Two types of benchmarks run for a dual core?

2009-03-24, 22:43

Hi when I run the benchmark, I get two readings, one for cpu O and one for cpu 1
Why are there two times like this, and when I compare these times to the times in the benchmark reporting, the one cpu sucks bad and the other is really really good?
I would appreciate any help on this, thanks!

[Mar 24 18:39] Worker starting
[Mar 24 18:39] Setting affinity to run worker on logical CPU #0
[Mar 24 18:39] Your timings will be written to the results.txt file.
[Mar 24 18:39] Compare your results to other computers at http://www.mersenne.org/report_benchmarks
[Mar 24 18:39] Timing 49 iterations at 768K FFT length. Best time: 20.178 ms.
[Mar 24 18:39] Timing 42 iterations at 896K FFT length. Best time: 24.204 ms.
[Mar 24 18:39] Timing 37 iterations at 1024K FFT length. Best time: 27.340 ms.
[Mar 24 18:39] Timing 29 iterations at 1280K FFT length. Best time: 34.537 ms.
[Mar 24 18:39] Timing 24 iterations at 1536K FFT length. Best time: 42.223 ms.
[Mar 24 18:39] Timing 21 iterations at 1792K FFT length. Best time: 50.384 ms.
[Mar 24 18:39] Timing 18 iterations at 2048K FFT length. Best time: 55.813 ms.
[Mar 24 18:39] Timing 14 iterations at 2560K FFT length. Best time: 73.970 ms.
[Mar 24 18:39] Timing 12 iterations at 3072K FFT length. Best time: 90.392 ms.
[Mar 24 18:39] Timing 10 iterations at 3584K FFT length. Best time: 107.726 ms.
[Mar 24 18:39] Timing 10 iterations at 4096K FFT length. Best time: 120.210 ms.
[Mar 24 18:40] Timing 10 iterations at 5120K FFT length. Best time: 153.859 ms.
[Mar 24 18:40] Timing 10 iterations at 6144K FFT length. Best time: 185.918 ms.
[Mar 24 18:40] Timing 10 iterations at 7168K FFT length. Best time: 225.292 ms.
[Mar 24 18:40] Timing 10 iterations at 8192K FFT length. Best time: 245.962 ms.
[Mar 24 18:40] Timing FFTs using 2 threads.
[Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:40] Timing 49 iterations at 768K FFT length. Best time: 10.752 ms.
[Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:40] Timing 42 iterations at 896K FFT length. Best time: 12.918 ms.
[Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:40] Timing 37 iterations at 1024K FFT length. Best time: 14.758 ms.
[Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:40] Timing 29 iterations at 1280K FFT length. Best time: 17.805 ms.
[Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:40] Timing 24 iterations at 1536K FFT length. Best time: 21.739 ms.
[Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:40] Timing 21 iterations at 1792K FFT length. Best time: 25.815 ms.
[Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:40] Timing 18 iterations at 2048K FFT length. Best time: 28.754 ms.
[Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:40] Timing 14 iterations at 2560K FFT length. Best time: 37.996 ms.
[Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:40] Timing 12 iterations at 3072K FFT length. Best time: 46.389 ms.
[Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:40] Timing 10 iterations at 3584K FFT length. Best time: 55.037 ms.
[Mar 24 18:41] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:41] Timing 10 iterations at 4096K FFT length. Best time: 61.391 ms.
[Mar 24 18:41] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:41] Timing 10 iterations at 5120K FFT length. Best time: 79.163 ms.
[Mar 24 18:41] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:41] Timing 10 iterations at 6144K FFT length. Best time: 96.253 ms.
[Mar 24 18:41] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:41] Timing 10 iterations at 7168K FFT length. Best time: 116.001 ms.
[Mar 24 18:41] Setting affinity to run helper thread 1 on logical CPU #1
[Mar 24 18:41] Timing 10 iterations at 8192K FFT length. Best time: 127.277 ms.
[Mar 24 18:41] Timing trial factoring of M35000011 with 58 bit length factors. Best time: 5.154 ms.
[Mar 24 18:41] Timing trial factoring of M35000011 with 59 bit length factors. Best time: 5.180 ms.
[Mar 24 18:41] Timing trial factoring of M35000011 with 60 bit length factors. Best time: 5.168 ms.
[Mar 24 18:41] Timing trial factoring of M35000011 with 61 bit length factors. Best time: 5.160 ms.
[Mar 24 18:41] Timing trial factoring of M35000011 with 62 bit length factors. Best time: 5.148 ms.
[Mar 24 18:41] Timing trial factoring of M35000011 with 63 bit length factors. Best time: 8.753 ms.
[Mar 24 18:41] Timing trial factoring of M35000011 with 64 bit length factors. Best time: 8.776 ms.
[Mar 24 18:41] Timing trial factoring of M35000011 with 65 bit length factors. Best time: 8.121 ms.
[Mar 24 18:41] Timing trial factoring of M35000011 with 66 bit length factors. Best time: 8.072 ms.
[Mar 24 18:41] Timing trial factoring of M35000011 with 67 bit length factors. Best time: 8.072 ms.
[Mar 24 18:41] Benchmark complete.
[Mar 24 18:41] Worker stopped.

