prime95, making steam come out of my ears!!

[sendatooli]

i am so mad! prime95 always gets an error on my machine. sometimes after 2 hours, sometimes after 1 minuet!

my machine is not overclocked!
it is a top of the line dual mp workstation. everything is the very best. i built this machine speciffically to be rock stable.

now what gets me so mad, is that according to prime95 docs. if it causes an error, your machine has a problem. it is stated that there are no bugs in prime95!
intrestingly, the old prime95 runs ok for much longer than the new one with the installer.

right now the machine's cpu's max out at 40-41c. full load.
the case gets up over 50c. thats unavoidable with all these drives and burners. the case has very good airflow and plenty of room.

so what can this be? a bug in prime95? bad memory stick? hardware conflict?
is prime95 the last word in stability testing like it says it is? or is the auther just touting his horn?

what else could i use to be absolutley shure my machine is stable and not getting errors? and maybe even something that if indeed there are errors will let me know what part of the machine is causing the errors?

i figure if two stability apps. create errors then the machine is in trouble. for now i am doubting prime95. anyone else with me on this?

i am not concerned with benchmarking the machine. it's plenty fast for what i use it for. it suits me well. i just CANNOT afford to have it lock up for no good reason.

on a side note it runs my very processor,memory,sse intensive pro audio applications flawlessly.


i just want to be shure its 100% stable, using a app speciffically designed to test stability. whats the most trusted stability test around here? i do not want to use 3dmark. it has too many bugs that are known.



thank you,
sendatooli.

[sendatooli]

i am now trying passmark,toast and k7burn. do you guys trust the results of these?



sendatooli.

[Tiger]

Originally posted by sendatooli

right now the machine's cpu's max out at 40-41c. full load.
the case gets up over 50c. thats unavoidable with all these drives and burners. the case has very good airflow and plenty of room.

This temperature diff is not possible unless the HSfs are drawing ambient air for cooling.(i.e. the CPU temp is 10C lower than case temp) So if you are using the air within the case to cool the CPU then this is impossible. Your CPUs must be running in the order of 60C which would make them borderline in terms of any stress placed on them.

[sendatooli]

ive got a ram air on the cpu fans. so that is why they are lower.

anyways, i figured out my problem!

memtest reporting many memory errors!!!!!!

now what? memory is not overclocked and running at stock settings. these jerks replaced the memory once and the motherboard once. i will never deal with them again. i doubt they will replace it now. anything i can do?

well, i guess prime95 works after all. it would be nice if instead of saying "hardware failuer" it reported what exactley faild!

if there is any way to fix my memory short of replacing it please let me know!





sendatooli.

[muddocktor]

I would print up some results of your mem test failure if possible, and bring it to the folks that sold you the computer and make them replace the memory with some quality stuff, no some no-name cheap cr*p.

[thefly]

Swap ram into other machine. run ram bench marks and tests to see if it's the ram.

Maybe adjust your ram timing, maybe that would help.

[mEKbOY]

Ram is the Number 1 problem on the Tyan Tiger and Tyan Thunder boards. Make sure you got some good quality stuff that is aporved by Tyan otherwise it can cause you headaches up the arse.

[sendatooli]

its micron ddr. it has been replaced once already. apparently this board is real picky on memory. even 2 sticks from the same brand one might work and the other not.

so, dont ask me how i did this. but, i managed to get new memory on christmas eve!
guess what! memtest86 ran 13 hours on "all tests" with 0(zero) errors!. what i dont understand is i am still getting memory errors in windows. according to passmark. and prime95 still quits? if this is more bad memory wouldnt memtest86 catch it? or is it another problem now? hardware conflict? windows corrupted? doesnt like my mobo or 2 cpu's? any ideas what i should try and troubleshoot here? i wish there was a program that said "this is the problem, and this is how to fix it!" i wish. this stinks for a machine that isnt even o/c'd. the guy that sold it to me says "this what customer gets when he tries to put together system by himself". i should tell him to stick it! i have put together many machines. the only ones that didnt work were because of bad hardware, not because of me.


anyways, mekboy if you are reading. should i flash my bios? i got the original one and it doesnt recognize my processors as mp's. i think it should? the guy where i bought it told me to flash bios "at my own risk" so this guy is a jerk! if i flash this board am i likely to get a dead board? did you flash yours? i got thunder.



happy holidays,
sendatooli.

[mEKbOY]

Well...i know you have a thunder as well. My board came stock with bios version 2.07. I later flashed it to 2.09. not 2.09a which is the beta one at tyan's website. it has been working fine befor and after i flashed it. The newer bios gives you some more chipset settings. I would say read tyan's faq and go for it. If you get a weird error while trying to flash it let me know. I can fix that for ya to. regarding the memory. I dont care if you have tried 20 sticks of it. If it aint ddr registered thats been tyan aproved thats listed on there site then it will give you hell. I just finaly experienced problems with mine today. I bought some micron no name stuff that has been working fine or at least i thought. my system has been totaly stable except for a radom freeze every now and then. Well today it happened and caused me not to be able to load wondows any more. Sometimes the machine wouldnt even start. I took out 2 of the ram chips and wallah... it worked. I am going to be buying some corsair ddr registered thats on tyan's aproved list. Lesson learned the hard way. These high class sever boards are picky and if you dont give them what they want then they will screw you. BTW..i have mp's and befor i had the 2.09 bios they were always reconized as normal athlons. Any other questions just ask.

