mersenneforum.org  

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

Reply
 
Thread Tools
Old 2017-04-12, 15:44   #155
db597
 
db597's Avatar
 
Jan 2003

7·29 Posts
Default

Quote:
Originally Posted by Lorenzo View Post
It's very simple Just go to Advanced/Time (in main menu). Then type: 332220523 (field "exponent to time") and click on OK. And just wait for few minutes when it have done
~ thanks a lot
Sure, here's the results I got from running "332220523"...

Code:
[Main thread Apr 12 23:40] Mersenne number primality test program version 29.1
[Main thread Apr 12 23:40] Optimizing for CPU architecture: AMD Zen, L2 cache size: 512 KB, L3 cache size: 16 MB
[Main thread Apr 12 23:41] Starting worker.
[Worker #1 Apr 12 23:41] Worker starting
[Worker #1 Apr 12 23:41] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 166.036 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 166.302 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 166.738 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 166.389 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 166.334 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 166.299 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 166.313 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 166.647 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 166.375 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 166.333 ms.
[Worker #1 Apr 12 23:41] Iterations: 10.  Total time: 1.664 sec.
[Worker #1 Apr 12 23:41] Estimated time to complete this exponent: 639 days, 17 hours, 49 minutes.
[Worker #1 Apr 12 23:41] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 2 threads
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 83.299 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 83.317 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 83.393 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 83.439 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 83.243 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 83.272 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 83.256 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 83.331 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 83.330 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 83.295 ms.
[Worker #1 Apr 12 23:41] Iterations: 10.  Total time: 0.833 sec.
[Worker #1 Apr 12 23:41] Estimated time to complete this exponent: 320 days, 8 hours, 49 minutes.
[Worker #1 Apr 12 23:41] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 3 threads
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 56.576 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 56.717 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 56.660 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 56.772 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 56.694 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 56.740 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 56.929 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 56.590 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 56.704 ms.
[Worker #1 Apr 12 23:41] p: 332220523.  Time: 56.882 ms.
[Worker #1 Apr 12 23:41] Iterations: 10.  Total time: 0.567 sec.
[Worker #1 Apr 12 23:41] Estimated time to complete this exponent: 218 days, 2 hours, 54 minutes.
[Worker #1 Apr 12 23:42] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 4 threads
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 43.378 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 43.511 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 43.436 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 43.516 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 43.662 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 43.449 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 43.516 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 43.520 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 43.506 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 43.566 ms.
[Worker #1 Apr 12 23:42] Iterations: 10.  Total time: 0.435 sec.
[Worker #1 Apr 12 23:42] Estimated time to complete this exponent: 167 days, 6 hours, 53 minutes.
[Worker #1 Apr 12 23:42] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 5 threads
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.449 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.723 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.554 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.894 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.744 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.658 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.724 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.602 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.680 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.864 ms.
[Worker #1 Apr 12 23:42] Iterations: 10.  Total time: 0.357 sec.
[Worker #1 Apr 12 23:42] Estimated time to complete this exponent: 137 days, 5 hours, 31 minutes.
[Worker #1 Apr 12 23:42] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 6 threads
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 31.922 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 32.019 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 31.831 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 31.855 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 32.167 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 31.813 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 31.897 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 32.012 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 31.765 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 31.850 ms.
[Worker #1 Apr 12 23:42] Iterations: 10.  Total time: 0.319 sec.
[Worker #1 Apr 12 23:42] Estimated time to complete this exponent: 122 days, 17 hours, 2 minutes.
[Worker #1 Apr 12 23:42] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 7 threads
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 29.776 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 31.162 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 31.822 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 30.880 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 30.634 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 30.922 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 31.070 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 30.778 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 31.368 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 31.456 ms.
[Worker #1 Apr 12 23:42] Iterations: 10.  Total time: 0.310 sec.
[Worker #1 Apr 12 23:42] Estimated time to complete this exponent: 119 days, 3 hours, 34 minutes.
[Worker #1 Apr 12 23:42] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 8 threads
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 27.948 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 28.502 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 28.472 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 28.460 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 29.614 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 29.285 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 30.282 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 29.652 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 29.499 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 29.594 ms.
[Worker #1 Apr 12 23:42] Iterations: 10.  Total time: 0.291 sec.