Mini-Geek · 2009-03-24, 23:09

Quote:

Originally Posted by Unregistered

Hi when I run the benchmark, I get two readings, one for cpu O and one for cpu 1
Why are there two times like this, and when I compare these times to the times in the benchmark reporting, the one cpu sucks bad and the other is really really good?
I would appreciate any help on this, thanks!

[Mar 24 18:39] Worker starting
[Mar 24 18:39] Setting affinity to run worker on logical CPU #0
[Mar 24 18:39] Your timings will be written to the results.txt file.
[Mar 24 18:39] Compare your results to other computers at http://www.mersenne.org/report_benchmarks
...
[Mar 24 18:40] Timing FFTs using 2 threads.
[Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1

The first test is using only CPU #0, the second is using both CPUs 0 and 1, so in almost all cases the second will be faster, and this is perfectly normal.

cheesehead · 2009-03-25, 06:30

Quote:

Originally Posted by Unregistered

Hi when I run the benchmark, I get two readings, one for cpu O and one for cpu 1
Why are there two times like this, and when I compare these times to the times in the benchmark reporting, the one cpu sucks bad and the other is really really good?
I would appreciate any help on this, thanks!

It's easy to misunderstand what the test was doing after:

"Timing FFTs using 2 threads.
Setting affinity to run helper thread 1 on logical CPU #1".

In addition to Mini-Geek's answer:

On a multi-CPU system, Prime95 can perform many of its calculations (the ones involving FFTs) either by using just one thread on one CPU, or by splitting the work between two (or more) threads on different CPUs. The first part of the benchmark is performed using one thread on one CPU, #0. (It's not necessary to perform the single-thread benchmark again on CPU #1 because normally its performance would be just the same as for CPU #0.)

Then, the benchmark starts a "helper thread" on the second CPU ("logical CPU #1"), in addition to the main thread that was already running on CPU #0. It re-runs the same calculations it did before, except that they are split between the two threads, so each CPU performs about half of the work. This results in finishing each FFT in slightly more than half the time it took the single-thread benchmark. (The reason the dual-thread time is always a bit more than half of the single-thread time is that the FFT calculation can't be totally split evenly 50%-50% between threads.)

Sure enough, you see that the single-thread iteration for 768K FFT was 20.178 ms, then the dual-thread iteration for the same FFT was 10.752 ms, close to half the time.

CADavis · 2009-03-26, 07:22

It uses both threads for one test.

2009-03-26, 14:06

Thanks so much! I kinda suspected. I did read a bit about worker threads and tasks or something, but I unfortunately I am not up to speed about computer science.

Take care to everyone who answered my question!

2012-12-31, 07:39

How do you run the same test on both cores? Would you just put the same test for both workers in worktodo.txt?

Dubslow · 2012-12-31, 19:40

No. In Prime95, in one of the menu options (I forget which) is an option called "threads per test" or perhaps "threads per worker". Set that to two, and set total workers to one.

2013-01-01, 07:00

Found it. In the workers window. Set "number of worker windows to run" to run at 1. Then "cpus to use(multithread)" to 2. Figured id post the actual option names incase anyone else had the same question. That brought my eta on my 55million ll test down a little over two months. Once my new tower gets here my new mb is going in and my OC on my E5300 begins. anxious to see my eta then :)

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
How to retire one core in a dual-core CPU?	Rodrigo	PrimeNet	4	2011-07-30 14:43
Dual Core to Quad Core Upgrade	Rodrigo	Hardware	6	2010-11-29 18:48
benchmarks on dual i7-xeon	fivemack	Msieve	1	2009-12-14 12:51
Importance of dual channel memory for dual core processors	patrik	Hardware	3	2007-01-07 09:26
Dual Core? BFD	R.D. Silverman	Hardware	12	2005-02-20 21:46

