N00B Question about system with multiple processors

[QuickCoder]

Hi all, New to the forum. I have 2 quad core processors in my system.

I have recently added the second one.

Is there a way to run the 4 exponents I already have on both processors giving 2 cores to each?

Or is there a better way to run with this system?


Thanks

[legendarymudkip]

Depending on the cache size of each CPU, it may be faster to run one thing on 4 threads, due to memory speeds and cache stuff. To assign more than one core to a task, click "Test", then "Worker Windows...". On here, you will see a menu with some numbers and other things. If you want to run 2 threads for 2 workers, change "Number of worker windows to run:" to 2 and "CPUs to use (multithreading)" to 2 as well. I would also test 1 worker and 4 CPUs to use, and compare the overall throughput to see which is faster overall.

EDIT: If there are 2 processors in it, double the numbers I said for the threads for the specific worker numbers. I don't know if it works, but if not then there (most likely) isn't a way to do that at the current time. It'll most likely be due to the OS/program not being able to access or recognise the second CPU, if I understand your setup correctly.

[QuickCoder]

Thank you for the reply.
I'm using a Dell Precision T5500 with 2x Quad Core Xeon 2.4

The current number of worker windows is set to 4 running on only 1 of the CPUs.

"CPUs to use" was set at 1 so if I am correct in saying I should change this to 2 and keep the worker windows at 4 it should run 2 cores for each worker?

Many thanks

[legendarymudkip]

I think it should do, yes.

[QuickCoder]

Great thanks, I will try it out.

[TheMawn]

I have often wondered about dual CPU systems...

Could you do me a favour? Copy Prime95.exe to a different location, open it, and run the benchmark utility. At the same time, I could tell you if you're better off running a number on each core or if you should split them up based on the results. Attach the results.txt that gets generated in the same place as Prime95.exe or put the outpit in CODE brackets.

[QuickCoder]

I will do that of course no problem, It will take me some time as I have other things to do this week.

[QuickCoder]

I have attached that for you now. I am unable to tell if it ran on both processors.

[MatWur-S530113]

hm, I never had a dual-processor system and I never had a processor with HT.
But yes, it seems that prime95 only detected 1 of the 2 processors. Are both processors being detected by
a) the BIOS ?
b) the OS ? Maybe you have to re-install the OS that both processors are able to run... (but not sure)

Could you tell us something more about the system (installed Memory, Graficcard).
This infos could be helpful to select the best kind of work for such a machine.
F.e. if there is enough memory some stages 2 of P-1-factoring could be done parallel.

Matthias

[QuickCoder]

My LL test is using both processors so I am not sure what the problem is.

Its running 2 cores per worker and 4 workers. The OS and BIOS both detect the 2 processors.

24GB Ram and Nvidia Quadro FX1800

[TheMawn]

So the two CPU's are dual-core with hyperthreading?

It looks to me like you should be fine. No serious bottlenecking that I can see.

However... Hyperthreading looks like it seriously fudges things up...

Best time for 1024K FFT length: 47.374 ms., avg: 64.597 ms.
Best time for 1280K FFT length: 86.777 ms., avg: 87.307 ms.
Best time for 1536K FFT length: 26.086 ms., avg: 26.290 ms.
Best time for 1792K FFT length: 31.247 ms., avg: 31.473 ms.
Best time for 2048K FFT length: 100.781 ms., avg: 101.256 ms.
Best time for 2560K FFT length: 44.771 ms., avg: 65.440 ms.
Best time for 3072K FFT length: 120.203 ms., avg: 120.837 ms.
Best time for 3584K FFT length: 65.449 ms., avg: 65.800 ms.
Best time for 4096K FFT length: 73.645 ms., avg: 74.196 ms.
Best time for 5120K FFT length: 94.953 ms., avg: 95.302 ms.
Best time for 6144K FFT length: 118.777 ms., avg: 119.645 ms.
Best time for 7168K FFT length: 273.774 ms., avg: 275.301 ms.
Best time for 8192K FFT length: 226.844 ms., avg: 227.370 ms.

As far as I can tell, this only happens when two workers run on one core.