What is minimum RAM needed for 6 Core CPU for P-1?

[cheesehead]

Originally Posted by lycorn

2) The parameters referred by garo are to be placed in prime.txt, not local.txt

Originally Posted by odin

Also I tried placing the parameters in prime.txt and they did not work.

File undoc.txt specifies, for each parameter, whether it is to be placed in prime.txt or local.txt (In v24, anyway)

It is hard for us to recall whether any particular parameter goes in prime or local. (There's a pattern, sort of, but ...) I recommend always checking undoc.txt and readme.txt to see what goes where.

Note that MaxHighMemWorkers and OnlyRunStage2WithMaxMemory are not supposed to be placed in the same file:

MaxHighMemWorkers=1

From undoc.txt:

"You can set MaxHighMemWorkers=n in local.txt"

OnlyRunStage2WithMaxMemory=1

From undoc.txt:

"In prime.txt set:
OnlyRunStage2WithMaxMemory=1"

When I placed them in local.txt however (and of course removed it from prime.txt) one worker ran a stage 2 and the other worker moved on to another exponent to start a stage 1.

You might want to move OnlyRunStage2WithMaxMemory=1 back to prime.txt -- otherwise, you might be puzzled by the behavior sometime.

[esqrkim]

I have a dedicated computer with a quad core and 3.5 GB of recognized RAM. I noticed that the max memory used by each core when running P-1 simultaneously was around 330 MB even though I allocated 2800 MB of RAM for Prime95. In retrospect, this limitation might have been caused by the FSB limitation. I am not sure. In any case, I increased the FSB to 424 MB since then, so I'll see what happens next time I do P-1.

[odin]

Originally Posted by esqrkim

I have a dedicated computer with a quad core and 3.5 GB of recognized RAM. I noticed that the max memory used by each core when running P-1 simultaneously was around 330 MB even though I allocated 2800 MB of RAM for Prime95. In retrospect, this limitation might have been caused by the FSB limitation. I am not sure. In any case, I increased the FSB to 424 MB since then, so I'll see what happens next time I do P-1.

You should try what I showed above.

You could get mprime to only do 1 stage 2 P-1 factoring test at a time so all the allocated ram will be dedicated to that. The other workers will move to a Stage 1 P-1 factoring test for another exponent in the meantime. Then when that one is finished another worker thread will go back to a previous exponent and finish Stage 2 and the cycle repeats. You will work through the same amount of exponents but each Stage 2 will have as much ram allocated to it as possible.

in prime.txt add:

OnlyRunStage2WithMaxMemory=1

in local.txt add:

MaxHighMemWorkers=1

Tell me how it works out.

[lfm]

Originally Posted by odin

You should try what I showed above.

You could get mprime to only do 1 stage 2 P-1 factoring test at a time so all the allocated ram will be dedicated to that. The other workers will move to a Stage 1 P-1 factoring test for another exponent in the meantime. Then when that one is finished another worker thread will go back to a previous exponent and finish Stage 2 and the cycle repeats. You will work through the same amount of exponents but each Stage 2 will have as much ram allocated to it as possible.

in prime.txt add:

OnlyRunStage2WithMaxMemory=1

in local.txt add:

MaxHighMemWorkers=1

Tell me how it works out.

If you have 6 cores doing stage 1s and only one core at a time doing stage 2s it will not keep up. Maybe better to have 4 capable of doing stage 2 to make sure it keeps up. 200mb each and 200mb for the OS still is just 1gb.
Should be fine. (fine for Linux, MS-win might want more)

[henryzz]

If you are paying for a 6-core cpu then why wouldn't you get a reasonabe amount of cheap(in comparison to the cpu) memory. I wouldn't suggest less than 3x1GB or 2x2GB depending on whether the cpu has 2 or 3 memory channels. Also try to get memory that runs at a good speed to help LL tests.
With DDR2 I might be more conservative because the pc will then have been designed to be cheap if your using that.