[sendatooli]

thank you mekboy. i will try to get better memory.





sendatooli.

[Crash893]

well i can tell you i am running a prime95 team and on alot of computers and i can say that with about 100% certianty that prime has never crashed any of the systems directly ( some times they wig out becuase of the extra heat but on a truly stable machine prime gives me no crap)

if you would like email me your prime.log and ill take a look at it.(crash893@hotmail.com (use a noticable subject line so i dont trash it))

also i told some other guy prime logs what time the erros take place.

so i would get mbm and have it set up a log and then look at the corosponding time in that log and see if the temps were up or if something else was going wigy.

anyway good luck.

[sendatooli]

thanks crash. you are right. i had settings pushed too far. as for my own system i had a service running that caused a conflict. or had a bug in it. hmonitor. temp sensing app.


sendatooli.

[Alex99]

Hey crash,
I let Prime 95 ran overnight. In the morning I see a Windows98 dialogue box saying "This program has performed an illegal function......" When I click OK on the box, it dispears and along with Prime95 (Prime exited). Usually when I get errors, I get the error message within Prime95 and Prime95 does not quit it still keeps going. I looked at Prime95 results.txt file, I see a bunch of Error logs stating possible hardware errors. BUT, nothing from the last night crash. So, I'm lead to believe that it was a Windows98 problem. Or was it? What do you think?

My system spec is as such:

AthlonXP 1700+ @ 1650 Mhz (11x150)
Shuttle AK31 v3.1
Crucial PC2100 @ CAS2, 4-way, 2T, 5T, 2T, 4 Level, 1T

My temps are very good. I just bought a modified case and CPU temps from MBM are 37 idle 43 Load. I know I'm at the limit of my CPU because at 11x152 I get Windows Protection error while Windows is loading. But, at 11x150, I ran Quake3 torture in loops for 24 hours straight w/o any problems. I don't have any other problems except for that one weird Prime95 crash above.

[Crash893]

so did all or 98 crash or just prime95

im just going to assume that just prime crashed
and yes it was probably 98 that just couldnt handel it anymore.

if you wanted to test it out i would do a fresh reboot of 98 then start up prime95 and see how long it can go before it crashes.


are you doing a workunit or just the stress test?

[Alex99]

Originally posted by crash893
so did all or 98 crash or just prime95

im just going to assume that just prime crashed
and yes it was probably 98 that just couldnt handel it anymore.

if you wanted to test it out i would do a fresh reboot of 98 then start up prime95 and see how long it can go before it crashes.


are you doing a workunit or just the stress test?

No, Windows didn't crash, just Prime95. And after I rebooted and restarted Prime95, I get the same error (This program has performed an illegal operation...) at almost the exact time. I was doing a work unit. At first I thought, that I was having Windows issues. But while I was working on StudioDV (converting .avi captures from my DV camcorder to Mpeg2) I got the same error for StudioDV. At this point I knew that I'm having hardware issues. So, I bumped my system clock down from 11x150 to 11x148. Sure enough StudioDV completed the job. Then I run Prime95, and it ran straight overnight w/o a hickup. It is still running when I left home this morning. This is great!! I must admit, I never had much faith in Prime95. I really thought the program is buggy and errors itself out at random for no good reason. Hopefully, when I get around to volt modding my AK31 v3.1 I'll be able to run my system higher then 150 FSB. But who knows, it could my RAM that's hitting its limits. Anyways, thanks.

[Crash893]

have you looked at what temps your computer is at when prime crashes.

try the stress test option

[Alex99]

It crashed after about an hour into the program. Usually my load temp was at 45-46 C after while running Prime95.

I have an update. Yesterday, my Millennium Glaciator 1 HSF that I ordered came. After installing it using the supplied Arctiv Silver, my load temp is at 39-40 C. Very nice heatsink, quiet, but very heavy though. I now can run at 150 FSB stable. I cranked up Prime95 after setting the FSB to 150. It ran straight over night. It was still running when I left home for work this morning, YES!!! I think 1650 Mhz is the limit of my CPU. I tried 151 FSB it boots up and I ran Prime for 15 minutes without any errors. Then I tried 152, but it wouldn't load Windows. Right before Windows come up, I keep getting Exception Errors. I've learned that eventhough my CPU is pretty stable at 1650 Mhz, that it's very sensitive to heat at that speed. In the summer time, I probably won't be able to run at that speed. But a few notches lower, at 11x148, it's very stable and pretty tolerable to heat.

[sendatooli]

i had prime95 running for four days straight. it locked well into the 4th day. does this mean my machine is still not stable?

i tried this twice, it seems to lock at about 4 days. but, not at the same place. so i know its not a bug in the program. it seems odd that it runs that long if it isnt stable. could it just be windows xp giving up?

should i worry that my machine isnt stable now?

i know my crummy memory plays a big factor. because i had to put the memory on the lowest settings, to get prime95 to run for more than an hour.

it should work on the lowest settings. either that or it doesnt work at all. but, memtest doesnt give any errors. so it must be ok?



thank you,
sendatooli.

[Crash893]

i think my world record for prime95 up time is like 2 years on a computer at work

if your talking about the memory you assign to prime at the begining that doesnt matter all that much (you just use it for p-1 factoring and even at that it just speeds things up)

[Crash893]

also i thought you should know there is a new version of pirme95 out i think 21.4 that backs of some of the p4 error correcting to avoid false alarms due to hardware settings

and it will also tell yo uhow long you were in test mode