
      L2 (size=256KB, linesize=64, ways=4, Inclusive=0)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x00000002)
            PU#1 (cpuset: 0x00000002)
      L2 (size=256KB, linesize=64, ways=4, Inclusive=0)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x00000004)
            PU#2 (cpuset: 0x00000004)
      L2 (size=256KB, linesize=64, ways=4, Inclusive=0)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x00000008)
            PU#3 (cpuset: 0x00000008)
Prime95 64-bit version 29.8, RdtscTiming=1
Timings for 2048K FFT length (4 cores, 1 worker):  2.53 ms.  Throughput: 395.74 iter/sec.
Timings for 2048K FFT length (4 cores, 4 workers): 10.69, 10.59, 10.39, 10.16 ms.  Throughput: 382.59 iter/sec.
Timings for 2304K FFT length (4 cores, 1 worker):  2.89 ms.  Throughput: 346.50 iter/sec.
Timings for 2304K FFT length (4 cores, 4 workers): 11.58, 11.32, 11.11, 11.06 ms.  Throughput: 355.15 iter/sec.
Timings for 2400K FFT length (4 cores, 1 worker):  3.00 ms.  Throughput: 333.51 iter/sec.
Timings for 2400K FFT length (4 cores, 4 workers): 12.26, 11.89, 11.78, 11.95 ms.  Throughput: 334.20 iter/sec.
Timings for 2560K FFT length (4 cores, 1 worker):  3.08 ms.  Throughput: 324.68 iter/sec.
Timings for 2560K FFT length (4 cores, 4 workers): 13.10, 12.72, 12.66, 12.57 ms.  Throughput: 313.49 iter/sec.
Timings for 2688K FFT length (4 cores, 1 worker):  3.39 ms.  Throughput: 295.12 iter/sec.
Timings for 2688K FFT length (4 cores, 4 workers): 13.30, 13.26, 13.18, 13.13 ms.  Throughput: 302.63 iter/sec.
Timings for 2880K FFT length (4 cores, 1 worker):  3.57 ms.  Throughput: 280.17 iter/sec.
Timings for 2880K FFT length (4 cores, 4 workers): 15.30, 14.57, 14.32, 14.28 ms.  Throughput: 273.86 iter/sec.
Timings for 3072K FFT length (4 cores, 1 worker):  3.74 ms.  Throughput: 267.17 iter/sec.
Timings for 3072K FFT length (4 cores, 4 workers): 15.53, 15.41, 15.03, 14.97 ms.  Throughput: 262.61 iter/sec.
Timings for 3200K FFT length (4 cores, 1 worker):  4.25 ms.  Throughput: 235.31 iter/sec.
Timings for 3200K FFT length (4 cores, 4 workers): 16.82, 16.34, 16.11, 16.10 ms.  Throughput: 244.87 iter/sec.
Timings for 3360K FFT length (4 cores, 1 worker):  4.45 ms.  Throughput: 224.61 iter/sec.
Timings for 3360K FFT length (4 cores, 4 workers): 17.95, 17.58, 17.29, 17.20 ms.  Throughput: 228.55 iter/sec.
[Sun Mar 29 19:13:44 2020]
Timings for 3456K FFT length (4 cores, 1 worker):  4.34 ms.  Throughput: 230.16 iter/sec.
Timings for 3456K FFT length (4 cores, 4 workers): 17.76, 17.24, 17.24, 17.07 ms.  Throughput: 230.91 iter/sec.
Timings for 3584K FFT length (4 cores, 1 worker):  4.70 ms.  Throughput: 212.98 iter/sec.
Timings for 3584K FFT length (4 cores, 4 workers): 18.46, 17.98, 17.77, 17.95 ms.  Throughput: 221.75 iter/sec.
Timings for 3840K FFT length (4 cores, 1 worker):  4.86 ms.  Throughput: 205.63 iter/sec.
Timings for 3840K FFT length (4 cores, 4 workers): 20.57, 20.09, 19.75, 19.80 ms.  Throughput: 199.52 iter/sec.
Timings for 4096K FFT length (4 cores, 1 worker):  5.18 ms.  Throughput: 192.93 iter/sec.
Timings for 4096K FFT length (4 cores, 4 workers): 21.19, 21.17, 20.45, 20.43 ms.  Throughput: 192.28 iter/sec.
Timings for 4480K FFT length (4 cores, 1 worker):  5.90 ms.  Throughput: 169.36 iter/sec.