[Worker #1 Apr 12 23:42] Estimated time to complete this exponent: 112 days, 0 hours, 17 minutes.
[Worker #1 Apr 12 23:42] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 2 threads
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 196.411 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 196.614 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 196.611 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 197.224 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 196.596 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 196.749 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 198.562 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 196.610 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 196.655 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 196.652 ms.
[Worker #1 Apr 12 23:42] Iterations: 10.  Total time: 1.969 sec.
[Worker #1 Apr 12 23:42] Estimated time to complete this exponent: 756 days, 23 hours, 42 minutes.
[Worker #1 Apr 12 23:42] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 4 threads
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 98.860 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 98.221 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 98.272 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 98.976 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 98.953 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 98.882 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 98.831 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 98.510 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 98.804 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 99.006 ms.
[Worker #1 Apr 12 23:42] Iterations: 10.  Total time: 0.987 sec.
[Worker #1 Apr 12 23:42] Estimated time to complete this exponent: 379 days, 15 hours, 17 minutes.
[Worker #1 Apr 12 23:42] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 6 threads
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 66.049 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 66.010 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 66.179 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 66.147 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 66.124 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 66.015 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 66.104 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 66.057 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 66.152 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 66.037 ms.
[Worker #1 Apr 12 23:42] Iterations: 10.  Total time: 0.661 sec.
[Worker #1 Apr 12 23:42] Estimated time to complete this exponent: 254 days, 2 hours, 46 minutes.
[Worker #1 Apr 12 23:42] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 8 threads
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 49.914 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 50.100 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 50.127 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 50.256 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 50.139 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 50.043 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 50.138 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 50.153 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 50.192 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 50.168 ms.
[Worker #1 Apr 12 23:42] Iterations: 10.  Total time: 0.501 sec.
[Worker #1 Apr 12 23:42] Estimated time to complete this exponent: 192 days, 17 hours, 31 minutes.
[Worker #1 Apr 12 23:42] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 10 threads
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 40.508 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 40.519 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 40.602 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 40.624 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 40.624 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 40.731 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 40.559 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 40.623 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 40.628 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 40.722 ms.
[Worker #1 Apr 12 23:42] Iterations: 10.  Total time: 0.406 sec.
[Worker #1 Apr 12 23:42] Estimated time to complete this exponent: 156 days, 4 hours, 0 minutes.
[Worker #1 Apr 12 23:42] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 12 threads
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.039 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 34.998 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 34.975 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.183 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.046 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.158 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.303 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.309 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 34.864 ms.
[Worker #1 Apr 12 23:42] p: 332220523.  Time: 35.168 ms.
[Worker #1 Apr 12 23:42] Iterations: 10.  Total time: 0.351 sec.
[Worker #1 Apr 12 23:42] Estimated time to complete this exponent: 134 days, 23 hours, 33 minutes.
[Worker #1 Apr 12 23:42] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 14 threads
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 32.017 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 32.452 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 31.986 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 32.418 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 32.196 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 31.831 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 32.025 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 32.195 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 32.260 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 31.923 ms.
[Worker #1 Apr 12 23:43] Iterations: 10.  Total time: 0.321 sec.
[Worker #1 Apr 12 23:43] Estimated time to complete this exponent: 123 days, 13 hours, 5 minutes.
[Worker #1 Apr 12 23:43] Using FMA3 FFT length 18M, Pass1=1536, Pass2=12K, 16 threads
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 30.566 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 30.904 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 30.629 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 30.742 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 30.836 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 30.854 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 30.650 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 30.884 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 30.660 ms.
[Worker #1 Apr 12 23:43] p: 332220523.  Time: 31.129 ms.
[Worker #1 Apr 12 23:43] Iterations: 10.  Total time: 0.308 sec.
[Worker #1 Apr 12 23:43] Estimated time to complete this exponent: 118 days, 8 hours, 59 minutes.
db597 is offline   Reply With Quote
Old 2017-04-12, 16:01   #156
db597
 
db597's Avatar
 
Jan 2003

3138 Posts
Default

Quote:
Originally Posted by pinhodecarlos View Post
What's the power consumption on both processors whilst doing the same type of work at full CPU occupancy? What's the overall investment for each type of machine?
Power consumption of my Ryzen is as follows, but also note that I've chosen to run it at only 3.32GHz for efficiency reasons. I have successfully overclocked it to 3.975GHz, but power draw is more than double at that speed just not worth it.

