View Full Version : Prime95 - valid test?
I'm running my XP2500+ at 200x11 (stock vcore). It sails through Hot CPU, Sandra's burn-in, Memtest, and PCMark2002. However, if I try to run Prime95, the machine freezes after about 5 minutes. I've encountered no problems with any of my apps in this config, and playing Unreal Tournament for several hours straight posed no difficulties either. Of course, it bugs me to no end that it can't get past Prime95, even though I'm not entirely sure how applicable it may be, given the results I'm seeing with other stability tests. Has anyone else come across the same problem with Prime95, and did you find that eventually the stability of your rig was compromised, even though other stability tests indicated otherwise?
GlitchOfDoom
08-06-03, 11:56 AM
Your computer freezed??? or did prime 95 just give you an error? If your computer freezed then I dont know, if you are on dialup and you used resume for some reason prime 95 could be corrupt. Or memory could be bad or set too high of timings.
Originally posted by GlitchOfDoom
Your computer freezed??? or did prime 95 just give you an error?
Thanks for the reply. No errors; the machine freezes after a few minutes in Prime at 200x11, but runs 170x12 and at 205x9.5 with absolutely no problems. As I mentioned, it runs 200x11 with Sandra, Hot CPU, etc, but for whatever reason doesn't want to work with Prime.
If your computer freezed then I dont know, if you are on dialup and you used resume for some reason prime 95 could be corrupt. Or memory could be bad or set too high of timings.
I'm using 2x256 Corsair XMS PC3200, and Memtest shows no errors after 8 hours. I've relaxed its timing, and that didn't make any difference with Prime. Perhaps there's something I didn't do correctly. When I installed Prime, I opted not to run the tests online. Could this be causing the problem I'm having? It doesn't really seem like it would, but stranger things have happened.
NookieN
08-06-03, 12:51 PM
Yes if you're just using Prime95 for stress testing you don't want to join GIMPS.
What version of Prime are you running? v23.6 is the latest and offers some new stability test options. Maybe you'll see different results from each of Prime's different testing modes, which could help you pinpoint the problem.
Originally posted by NookieN
What version of Prime are you running? v23.6 is the latest and offers some new stability test options.
Indeed, I was using an older version. I downloaded and installed v23.6 and things are going much better; thanks for the tip! It's been running at 190x11 for a few hours now with no errors. I tried it at 200x11 and 195x11 and the thing froze every time. Raising vcore only made matters worse, making the machine freeze even sooner. Perhaps my logic is a bit off, but this makes me think the CPU is too hot. Motherboard Monitor shows the temp as 41 under full load, as does Sandra and bios (I'm using a MSI K7N2 Delta-L mobo and a Thermalright SK-7). It is said that MSI's bios intentionally show lower temps, but the HS feels relatively cool to the touch under full load, so I dunno...
ll see different results from each of Prime's different testing modes, which could help you pinpoint the problem.
Great idea! I'm on it :-)
What do you think about my temps? I mean, is it possible that the CPU is overheating, even though I'm getting reasonable-looking readings and the HS doesn't feel particularly hot? It's not cold by any stretch of the imagination, but it's hardly glowing either. I made *very* sure that it's installed correctly and applied AS3 between it and the processor. My case is very well ventilated, but to rule-out that variable, I've been using it without its side panels attached. My room is at 25-30 degrees max.
I guess I'd feel better about this if Prime was showing errors, but the freezing has me scratching my head, especially since the machine can run other tests with no problems.
Look into your Powersupply. when you raise the CPU speed it draws more power and if the PSU can't keep up the Vcore starts bounceing around instead of hanging out in a fairly narrow band. CPUs rather dislike such voltage fluctuations.
Use MBM5 to track your vcore while under load and see how much it varies. Granted it isn't spot on with it's measurements but it will expose the trend never the less.
i can't run prime either. THe latest version of prime is no dice as well. At stock speeds it'll even give me errors. I don't know what's the deal. I can run Sandra burn in, f@h, and s@h for hours on end. Not to mention bench in 3dm03 and 01. I don't understand what's the issue. My proc is @ 100% load right now and mbm shows my vcor @ 1.76v. :confused:
Originally posted by AKULA
Look into your Powersupply. when you raise the CPU speed it draws more power and if the PSU can't keep up the Vcore starts bounceing around instead of hanging out in a fairly narrow band. CPUs rather dislike such voltage fluctuations.
Use MBM5 to track your vcore while under load and see how much it varies. Granted it isn't spot on with it's measurements but it will expose the trend never the less.
It's interesting you say that, as I've been looking at the power rather closely. I've noticed no variance under load. I'm using a Chieftec 420 watt, which is a rebadged Enlight EN-8420934 (http://www.newegg.com/app/Showimage.asp?image=17-103-604-05.jpg/17-103-604-06.jpg/17-103-604-07.jpg/17-103-604-04.JPG). To tell ya the truth, I almost wish the PSU performed worse than it does, because that would explain the problem I'm having and provide an easy fix.
