-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
prime 95 not working rite?
- Thread starter d94
- Start date
- Joined
- Jul 31, 2005
errrm..
get rid of it.... download the latest... redo it
BTW
all you need to know is if you stopped it... and it didnt stop itself by an error... THATS A GOOD THING!
get rid of it.... download the latest... redo it
BTW
all you need to know is if you stopped it... and it didnt stop itself by an error... THATS A GOOD THING!
- Joined
- Apr 22, 2002
- Location
- Austin, Texas
I have to minimize than maximize prime95 to see how long it ran for and if there were any errors. For some reason the text isn't there but when I maximize it again the info shows up. Not sure if that's your problem.
- Joined
- May 30, 2004
d94 said:well, i ran prime while i slept all nite (6hrs)
but when i stopped the test it didnt have anything calculated or w/e it does! no errors.. but no new lines either...
whats goin on here!?
Windows may have locked up. Were you still able to move the mouse pointer?
If you weren't, Windows probably locked up on you. Lock up when running Prime95=bad, probably means that it's overheating or at a wall.
I've found that unless I'm running v2.46, or v24.6(not sure..lol) that the same happens. Or similar. It will not max out my cpu usage. It will spike to 100% and drop, spike again and drop, but won't keep it maxxed. As a result, it takes forever to finish a test.
Try searching for Prime 95 v246 and you should be able to find it.
Edit: It's ver 24.6... just checked in the program
Try searching for Prime 95 v246 and you should be able to find it.
Edit: It's ver 24.6... just checked in the program
- Joined
- Oct 7, 2004
- Location
- Cincinnati
Boy 24.6 must be really new. v24.13 was last updated 7/5/05.
/edit: I found it for you. Took more effort than I thought it would. Well looking at the times between my first post and my edit I guess it was under 10 minutes, but in "Google-time" that's an eternity.
http://users.forthnet.gr/lar/papast...ogs/p95v246.zip
/edit: I found it for you. Took more effort than I thought it would. Well looking at the times between my first post and my edit I guess it was under 10 minutes, but in "Google-time" that's an eternity.
http://users.forthnet.gr/lar/papast...ogs/p95v246.zip
Last edited:
d94 said:
As I tried to explain before, perhaps poorly, if you run two copies of prime95 and each is running the blend test, then each will try to use more than half of your physical RAM. This will cause thrashing - the OS swaps RAM out to hard disk. Since hard disk is much, much slower than RAM it takes a long, long time to complete a single test.
The cure is to run two prime95s each running the in-place FFTs. This will use little RAM and avoid thrashing. Alternatively, you could run two prime95 custom torture tests. In the torture test dialog box, give each prime95 about 40% of your physical RAM (leaving 20% for Windows).
The "bug" is that prime95 was designed in pre-hyperthreading, single processor days. What I should do is make the blend test by default allocate 80% of RAM divided by the number of physical + virtual processors in the machine.
Why then, wouldn't it work correctly if you are running only one instance of prime 95, as is the case when I get the problems as well. I could see perhaps running 2 instances by mistake once or twice, but not every time I run it. Besides, they would both show up in the task bar.
This also happens on non-hyperthreading, amd cpus in my case. (2500 mobile and 2400 desktop)
This also happens on non-hyperthreading, amd cpus in my case. (2500 mobile and 2400 desktop)
yo4444 said:
In the blend test, prime95 guesses how much memory it can use without causing thrashing. This is a guess because the OS, drivers, and other programs you might have loaded all eat up some RAM. Prime95 version 23.x was way too aggressive in guessing how much memory it could safely use. Version 24.13 is much more conservative but the guessing code for the blend test may need further work.
The first step is to see if my theory that it is thrashing that is causing the problem is correct. If the in-place torture tests or the custom test with the "memory to use" entry reduced run without problem then this is a thrashing problem. Another sign of thrashing would be hearing disk activity during the slow torture test.
If thrashing is not the cause, then either you have a machine that is past its limits or you have a brand new prime95 bug. If the same torture test works at slower clock speeds then it is the former, otherwise it is the latter.
