[AlcatrazT2]

Hey everyone:
So yesterday I installed my 3.2c on my old Windows XP Pro w/ SP2 installation...all went well. I went ahead and OC'ed the CPU to 3.6 on stock voltages...all is well.
Yesterday, I was running Prime95, Torture Test, Option 3 (where it blends everything) and each test ran for about 5-6 minutes. All in all, all tests for 1024k, I think it's about six or so, finished in less than 15 minutes.
Today, I was told, that because I have a HT Processor, I need to run two Prime95's at once. When I run both, it's taken about 25+ minutes and neither Test 1 on either of the two has finished.
The computer is fully functionable, even though there is a small lag...
However, CPU is being used at 7% maximum, just PF Usage is high...
My friend, who is doing the same test on his 2.8 HT (He is just running one Prime95 application vs. me running two)...with a clean reformat just like mine, and he said his test has been going on for about one hour and 30 minutes...

This is weird, becuase yesterday, I did the test and it took me much less time. Today, it's taking me the same amount of time plus more to just do one test...any suggestions on why this may be?

The only difference is, today I'm using a brand new format which was done with this 3.2c OC'ed to 3.6...vs. yesterday, when I was using it on the format which I did with a 2.4b OC'ed to 3.0

[mdameron]

i'm pretty sure there is no 'finish line' on prime95, you just run it until you are comfortable that your system is stable.

its also debateable on where or not to run 2 instances of it or not. I just go by how long i can play hl2 and if it crashes or not

[AlcatrazT2]

It's understandeable that there is no time limit for Prime95...the thing which I'm trying to get accross is...how come yesterday six tests for ex. would finish in like 15 minutes when today, one test is taking more than 30 minutes...

i mean, i ran the same test with the same processor at the same OC (3.2 ~~> 3.6) and it completed fast and today, it's taking years...

the same for my friend, who is running one instance of Prime95, just on a less powerful CPU

[mgoode]

This is a memory bug in prime95. chose custom test then subtract 20 mb from the amount of memory tested. On my rig it defaults to 927mb and I change it to 907mb and then prime runs fine

[AlcatrazT2]

Does this bug happen when there are two instances of Prime95 running only or can it happen with one?
Thanks for your input everyone.

[deathman20]

1 reason why its taking a long time with the tests, is that its high PF usage. And well nearly 500 megs an prime instance, that chews quiet a bit of your ram down. If you don't have a gig, the main reason is that its swaping from the harddrive alot, which will slow the tests down considerably.

[mgoode]

Originally Posted by AlcatrazT2

Does this bug happen when there are two instances of Prime95 running only or can it happen with one?
Thanks for your input everyone.

On my rig it happens with both and I am running 1 gig(2x512mb) ocz pc4000 rev 2.

[AlcatrazT2]

Thanks a lot for all of your input guys.
Well, I ran the test once and it was working fine...CPU usage was at about 50% percent at average...nothing higher. I guess, if I ran another instance of Prime95...then it would go higher. However, once I did, CPU usage went down and I guess, it would take years for anything to complete like I mentioned...

So...should I just run Options, Turture Test, Custom...and in my case, if it's 415MB, subtract 20MB like mentioned above by mgoode and let it run at that? While the first test runs it on Blend?

Thanks again.

P.S. mgoode: When you said its happening on both, did you mean, just one test, and the same for both tests at once?
My friend was having the same issue running Prime95 (ONE instance) on his computer which has 1GB of RAM (PC2700 - NOT an OC'ed rig).

[mgoode]

My bad , I meant prime95 hangs hangs when running single or dual instances. Yes subtract 20mb from your default prime95 memory allocation.

[AlcatrazT2]

mgoode: Thanks for all of your input man. And I thought, I needed to reformat!

Well, for me it runs fine when I have a single instance running. However, it starts messing up when I have two instances running, so I guess, I'll make my default Prime95 run the Torture test which Blends everything, and the other one, will be the Custom with 20MB subtracted.

My friend, is having the same issue as you are. I guess, he'll need to subtract 20 like you did to get the thing working.

It's interesting how 1GB starts messing up when one instance is running...
And how 512MB doesn't mess up...However, when you do run 2 instances of Prime95 (I think of it as doubling to one GB) it starts messing up just then.
Makes sense why for 1GB it messes up with one instance, and how for 512MB, it will only do it when two instances run.

Do you have an idea why it's 20MB which should be extracted? Not more? Not less?

And just to double check...we're talking about the "Memory to use" feature in the Custom settings, correct?

[mgoode]

no problem, yes it the "memory to use" box

[AlcatrazT2]

Well...I ran one test on the "Blend mode" and the other on "Custom" with 20MB subtracted...and no luck. Things droped back to 4-9% range and didn't go any higher...at least one test goes up to 50%, with both, it's less!

despite this, i'm still wondering, why my friend was having an issue just running one instance and why i was having an issue and today, all is fine.

[SolidxSnake]

I wouldn't reccommend using "Blend" as a CPU tester... Use Large In-Place FFTs (the second option)... it tests the CPU a LOT

also, if you want to test ram, try Memtest86+