![]() |
![]() |
#1 |
77448 Posts |
![]()
Hi guys,
I've got an assigment few day ago through Test/Worker Windows... menu. I wanted to use just one defined core for this work (#4). So I did. Now I have 1 worker making his hard job on 4th core. But now I'm assuming that worker#1 may lack for life duration. So can I spread the work over one two or three other cores? I'm interesting to make it in night hours. |
![]() |
![]() |
#2 |
24×563 Posts |
![]()
Still no help... I've been trying to make some research meantime and found that I need to run helper threads to involve other cores in the process. I tried different changes in local.txt file (Affinity= and ThreadsPerTest= ) but it doesn't add any helper threads...
Any help appreciated... Thanks |
![]() |
![]() |
#3 |
6809 > 6502
"""""""""""""""""""
Aug 2003
101×103 Posts
23×457 Posts |
![]()
Are you trying to work on only 1 number as fast as possible?
Is your processor 4 cores? If so, set "Number of worker windows to run" to 1 and "Multithreading (CPUs to use)" to 4. And let it run on any CPU (don't assign it). If that is not what you want try posting your request again, either reworded, or in you mother language. |
![]() |
![]() |
![]() |
#4 |
2·23·157 Posts |
![]()
Hi, thank you for reply and sorry for my poor English. I doubt there is somebody Russians so will try my best to explain my problem. Actually you understood it right. I have 4 core CPU and I want to run only one test using all cores at night and one or two cores daytime. In otherwords, I need to be able to switch between these two states. I have run a task already and I defined certain one core for this. I have tried to change "Multithreading (CPUs to use)" to 4 already and it split a work between all cores but overall efficiency not change. It stays around 25% which seems odd to me. I tried to run other prime95 instance separately and received another assignment. I put "Multithreading (CPUs to use)" to 4 there and it worked just perfect. So why can't I reach the same in my main task? Maybe it can't be changed to multithread once chosen or it depend on what kind of test/work I assigned on?
Waiting for you farther explanations, sorry again for my language and thank you! |
![]() |
![]() |
#5 |
6809 > 6502
"""""""""""""""""""
Aug 2003
101×103 Posts
23·457 Posts |
![]()
There are some Russian speakers around here.
![]() You should not normally need to change the number of cores from day to night. I don't know if it is possible. Someone else will be able to tell you for sure and if so, how. Last fiddled with by Uncwilly on 2009-11-01 at 06:04 |
![]() |
![]() |
![]() |
#6 |
"GIMFS"
Sep 2002
Oeiras, Portugal
29×3 Posts |
![]()
The Prime95 program runs always at the lowest priority, which means that the CPU only use the "spare cycles". Whenever other task needs processor power it will get it. So you do not have to worry about the number of cores it uses.
If your goal is to perform a Lucas-Lehmer test in the shortest time possible, throw all cores at it: in the Prime95 window Select Test->Worker windows, and then set Number of worker windows to run to 1, Type of work to get to First time tests, CPU affinity to Run on any CPU, and Multithreading (CPUs to use) to 4. Do not edit manually the configuration files. This should have you set. If you already have a second exponent assigned that you do not want to test, use the Advanced->Unreserve Exponent option. Tell us how it went. Good luck and enjoy the search. |
![]() |
![]() |
![]() |
#7 |
2×5×7×109 Posts |
![]()
Thank you, looks like it works:)
|
![]() |
![]() |
#8 |
"GIMFS"
Sep 2002
Oeiras, Portugal
153610 Posts |
![]()
OK, that´s fine.
Now note that although its true that the fastest way of testing a number is to use the maximum number of cores for the test, it may not be the most productive in terms of the overall project throughput. This is due to limitations of the hardware architecture, mainly memory bandwidth. This means that using 4 cores on a test doesn´t make it run 4 times faster. If you´re using a Core2 Quad, it´s more efficient to run 2 tests (2 worker windows) each one using 2 threads. The Core i7 (or i5) suffers less from this problem, so using 4 cores on a single test it´s not so big a waste. AFAIK the AMD CPUs (Athlon II, Phenom,...) have better mem bandwudth than the Core2, but it´s still better to run 2 tests with 2 threads each than a single test with 4 threads. There are also other possible combinations, which you may wish to try in the future, like using 3 cores for the LL test and the 4th for Trial Factoring, or running 4 tests each one single threaded. Give it a try, it´s good fun and you´l end up learning something more about your machine, while at the same time you contribute to the GIMPS project. |
![]() |
![]() |
![]() |
Thread Tools | |
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
Best use of six cores for L-L testing | Chuck | Hardware | 4 | 2011-05-10 03:04 |
32 cores limitation | gabrieltt | Software | 12 | 2010-07-15 10:26 |
Running on 4 Cores | Unregistered | Information & Answers | 9 | 2008-09-25 00:53 |
6 Intel Cores | petrw1 | Hardware | 3 | 2008-09-16 16:33 |
A program that uses all the CPU-cores | Primix | Hardware | 7 | 2008-09-06 21:09 |