Timings for 4480K FFT length (4 cores, 4 workers): 23.51, 22.80, 22.65, 22.76 ms.  Throughput: 174.50 iter/sec.
Timings for 4608K FFT length (4 cores, 1 worker):  5.78 ms.  Throughput: 173.03 iter/sec.
Timings for 4608K FFT length (4 cores, 4 workers): 24.61, 24.00, 23.96, 23.51 ms.  Throughput: 166.57 iter/sec.
Timings for 4800K FFT length (4 cores, 1 worker):  6.16 ms.  Throughput: 162.33 iter/sec.
Timings for 4800K FFT length (4 cores, 4 workers): 24.56, 24.78, 24.18, 24.02 ms.  Throughput: 164.06 iter/sec.
Timings for 5120K FFT length (4 cores, 1 worker):  6.81 ms.  Throughput: 146.90 iter/sec.
Timings for 5120K FFT length (4 cores, 4 workers): 27.33, 26.39, 26.15, 25.96 ms.  Throughput: 151.23 iter/sec.
Timings for 5376K FFT length (4 cores, 1 worker):  7.06 ms.  Throughput: 141.55 iter/sec.
Timings for 5376K FFT length (4 cores, 4 workers): 29.35, 28.62, 28.16, 28.20 ms.  Throughput: 139.97 iter/sec.
Timings for 5760K FFT length (4 cores, 1 worker):  7.67 ms.  Throughput: 130.29 iter/sec.
[Sun Mar 29 19:18:56 2020]
Timings for 5760K FFT length (4 cores, 4 workers): 31.20, 30.60, 30.13, 30.14 ms.  Throughput: 131.10 iter/sec.
Timings for 6144K FFT length (4 cores, 1 worker):  8.59 ms.  Throughput: 116.37 iter/sec.
Timings for 6144K FFT length (4 cores, 4 workers): 33.18, 33.27, 32.70, 32.32 ms.  Throughput: 121.72 iter/sec.
Timings for 6400K FFT length (4 cores, 1 worker):  8.38 ms.  Throughput: 119.28 iter/sec.
Timings for 6400K FFT length (4 cores, 4 workers): 34.66, 34.26, 33.99, 33.72 ms.  Throughput: 117.11 iter/sec.
Timings for 6720K FFT length (4 cores, 1 worker):  9.01 ms.  Throughput: 110.97 iter/sec.
Timings for 6720K FFT length (4 cores, 4 workers): 36.54, 36.32, 35.39, 35.57 ms.  Throughput: 111.27 iter/sec.
Timings for 6912K FFT length (4 cores, 1 worker):  9.47 ms.  Throughput: 105.60 iter/sec.
Timings for 6912K FFT length (4 cores, 4 workers): 37.08, 36.96, 36.90, 36.56 ms.  Throughput: 108.48 iter/sec.
Timings for 7168K FFT length (4 cores, 1 worker):  9.49 ms.  Throughput: 105.35 iter/sec.
Timings for 7168K FFT length (4 cores, 4 workers): 39.29, 39.24, 37.98, 38.03 ms.  Throughput: 103.57 iter/sec.
Timings for 7680K FFT length (4 cores, 1 worker): 10.01 ms.  Throughput: 99.90 iter/sec.
Timings for 7680K FFT length (4 cores, 4 workers): 42.06, 40.07, 39.64, 39.48 ms.  Throughput: 99.29 iter/sec.
Timings for 8064K FFT length (4 cores, 1 worker): 11.01 ms.  Throughput: 90.86 iter/sec.
Timings for 8064K FFT length (4 cores, 4 workers): 44.00, 44.86, 42.86, 42.76 ms.  Throughput: 91.73 iter/sec.
Timings for 8192K FFT length (4 cores, 1 worker): 11.02 ms.  Throughput: 90.78 iter/sec.
Timings for 8192K FFT length (4 cores, 4 workers): 45.76, 46.28, 43.97, 44.50 ms.  Throughput: 88.68 iter/sec.
Rodrigo is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
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 "interesting video" thread... Xyzzy Lounge 9 2006-12-24 20:06
Perpetual autostereogram thread... Xyzzy Lounge 10 2006-09-28 00:36
Perpetual ECM factoring challenge thread... Xyzzy Factoring 65 2005-09-05 08:16

All times are UTC. The time now is 00:17.

Sat Apr 4 00:17:08 UTC 2020 up 9 days, 21:50, 1 user, load averages: 1.47, 1.67, 1.56

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.