- Ryzen 1700 (non-X) running full load (16 threads)
- All 8 cores all clocked at @ 3.32GHz (33.25x multiplier)
- VCore 1.031V / VSoC auto
- Power draw of all cores + SoC = 64.5W (as reported by HWiNFO)

In terms of cost, when I purchased the Ryzen I was comparing the following CPU + motherboard combo deals (the Ryzen was just a little cheaper):

1) Ryzen 1700 + Asus X370-Pro... S$735
2) Intel 7700K + Asus Z270-K... S$761
db597 is offline   Reply With Quote
Old 2017-04-22, 12:38   #157
nordi
 
Dec 2016

71 Posts
Default

Has anyone benchmarked ECM or P-1 factoring? At least for smaller numbers, they require a lot less RAM bandwidth than LL tests, so I'd expect Zen CPUs to perform well.
nordi is offline   Reply With Quote
Old 2017-05-02, 20:23   #158
mackerel
 
mackerel's Avatar
 
Feb 2016
UK

1101000112 Posts
Default

Agner Fog has just released an updated architecture guide including Ryzen. Some fun reading.

http://www.agner.org/optimize/blog/read.php?i=838
http://www.agner.org/optimize/

Summary at 1st link, detail download item 3 on 2nd link.
mackerel is offline   Reply With Quote
Old 2017-05-05, 21:16   #159
airsquirrels
 
airsquirrels's Avatar
 
"David"
Jul 2015
Ohio

11×47 Posts
Default

I needed to build a new GPU Dev Box, and I decided to try the Ryzen route.

I went with the 1800X with ASUS Prime X370-Pro. The GPU in it is currently a Radeon Pro Duo.

Initially the Motherboard arrived with the 0504 Bios and I had very significant problems getting the system to even POST. Eventually after a serious of reboots and hand-waving I heard a beep from the speaker and was able to enter the BIOS and get an update to 0604.

After this BIOS update things seem pretty stable. Running some prime95 and other tests now, although I do not expect the results to differ from the others that have been posted.
airsquirrels is offline   Reply With Quote
Old 2017-05-06, 11:21   #160
tului
 
Jan 2013

22×17 Posts
Default

Quote:
Originally Posted by airsquirrels View Post
I needed to build a new GPU Dev Box, and I decided to try the Ryzen route.

I went with the 1800X with ASUS Prime X370-Pro. The GPU in it is currently a Radeon Pro Duo.

Initially the Motherboard arrived with the 0504 Bios and I had very significant problems getting the system to even POST. Eventually after a serious of reboots and hand-waving I heard a beep from the speaker and was able to enter the BIOS and get an update to 0604.

After this BIOS update things seem pretty stable. Running some prime95 and other tests now, although I do not expect the results to differ from the others that have been posted.
I was originally planning to double my motherboard cost and get the $300 MSI X370 just to have the USB based bios flash option because I had a feeling updates might be hairy. Went with the X370 SLI Plus instead as it was in stock. I'm hoping there is another wave of 2.0 board with more high end features and I'll likely relegate this board to a Ryzen 3 or 5 CPU and quad port NIC for running pfsense.
tului is offline   Reply With Quote
Old 2017-05-15, 03:41   #161
ewmayer
2ω=0
 
ewmayer's Avatar
 
Sep 2002
República de California

24×727 Posts
Default

I posted my Mlucas AVX2/FMA3 timings on David's [airsquirrels] Ryzen here: http://www.mersenneforum.org/showpos...4&postcount=58

I don't get very good scaling beyond 4-threads, possibly due to not properly taking account of the NUMA-ness of the 8-core architecture, but 1-4 thread timings look pretty good. Based on my subsequent running-DCs-on-all-8-cores timings (@2816K, 2-thread time rises from 16.7 msec to 21 msec with 4 such jobs running), we should add ~25% to the 1 and 2-threaded timings to get the system-fully-loaded one. Thus we estimate the following max-throughputs:

Code:
	1-thread
FFT(K) msec/iter iters/sec @full load:
 1024	10.42	614.2
 1152	12.14	527.2
 1280	13.23	483.7
 1408	15.40	415.6
 1536	15.96	401.0
 1664	18.57	344.6
 1792	18.67	342.8
 1920	21.53	297.3
 2048	21.68	295.2
 2304	25.47	251.3
 2560	27.57	232.1
 2816	32.14	199.1
 3072	33.18	192.9
 3328	38.69	165.4
 3584	39.13	163.6
 3840	44.07	145.2
 4096	44.67	143.3
 4608	51.83	123.5
 5120	56.91	112.5
 5632	66.01	 97.0
 6144	68.74	 93.1
 6656	79.48	 80.5
 7168	80.03	 80.0
 7680	89.57	 71.5

Last fiddled with by ewmayer on 2017-08-20 at 21:19 Reason: sec -> msec
ewmayer is offline   Reply With Quote
Old 2017-05-28, 15:23   #162
Mark Rose
 
Mark Rose's Avatar
 
"/X\(‘-‘)/X\"
Jan 2013

3×977 Posts
Default

http://wccftech.com/amd-threadripper...ors-x399-x390/

It seems there will be 10, 12, 14, and 16 core Ryzen chips with quad channel memory.

I bet the 12 core version will have 3 working cores per CCX, but the 10 and 14 core counts cannot be balanced equally.
Mark Rose is offline   Reply With Quote
Old 2017-08-20, 20:20   #163
sanaris
 
"Yury Vorobyov"
Jul 2013
Chelyabinsk

19 Posts
Default

Quote:
Originally Posted by ewmayer View Post
I don't get very good scaling beyond 4-threads, possibly due to not properly taking account of the NUMA-ness
What is the way for NUMA-oriented builds? Run LL test with OpenMPI? :)
sanaris is offline   Reply With Quote
Old 2017-08-20, 21:19   #164
henryzz
Just call me Henry
 
henryzz's Avatar
 
"David"
Sep 2007
Cambridge (GMT/BST)

2·5·587 Posts
Default

Quote:
Originally Posted by sanaris View Post
What is the way for NUMA-oriented builds? Run LL test with OpenMPI? :)
A test per NUMA node is probably optimal.
henryzz is offline   Reply With Quote
Old 2017-08-20, 21:27   #165
ewmayer
2ω=0
 
ewmayer's Avatar
 
Sep 2002
República de California

1163210 Posts
Default

Quote:
Originally Posted by henryzz View Post
A test per NUMA node is probably optimal.
As I noted in my previous post in this thread, I get best performance by run one single-thread job per physical core of Ryzen. As described on my Mlucas README page, due to AMD's logical-core numbering system (which differs from Intel's), this means running such tasks on either the odd or even-numbered cores*, e.g. on the Ryzen airsquirrels lets me use, my 8 Mlucas jobs use '-cpu n' with n running from 0 to 14 in increments of 2, here I have 8 rundirs named run0-7, each with a copy of the post-build self-test-created mlucas.cfg file and its own worktodo.ini file, in each I run a copy of the AVX2-build binary:

cd run0 && nohup nice ../obj_avx2_pthr/Mlucas -cpu 0 &
cd run1 && nohup nice ../obj_avx2_pthr/Mlucas -cpu 2 &
cd run2 && nohup nice ../obj_avx2_pthr/Mlucas -cpu 4 &
cd run3 && nohup nice ../obj_avx2_pthr/Mlucas -cpu 6 &
cd run4 && nohup nice ../obj_avx2_pthr/Mlucas -cpu 8 &
cd run5 && nohup nice ../obj_avx2_pthr/Mlucas -cpu 10 &
cd run6 && nohup nice ../obj_avx2_pthr/Mlucas -cpu 12 &
cd run7 && nohup nice ../obj_avx2_pthr/Mlucas -cpu 14 &

-----------
* I suppose you could mix odds and evens as long as each of the [2k,2k+1] logical-core-index pairs mapping to the physical cores of the system has just one of said pair being used by a task, but I see little point in such odd/even mixing.

Last fiddled with by ewmayer on 2017-08-20 at 22:55
ewmayer is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Intel Processor Speculations Mark Rose Hardware 109 2017-10-13 16:55
Cannonlake speculations henryzz Hardware 0 2017-03-03 19:49

All times are UTC. The time now is 15:27.

Mon May 17 15:27:54 UTC 2021 up 39 days, 10:08, 0 users, load averages: 1.92, 2.17, 2.32

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, 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.