Question - HT chips ... 2 primes or not?

I have my 2.4C @ 3.432 ghz and instance 1 of prime95 has lasted as far as 27 hours before I stopped it but my 2nd instance always crashes somewhere around 8 hours.

I've read a few posts on the GIMPS forum with people saying that Prime95 v23.8 only needs 1 instance to test stability on HT chips... But everywhere else people still say you need 2 instances.

So which is it? Thanks.

edit: This is the thread I was talking about. http://www.mersenneforum.org/showthread.php?t=443&highlight=instances

My personal experince is that the "old" Prime, with just one torture test, puts more stress on the system.
Dual of course...

If one instance drops out, you on the edge... I would guess memory. I had that problem after 11 hours one time... dropped 1 FSB and totally stable...

I generally agree that two instances should be used, but I have the same problem as you do. One instance can run for days, but a second instance will only run on my machine for about 10 hours before dying. No amount of higher voltage will fix it; CPU, RAM, or AGP.

I figure that none of my current games are SMP-aware, and even if they were, I've probably got at least eight hours worth of play time before they GPF due to instability. Can I live with that?

Yeah

Anyone tried running 2 primes on stock speed & voltage? Now if one of them crashes..it's be interesting

IMHO - Two instances of Prime95 are needed using version 2.1x/2.2x
but 2 are not necessary with 2.3x due to the changes in the setup/options.

barton2500 said:

Anyone tried running 2 primes on stock speed & voltage? Now if one of them crashes..it's be interesting

Click to expand...

I have and mine kept going util I stopped it. In my sig is also 2 instance of prime95 stable on priority 10. if I go 1 fsb higher I'll get error in one of my prime95.

IMO I'd like to keep at least 2 instance of prime stable for at least 24hrs, if one fail then I just have to lowered my fsb until both run without error.

For stability I normally run at least 2 instance prime for at least 24hrs on priority 10, at least 24hrs on 3DMark and 24hrs on memtest86 for me to consider my system stable.

barton2500 said:

Anyone tried running 2 primes on stock speed & voltage? Now if one of them crashes..it's be interesting

Click to expand...

I will have to try that...

texasfit said:

IMHO - Two instances of Prime95 are needed using version 2.1x/2.2x but 2 are not necessary with 2.3x due to the changes in the setup/options.

Click to expand...

Yeah thats what the people at the prime95 forums say but it seems nobody agrees with them... What settings would you recommend to use?

Yes IMHO it's still necessary to use 2 instances of Prime95 to test your system fully with a HT P4. The reason being that one instance of Prime95 will not stress your system fully as there is still headroom available to do other tasks simultaneously. When running 2 instances, you will get higher thermal loads and power requirements which may be all that's needed to push your system over the edge. I routinely notice a 2C difference in temps between running 1 versus 2 instances of Prime95. Also, I highly suspect the people who say there's no difference between 1 and 2 instances are not correctly running 2 instances of Prime95 on their computer. When I run 1 instance of Prime, I can easily multitask at the same time. If I run 2 instances of Prime95, everythings draaaaaags down to a halt. There's no way you can do anything with the computer with 2 instances running correctly.

Anyhow, to the naysayers, try this. Open up 2 separate copies of Prime95. Unlock each copy by entering the password if you haven't already done so. Set the affinity of one copy to run on Processor 0 and the other copy to run on Processor 1. Set the Priority level on both to level 10. Start tortue test on both copies using the Large FFT option. Attempt to open Task Manager and observe your system resources pegged at 100%

Toaster Oven said:

Yes IMHO it's still necessary to use 2 instances of Prime95 to test your system fully with a HT P4. The reason being that one instance of Prime95 will not stress your system fully as there is still headroom available to do other tasks simultaneously. When running 2 instances, you will get higher thermal loads and power requirements which may be all that's needed to push your system over the edge. I routinely notice a 2C difference in temps between running 1 versus 2 instances of Prime95. Also, I highly suspect the people who say there's no difference between 1 and 2 instances are not correctly running 2 instances of Prime95 on their computer. When I run 1 instance of Prime, I can easily multitask at the same time. If I run 2 instances of Prime95, everythings draaaaaags down to a halt. There's no way you can do anything with the computer with 2 instances running correctly.

Anyhow, to the naysayers, try this. Open up 2 separate copies of Prime95. Unlock each copy by entering the password if you haven't already done so. Set the affinity of one copy to run on Processor 0 and the other copy to run on Processor 1. Set the Priority level on both to level 10. Start tortue test on both copies using the Large FFT option. Attempt to open Task Manager and observe your system resources pegged at 100%

Click to expand...

I totally agree with you on this one mann. This is how mine being run.

Well for me the only reason I would run 2 instances of Prime was to get the CPU Load to 100%. Running one instance would only bring the load to 50%. Sometimes i would run Sandra Burn in on low priority in the background and Prime95 and that would bring the load to %100.