2009-03-24, 22:43	#1
Unregistered 2×29×47 Posts	Two types of benchmarks run for a dual core? Hi when I run the benchmark, I get two readings, one for cpu O and one for cpu 1 Why are there two times like this, and when I compare these times to the times in the benchmark reporting, the one cpu sucks bad and the other is really really good? I would appreciate any help on this, thanks! [Mar 24 18:39] Worker starting [Mar 24 18:39] Setting affinity to run worker on logical CPU #0 [Mar 24 18:39] Your timings will be written to the results.txt file. [Mar 24 18:39] Compare your results to other computers at http://www.mersenne.org/report_benchmarks [Mar 24 18:39] Timing 49 iterations at 768K FFT length. Best time: 20.178 ms. [Mar 24 18:39] Timing 42 iterations at 896K FFT length. Best time: 24.204 ms. [Mar 24 18:39] Timing 37 iterations at 1024K FFT length. Best time: 27.340 ms. [Mar 24 18:39] Timing 29 iterations at 1280K FFT length. Best time: 34.537 ms. [Mar 24 18:39] Timing 24 iterations at 1536K FFT length. Best time: 42.223 ms. [Mar 24 18:39] Timing 21 iterations at 1792K FFT length. Best time: 50.384 ms. [Mar 24 18:39] Timing 18 iterations at 2048K FFT length. Best time: 55.813 ms. [Mar 24 18:39] Timing 14 iterations at 2560K FFT length. Best time: 73.970 ms. [Mar 24 18:39] Timing 12 iterations at 3072K FFT length. Best time: 90.392 ms. [Mar 24 18:39] Timing 10 iterations at 3584K FFT length. Best time: 107.726 ms. [Mar 24 18:39] Timing 10 iterations at 4096K FFT length. Best time: 120.210 ms. [Mar 24 18:40] Timing 10 iterations at 5120K FFT length. Best time: 153.859 ms. [Mar 24 18:40] Timing 10 iterations at 6144K FFT length. Best time: 185.918 ms. [Mar 24 18:40] Timing 10 iterations at 7168K FFT length. Best time: 225.292 ms. [Mar 24 18:40] Timing 10 iterations at 8192K FFT length. Best time: 245.962 ms. [Mar 24 18:40] Timing FFTs using 2 threads. [Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:40] Timing 49 iterations at 768K FFT length. Best time: 10.752 ms. [Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:40] Timing 42 iterations at 896K FFT length. Best time: 12.918 ms. [Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:40] Timing 37 iterations at 1024K FFT length. Best time: 14.758 ms. [Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:40] Timing 29 iterations at 1280K FFT length. Best time: 17.805 ms. [Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:40] Timing 24 iterations at 1536K FFT length. Best time: 21.739 ms. [Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:40] Timing 21 iterations at 1792K FFT length. Best time: 25.815 ms. [Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:40] Timing 18 iterations at 2048K FFT length. Best time: 28.754 ms. [Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:40] Timing 14 iterations at 2560K FFT length. Best time: 37.996 ms. [Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:40] Timing 12 iterations at 3072K FFT length. Best time: 46.389 ms. [Mar 24 18:40] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:40] Timing 10 iterations at 3584K FFT length. Best time: 55.037 ms. [Mar 24 18:41] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:41] Timing 10 iterations at 4096K FFT length. Best time: 61.391 ms. [Mar 24 18:41] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:41] Timing 10 iterations at 5120K FFT length. Best time: 79.163 ms. [Mar 24 18:41] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:41] Timing 10 iterations at 6144K FFT length. Best time: 96.253 ms. [Mar 24 18:41] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:41] Timing 10 iterations at 7168K FFT length. Best time: 116.001 ms. [Mar 24 18:41] Setting affinity to run helper thread 1 on logical CPU #1 [Mar 24 18:41] Timing 10 iterations at 8192K FFT length. Best time: 127.277 ms. [Mar 24 18:41] Timing trial factoring of M35000011 with 58 bit length factors. Best time: 5.154 ms. [Mar 24 18:41] Timing trial factoring of M35000011 with 59 bit length factors. Best time: 5.180 ms. [Mar 24 18:41] Timing trial factoring of M35000011 with 60 bit length factors. Best time: 5.168 ms. [Mar 24 18:41] Timing trial factoring of M35000011 with 61 bit length factors. Best time: 5.160 ms. [Mar 24 18:41] Timing trial factoring of M35000011 with 62 bit length factors. Best time: 5.148 ms. [Mar 24 18:41] Timing trial factoring of M35000011 with 63 bit length factors. Best time: 8.753 ms. [Mar 24 18:41] Timing trial factoring of M35000011 with 64 bit length factors. Best time: 8.776 ms. [Mar 24 18:41] Timing trial factoring of M35000011 with 65 bit length factors. Best time: 8.121 ms. [Mar 24 18:41] Timing trial factoring of M35000011 with 66 bit length factors. Best time: 8.072 ms. [Mar 24 18:41] Timing trial factoring of M35000011 with 67 bit length factors. Best time: 8.072 ms. [Mar 24 18:41] Benchmark complete. [Mar 24 18:41] Worker stopped.

2009-03-26, 07:22	#4
CADavis Jul 2005 Des Moines, Iowa, USA 2×5×17 Posts	It uses both threads for one test.

2009-03-26, 14:06	#5
Unregistered 3·5·7³ Posts	Thanks so much! I kinda suspected. I did read a bit about worker threads and tasks or something, but I unfortunately I am not up to speed about computer science. Take care to everyone who answered my question!

2012-12-31, 07:39	#6
Unregistered 2×7×419 Posts	How do you run the same test on both cores? Would you just put the same test for both workers in worktodo.txt?

2012-12-31, 19:40	#7
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3·29·83 Posts	No. In Prime95, in one of the menu options (I forget which) is an option called "threads per test" or perhaps "threads per worker". Set that to two, and set total workers to one.

2013-01-01, 07:00	#8
Unregistered 6,967 Posts	Found it. In the workers window. Set "number of worker windows to run" to run at 1. Then "cpus to use(multithread)" to 2. Figured id post the actual option names incase anyone else had the same question. That brought my eta on my 55million ll test down a little over two months. Once my new tower gets here my new mb is going in and my OC on my E5300 begins. anxious to see my eta then :)