Prime95 question....

[Trader05]

I just got a new chip, 2.8c, i was just going to test to see what i can do with it. When i start Prime95 torture test, it always closes out, and windows says that the swap file is low. Does Prime95 need a swap file, because i currently don't have any swap file set in WinXp, i didn't think i needed one cause i have 1gb of ram. Any hints/suggestions/comments, all highly appreciated!

 

[BlueWeasel]

That's odd...I've never heard of that before.

Are you using the latest version of P95? The newer versions allow you to control the type of torture test you want to run. One of the options allows you to load the CPU only, and pretty much leave RAM out of it. Not sure if it would help your problem, but worth a shot?

 

[Don66]

Choose (In-Place Large FFT's.)
See how that works for you.

 

[magratton]

And setup that swap file! 1-2GB or let windows manage the size! Also, for prime95 I believe the recommended test is the "Custom" then set the min fft to 2048 leave rest of settings alone. That has proven to me and many others on the board to be a rapid identifier of errors if there are going to be any.

 

[irenealan]

I was just wondering the same so here's my setting for the custom test, let me know if I need to change anything thanks:

Min FFT size (in K): 2048
Max FFT size (in K): 4096
[ ] Run FFTs in-place
Memory to use (in MB): 927
Time to run each FFT size (in minutes): 15

Thanks.

Originally posted by: magratton
And setup that swap file! 1-2GB or let windows manage the size! Also, for prime95 I believe the recommended test is the "Custom" then set the min fft to 2048 leave rest of settings alone. That has proven to me and many others on the board to be a rapid identifier of errors if there are going to be any.

 

[THUGSROOK]

Originally posted by: magratton
And setup that swap file!

 

[ectx]

Originally posted by: magratton
And setup that swap file! 1-2GB or let windows manage the size! Also, for prime95 I believe the recommended test is the "Custom" then set the min fft to 2048 leave rest of settings alone. That has proven to me and many others on the board to be a rapid identifier of errors if there are going to be any.

Could have agreeed with you more.

 

[Margalus]

don't run the latest version of prime, it is very buggy and doesn't work for a LOT of people because of a pagefile bug. You should download and use the 22.13 version here

 

[ectx]

What is the pagefile bug? I ran it 23.8 on 4 systems for extensive period of time and never run into anything unusual? I am actually under the impression that 23.7(?) was buggy and 23.8 is ok. Any information, link is helpful. Thanks.

 

[Margalus]

the page file bug that was introduced in the 23.xx version causes computers to actually stop testing the cpu and go into a permanant paging loop that will use up 1gb of page file or more, while the cpu usage goes to zero and stays there.

 

[myocardia]

Originally posted by: Margalus
don't run the latest version of prime, it is very buggy and doesn't work for a LOT of people because of a pagefile bug. You should download and use the 22.13 version here

Use 23.8, 22.1 doesn't even identify stability problems. It was 23.7 that was buggy, 23.8 is the best so far at finding instability.

 

[Margalus]

Originally posted by: myocardia
Use 23.8, 22.1 doesn't even identify stability problems. It was 23.7 that was buggy, 23.8 is the best so far at finding instability.

I am going to have to disagree big time with that statement. 22.13 is by far the best at identifying errors and 23.8 is worthless. I have quite a bit of experience with it, and I have seen quite a bit more info in various overclocking forums that back up my feelings. 22.13 is the best at showing errors, I have never seen it let anyone down yet. Besides, even 23.8 still has the pagefile bug, as I and numerous other people can attest to.

 

[maluckey]

23.8 also has a bug that allows it to suspend FFT testing if you start another intensive application. It WILL disable the XP, CPU idle command by some method, so the CPU reads 100 percent but it will not move on through the tests. At first I thought it was just my imagination, but as I watched the temperatures creep down, not up, and when the test was restarted the temps went back up, my suspicions returned.

I happens when I run Prime95 and running SuperPI every single time. I can also get it to do this with some games running at the same time.

I'm looking for a way to run the FFT test while running other intense apps, and not have Prime95 go on vacation. Any suggestions?

 

[ectx]

Set your priority to 9 to see if it helps. I beleive p95, by defualt, sets the priority to 1(?) like ascreen saver. So if you raun somthing else, p95 only runs when cpu is frrred up by the other app.

In my experience, if you have a HT chip or dual cpus, it does require 2 apps to test the chips. For other chips, it really does not help if you run 2 apps.

Margalus,

What is the sysmtom of the pagefile bug? System hangs or reboots?

When I run large fft custom tests - those are the most likely errors - I don;t get an error message like rounding error. But I never though it was a bug since the system will be stable (no reboot or hang) if I lower the oc. Or maybe reboot and hang are not pagefile bug?

Since I started the large fft custom test, I have lower the oc on all my systems. If the instability is actually a program bug, then it really wastes a lot of my time.

Any link to a thread discussing this problem? Would the bug be more pronounced for large fft's?

 

[Margalus]

with the pagefile bug you will see your cpu usage go to zero and you will see your harddisk is going constantly. If you check task manager you will see that your pagefile is up to 1 gb or more and constantly being accessed. The program just sits there like that, stuck in an infinate loop.

 

[ectx]

Is the pc completely frozen or you can shut down p95 when it happens?

 

[Margalus]

the computer still responds and you can just shut down prime. It just stops testing the cpu is all and starts heating up the harddrive with the constant paging.

 

[ectx]

Thanks for the info. I am too lazy to search for it now (busy modding my mb). I have not personally experienced the problem myself (4 different mb's, 6-7chips, and 4 different pairs memeory sticks in the past month). But I cannot say there is no bug.

Downloaded 22.13 and may do a comparison when I have a chance.

I switched to 23.8 when I had problem w/22.12: " cannot read or write to temp files" in one of my rigs.