LtBlue14
08-07-03, 09:19 AM
chowdy - if it's givin you errors even at stock speeds, then something's definitely wrong, but i don't know what
RoadWarrior
08-07-03, 10:27 AM
Try your RAM at CAS 2, this sounds stupid, but I've been hearing of a few tbreds and bartons that have wierd errors at slower settings.
01_VTEC SE
08-07-03, 12:56 PM
what mobo are you using?
try 12.5x 176 FSB = 2200. Maybe it's your FSB that's causing problems??
Originally posted by RoadWarrior
Try your RAM at CAS 2, this sounds stupid, but I've been hearing of a few tbreds and bartons that have wierd errors at slower settings.
I tried it, thanks. Setting the timing at 2-2-2-11 yelled the best results yet at 200x11. Prime locked-up after about 1/2 hour instead of the usual 2-5 minutes. I still can't figure out why HotCPU, Memtest and Sandra's burn-in can run seemingly forever, but Prime locks-up the machine. Similarly, all apps run perfectly, too.
Originally posted by 01_VTEC SE
what mobo are you using?
It's a MSI K7N2 Delta-L. I've noticed that other folks with the same mobo are also having problems with Prime, so perhaps it's some weird interaction with the mobo that's responsible. I'd really like to get to the bottom of this.
PoX Freak
08-08-03, 09:00 PM
Still looks to be a volt problem though, even with the cpu at stock.
I'd opt out the PSU and grab an antec 400 watter or something of the sort. I had the same problem with my T-Bred. Everything els ran perfectly, just Prime95 would hang after 5-10 minutes. Then i swapped the PSU and POW.......no more problems.
simulator
08-08-03, 09:42 PM
Originally posted by blkgti
I tried it, thanks. Setting the timing at 2-2-2-11 yelled the best results yet at 200x11. Prime locked-up after about 1/2 hour instead of the usual 2-5 minutes. I still can't figure out why HotCPU, Memtest and Sandra's burn-in can run seemingly forever, but Prime locks-up the machine. Similarly, all apps run perfectly, too.
Prime95 puts a loooot of stress on your CPU and memory, I think
more than what any other application possibly does
For sure it's way more stressful than Sandra, 3dmark, etc...
they don't even come close to this thing...
for me 1 hour Prime w/ no error means a highly stable machine
if I can repeat the experiment a few times of course.
more than 1 hour is ice on the cake...some people are not
satisfied until they run P95 for a couple of days!
I agree with the power problem people. I have had similar problems with my dual athlon board. I think that the voltage regulators on my board weren't too good....Anyway, it would run at reduced speed and lowered voltage, but not at stock.
The reson people like prime 95 as a hardware test is that it is pretty much OS independent. The code for it is almost all ASM, and so it works on the hardware level stressing the cpu to the max....Anyway, I'd look into your PSU first and then the mainboard....
PS. I often couldn't see the voltage dop that would make things freeze.
simulator
08-08-03, 09:54 PM
Originally posted by TLhead
I agree with the power problem people. I have had similar problems with my dual athlon board. I think that the voltage regulators on my board weren't too good....Anyway, it would run at reduced speed and lowered voltage, but not at stock.
The reson people like prime 95 as a hardware test is that it is pretty much OS independent. The code for it is almost all ASM, and so it works on the hardware level stressing the cpu to the max....Anyway, I'd look into your PSU first and then the mainboard....
PS. I often couldn't see the voltage dop that would make things freeze.
if the onboard regulators are not too good, do u think replacing the PSU would help??
I am having pretty much the same problem with Prime95
and I do suspect a power problem somewhere but my voltage
readings are quite stable...
You can tell if it is the PSU by monitoring the voltages in MBM or even better with a voltmeter. If all your voltages are stable except the vcore and or vdimm then I would say it is likely a board problem, but if you see fluctuations in the 5, 12, or 3.3V lines the PSU would be the culprit.
To answer your question more directly, if the MB regulators are the problem, it's time for an RMA. MSI was pretty easy to deal with when I got mine. I hope this info helped and you can get it working! It's soooo frustrating to get a nice new part and it doesn't work as advertised.
TLH
Originally posted by TLhead
You can tell if it is the PSU by monitoring the voltages in MBM or even better with a voltmeter. If all your voltages are stable except the vcore and or vdimm then I would say it is likely a board problem, but if you see fluctuations in the 5, 12, or 3.3V lines the PSU would be the culprit.
I ran HotCPU and kept a very close eye on the voltages at full load with MIM5. There was a slight fluctuation in vcore - 1.63-1.66 - but all other voltages were totally stable. The variance in vcore amounts to ~2%. That seems like it's within spec, or is this something I should be concerned about?