I say if the new prime brings the CPU load to 100% then you only need one instance of prime.

why does everyone think that with one app running and in task man it showing 50% or 100% on one cpu that the cpu is only running at 50% load. Ht is tricking the OS into thinking there are two cpu's in the system. So a non smp/smt proggy will only load up on one virtual cpu and not the other. If your running prime you will maybe get another 5% load on the cpu since both primes will be demanding use of the same chunks of the chip which will cause the cpu to have to switch tasks constantly and thats where your extra heat and load come into play is the cpu has to work harder to keep switching between almost idenitical tasks.

LIke I said, it doesn't really bother me. If it makes my system a "lesser overclocker" because I can only keep two instances of prime running for 10 hours, then so be it. Or I could just turn off hyperthreading and let it run one instance for the next several weeks...

Until I find an app that fully takes advantage of my HT processor and FULLY loads it on both virtual processor tasks, AND be something that I'm willing to do for more than ten hours at a time, then I'll replan my overclocking strategy.

Actually, I take that back. At that time, I'll probably upgrade to an A64

My load temps do not change one bit if I run one or two instances of PRIME95. I am using v2.3xx

gamefoo21 said:

why does everyone think that with one app running and in task man it showing 50% or 100% on one cpu that the cpu is only running at 50% load. Ht is tricking the OS into thinking there are two cpu's in the system. So a non smp/smt proggy will only load up on one virtual cpu and not the other. If your running prime you will maybe get another 5% load on the cpu since both primes will be demanding use of the same chunks of the chip which will cause the cpu to have to switch tasks constantly and thats where your extra heat and load come into play is the cpu has to work harder to keep switching between almost idenitical tasks.

Click to expand...

Well how else are you suppose to measure the load? I used mbm5 to monitor load and i just assume if cpu1 was 100% and cpu 2 was 0% then only 50% load was used. Whether in actual fact more load was put on the cpu, i can't tell nor can i measure so i just say 50%.

Either way, finding the exact % of load wasn't the point, the point I made was by running another instance of prime or a prog such sandra burn- in in the background, i made sure the processor is stressed to its max to test for stability.

If it doesn't run two instances then it is not stable period!

If you like subtle errors occuring after a few days of being on then do it only 1 prime stable.

Nothing better than burning a cd or a DVD and get errors. Maybe you want to update your OS with a patch or etc... Maybe look for viruses. Those are all nice tasks to run and get errors I would say.

ol' man said:

If it doesn't run two instances then it is not stable period!

If you like subtle errors occuring after a few days of being on then do it only 1 prime stable.

Nothing better than burning a cd or a DVD and get errors. Maybe you want to update your OS with a patch or etc... Maybe look for viruses. Those are all nice tasks to run and get errors I would say.

Click to expand...

That, or we could (gasp!) maybe not leave our machine running for weeks on end at absolute full load? Damn! I'm a genius! There is a difference between my gaming rig and my webserver rig and my DMZ machines. The former one is used for gaming and is only powered on for 6-8 hours at a time, and overclocked to the hilt. The latter five are completely left at stock speed (or in two cases, are UNDERclocked) and have uptimes in the area of 100+ days. One machine in my DMZ has an uptime record of more than 400 days on Windows 2000...

None of my constant-on machines are running at more than probably 50% processor utilization, although one or two of them do run pretty constantly on the disk drives. The only rig that sees full load is the one that isn't left running all day.

Go figure

I just get a icky feeling using a computer I know is not 100% stable. I guess maybe if you are used to the old Kt133 chipset then you can accept hard locks and errors every few hours. Persoanally it is kind of annoying when my system locks up in a game at a LAN. Especially if it is a tournament.

Welp, since my gaming rig has yet to ever lock up over the approximately three months that I've owned it, I'm really not feeling very "icky". And yes, I've done some serious gaming marathon sessions on it: Halo single and multiplayer for approximately 18 hours, NFSUG for about 16 hours, UT2K4 with a buddy over Cable for about 12 hours...

If it can sustain a single session of Prime95 for days on end, then I'm pretty sure it can sustain a single session of ANY of the games I play for just as long.

Howabout you?

Albuquerque said:

Welp, since my gaming rig has yet to ever lock up over the approximately three months that I've owned it, I'm really not feeling very "icky". And yes, I've done some serious gaming marathon sessions on it: Halo single and multiplayer for approximately 18 hours, NFSUG for about 16 hours, UT2K4 with a buddy over Cable for about 12 hours...

If it can sustain a single session of Prime95 for days on end, then I'm pretty sure it can sustain a single session of ANY of the games I play for just as long.

Howabout you?

Click to expand...

Nope, ICK.

It is not telling the truth if two are not stable in HT mode anyway. WHat is the big deal knock it down 50MHz. What are you worried about losing 1 fps?

I always use 2 as well. People that think one is sufficient are generally in denial about their OCs.

Click to expand...

http://www.hardforum.com/showthread.php?threadid=734033

Yep I agree. You are in denial. It is not 100% stable.