Prime95, run for how long?

The Asgard said:
You can use prime 95 on dual cores by having 2 copies of the program running ( make 2 dirs of prime95) and setting the instance to each core
Click to expand...
That's not the way you do it, just install Prime95 once and then make two shortcuts.

dualprimeqq0.jpg
 
Big.Wayne said:
That's not the way you do it, just install Prime95 once and then make two shortcuts.
Click to expand...


How dismissive ;) His way works as well as it loads one core per instance anyway without setting affinity. An even easier way i found was just to keep it in the zip file and open it twice, it makes 2 different temporary folders and loads each core up. So while you way works, theres more than one way to skin a cat.
 
When do you guys run your tests after overclocking? When your in bed? Just don't fancy leaving mine under stress and not being able to be there incase something happens!
 
justinwilkin said:
Basically I've been dabbling in some amateur overclocking of my p4 620 (3.2ghz running at 3.5ghz).

It seems ok but I want to stress it and find out if it's stable.

I've just downloaded prime95, and I was wandering how long I should run it to see if my overclock is trully stable, and what will happen if it isn't?


I'm slightly worried as before this overclock, I tryed to get it to 3.6 but it screwed up my windows installation hence why I'm not going that far.

Any help would be appreciated :)
Click to expand...


Stability testing? thats a big chapter buddy...

I have had prime crash even at 24 hours testing with phase change cooling etc.

You need 2 instances of prime...not on blend tests...

try small and large FFTs...

then blend too
but if you want to test ram memtest is the best running loops of test 5 for example for quite a few hours...

Most of the guys think that running prime on blend even with one instance will determine whether your rig is stable...

dunno about now....but back in my days stability had a big meaning determining how high our rig would run 24/7 as thitsd was always my case even with phase change.

so check your cpu stability first...then..bootable disk...memtest..and loops of certain tests...max 3 errors is fine and with some burning in your ram may go higher...
 
Brwmogazos said:
You need 2 instances of prime...not on blend tests...

try small and large FFTs...

then blend too
Click to expand...

I can't run two instances :confused: It's not a dual core CPU, unless you mean it can run two because of hyper threading?

I ran it with large FFT's for 8 hours over night and it was fine. No errors at all, so I'll give it a go with small FFT's later :)
 
There are some really odd things being said in this thread;

You can do other things while Orthos is running - Priority 1 means that Orthos takes least priority on your PC, Priority 10 means Orthos take top priority over all other tasks, this can even make it hard to shut down Orthos in my experience.

You should stress test your PC under normal conditions. There is no point running a stress test overnight or through the day with the heating off if you are going to be using the machine in the evening with the heating on full blast. This is the biggest source of long-term failures - it'll run just cool enough overnight then in the morning or evening when the heating comes on, the CPU overheats and the stress test fails. That's why a 24-hour stress is the only true stability test. 8 hours is fine, but it should encompass all the temperature conditions the PC is likely to encounter.

Big.Wayne got a monster overclock out of an E4300 (if I recall correctly) by running the machine on a cold night with the bay windows open. Kudos for 'can-do' attitude and personal fortitude, but no-way was he claiming it was stable under normal conditions.



Orthos has multiple stress tests. It's worth running them all if you are having failures as more often than not in my experience it's the RAM that's failing, not the CPU overclock.
 
justinwilkin said:
I can't run two instances :confused: It's not a dual core CPU, unless you mean it can run two because of hyper threading?

I ran it with large FFT's for 8 hours over night and it was fine. No errors at all, so I'll give it a go with small FFT's later :)
Click to expand...


yes buddy...its due to hyper threading...this is the only way to stress 100% (load) in your cpu...

The only thing is that you need to be patient.

For example...back in the old days :P i used to start stability testing my ring before going to bed....leave it for 24+ hours(i know quite a long waiting time for a student with just one pc) and then see if the test was running fine...

small ffts and large ffts.


then if that was fine...i run blend...


to test mobo cpub ram etc...

that was the only way to find out if my cpu was running rock solid(still 24 hours doesnt mean 100% stable) while running 5/4 the mem divider...



then run 1:1 and try memtest loops..not the usual test which completes all the tests but loops of certain tests...thats the only way to find out if your ram is stable...

then you can know your rig is 24/7 stable...+ finally run your rig at the optimal clock and start stability testing again.

At least thats what i did cos i was always running 24.7 my one and only phase change cooled rig...

also had a few mobos die on me due to condensation even though there was no way of air slipping into my socket with the methods i used back then...heater, neoprene, dielectric grease, loads of seal string everywhere, and the air trapped inside was enought to due the job...

Anyways...i am mentioning that because you will definitely hear many opinions about everything...its up to you how important you take stability testing etc...theres definitely someone else out there who stresses his rig for stability a lot more than i do etc...but anything less than 24 hours for me is not stability testing...
 
I always have and always will follow my 24hr rule, no matter that the 'experts' may say. I've seen enough crashes well past 8 hrs, going in to the 19 20th hour etc... for me not to trust 8 hours. I've done this with a multitude of setups, peltier, air, water and phase. I generally do a 24hr memtest as well though i'm not tooo strict on that, have cut it off after 12hrs or so. Funny how my rigs never crash apart from dodgy software once in a blue moon (and not a windows crash mind you) or real hardware failure due to something just deciding to pop :D Wish i was mad into it these days but i just overclock whatever the cpu can do on air and stock volts, most i do is a 3rd party coolers these days and thats for quiet more than overclocking.
 
For a real solid stress test i normally run it for 8-12 hours overnight, I start it about an hour before bed so i know its ok before i leave it. On the advice of Big Wayne i think its also good to let it finish a complete round as this will have stressed the pc using as broad a spectrum as orthos can. I also use priority 9 instead of 10 incase i need to do something or stop orthos easily.
 
You must log in or register to reply here.
Back
Top Bottom