Blueacid
08-09-03, 08:09 AM
First up, Welcome to the forums, blkgti!
Second, if you are not bothered about one stability test, do not worry - My machine will only tolerate about 30 minutes of Prime95 before giving up, but for what I want (Gaming, Internet, Multimedia) the system is rock stable, even when running it at 100% load with Folding@Home (If you don't know what this is, look for the Folding Team forum, located on these message boards).
Admittedly some people will never sleep until their computer will run Prime95 flawlessly for days on end, but I'm not one of those people :p
SOunds like either a heat problem or a PSU problem, judging from stability going down when voltage goes up. Measure the voltages coming off your PSU with no load and running prime95 or cpuburn (even more stressful). Where do they start? How much do they drop? What are your on-die idle and load temps?
Originally posted by Gnufsh
SOunds like either a heat problem or a PSU problem, judging from stability going down when voltage goes up. Measure the voltages coming off your PSU with no load and running prime95 or cpuburn (even more stressful). Where do they start? How much do they drop? What are your on-die idle and load temps?
Thank you for your reply. I've already addressed the questions above in other posts to this thread. If you have a chance, look them over. I'd appreciate hearing your thoughts once you see the info.
I did read the thread, I didn't see multimeter readings off the PSU lines, nor cpu on-die temps. I did see your vcore readings, though I suggest you try to get those with a multimeter too, as motherboard voltage sensors have a way of being very inaccurate. The 2% drop may be normal, from the swithing on and off of the mosfets (two or thee phase voltage regulation?).
edit: on my main rig, which I don't have with me now, my 5v line drops .01 volts under load according to my voltmeter.
Originally posted by Gnufsh
I did read the thread, I didn't see multimeter readings off the PSU lines, nor cpu on-die temps.
I'll try to measure the lines with a miltimeter this afternoon. I assume you're talking about the 5v lines, correct? Though, how would I get the on-die temp? Too, how would I measure the 12v line with the machine under load, as the 12v line is used to power the CPU?
I did see your vcore readings, though I suggest you try to get those with a multimeter too, as motherboard voltage sensors have a way of being very inaccurate.
How would I do that? I mean, measure the vcore with a multimeter.
I measured the 5v and 12v lines under a full load with a multimeter:
5v ---> 5.12-5.13v
12v ---> 11.85-11.88v
That looks like it's within spec.
Those sound alright. I think most newer boards measure on-die by default, some people use in-socket anyway for some reason. I highly doubt that a 2500+ would do 2200 MHz at default voltage. If bbumping the voltage (a very small amount at a time!) causes more instability, there is either a heat or power problem. I still don't trust your PSU, An antec or enermax would br preferable. You could try adding heatsinks to your mosfets. You may also want better cooling on your CPU. It seems like you need to spend more time on your OC also. WOrk up slowly, a few MHz at a time, when you start to get a little unstable, bump the vcore a little and see what happens.
Originally posted by Gnufsh
I highly doubt that a 2500+ would do 2200 MHz at default voltage.
According to the CPU database (http://www.cpudatabase.com/CPUdb/Showamd.cfm) and quite a few user experiences, it seems like a reasonable goal. Right now I'm running 185x12 at stock Vcore - ie. 2221MHz - (Prime ran for 10 hours with no errors) and it's very stable, but I'd like to increase the FSB a bit more. Are you using a 2500+?
If bbumping the voltage (a very small amount at a time!) causes more instability, there is either a heat or power problem.
That seems logical. I posed a similar theory earlier in the thread, but without a way to test the variables it doesn't give me much to go on.
I still don't trust your PSU. An antec or enermax would br preferable.
Why so? The voltages seem right and it's identical to the Enlight which from what I've gathered is well-regarded. Some folks have problems with Enermax and Antec putting-out stable voltages, which is something I've yet to encounter with mine.
You may also want better cooling on your CPU.
What's wrong with the SK-7? According to the tests I've read, a bigger Thermalright only yields a couple of degrees difference. My fan puts-out over 40 cfm, but I guess I could try a higher output model...
It seems like you need to spend more time on your OC also.
Well, I've spent countess hours on the project already, testing at *every* step, but I'm sure more time couldn't hurt.
WOrk up slowly, a few MHz at a time, when you start to get a little unstable, bump the vcore a little and see what happens.
Been there, done that over and over and over. Thanks just the same. I posted my question only after I exhausted every permutation at my disposal, because the box has problems running Prime and no other stability test. BTW, I'm still curious how one would measure Vcore with a multimeter as you suggested, and I'd appreciate some pointers.
vBulletin® v3.8.4, Copyright ©2000-2009, Jelsoft Enterprises Ltd.