hardware, winxp or ocing...what is the problem

[Duvie]

Ok I am running at 2.5ghz and prime95 ran for 12 hours, temps 51-52c under load....

All of a sudden I change the priority of a program in winxp...it gives me the standard "this could cause instability"...I do it anyways and minutes later I get BSOD (first one seen)...By the way I was running GKnot dvd to divx encoding....

eror message:

***stop: 0x0000008e (0xc0000005,0xbf86ab10,0xf2b62c10,0x00000000)

*** win32k.sys - address BF86AB10 base at BF800000

if I don't go into safe mode or startup I will get same error but

*** ntfs.sys


I thought it was cpu but I can reproduce error with no ocing...or 1:1 or 4:3 ratio....

NOt heat as it can happen real early or later and I watched the temps and they were below 40c...


I try repair and recovery and it doesn't work...

I reformatted and reinstalled again NTFS ( all other drives are fat32...issue???)

Everything is fine run prime95 for 4 hours no problems...install msn and whamoo same error...on new install




sidenote: run fine in win98 with no issues in cadd and media player....



1) is it my ram crapping out...it is pc2700and I am only running it at 350ddr and I have never messed with cas timings...default 2.5t...

2) is it winxp?? maybe problems of working back and forth from a fat 32 and ntfs drive??

3) is it my oc'd processor??? I am running cadd and prime95 as we speak in prime with no issues...

4) Is it my harddrives???

 

[Duvie]

Does anybody have a memory tester program that runs in win98 environment??? I don't have a floppy hooked up...

I am testing now to see if it is the memory....


funny thing is pcmark2002 which often seems to stress system more them prime runs fine....hmmmm....

 

[WarCon]

On the new install did you switch the priority of the program again?

 

[Duvie]

No...I didn't even install that program this time...

I think I have ruled out heat as I have open case up and idle temps are 34.5c...load never gets much above 41c in programs...


I ruled out memory as I got same error no matter what ratio


I think it is the power supply as at same speed as it was last week the 12v rail is about .15v lower...plus if you hit sandra cpu repeatedly the alu score will fluctuate...heat doesn't climb so it is not throttling...if you wait about 10-15 seconds the scores stay close...I think the power supply is dropping and not recovering....

Edit: ***I noticed that it doesn't necessarily matter you can get a low score at any time...I noticed a 3.8% vcore swing in 1 second intervals during one of the low marks....


What do you think power issue maybe???


I don't want to think I toasted this thing...I have been wary of temps and felt temps at 50c load were reasonable...electromigration as I was tutored on has nothing to do with just voltage but heat breakdown...so I don't know...

I should have left it at 1.8@2.4ghz w/ 1.65v. temps were 36c/47c

 

[Duvie]

update: extensive testing....

NO matter voltage even at 1.8ghz (default) I get fluctuations in sandra cpu scores...slight voltage fluctuations...temps were 32.5c idle/ never climbed above 40c

At 2.4ghz at 1.6v,1.65v,1.7v,1.75v...I get same fluctuation in scores plus I notice 3.8-4 percent swings in vcore voltage....


Do you think this sounds like POwer supply or electromigration and I just f^d this chip up....


I can't think of a virus that could be doingthis unless it hides and sucks up cpu cycles to throttle chip down....

 

[WarCon]

I checked and I get the same fluctuations. I usually never run the same test twice in a row, figuring the part of windows (At least with Win98, might be the same in other OS's though) that has a real issue with releasing resources would cause my scores to decrease the next time I ran it. I never considered it might actually end up releasing them and have the scores go back up the 3rd time it is ran. I think your seeing a "sticky" kernel, not choosing to release the memory used by the last test or at least parts of it. Check your resources before you start the test and before you start the next test and keep checking until the score goes back up and see if it corresponds with resources being released.

This is just a thought, maybe I will try it too after I post this and see.

This is how it went for me during a brief test. I ran the CPU test and for the first two times it maintained around the same score, on the third time running it, resources were tapped for 3 percent more and the score went down 200 points. It stayed down and didn't flucuate for several more tests.

 

[Duvie]

BUt why just the alu...not the fpu or multimedia marks...

I have in the past I have ran the program repeatedly and the number didn't fluctuateway more then a few points either...


I have noticed in winxp longer waiting results in better recovery of score...also at 2.4ghz with 1.75v the score didn't fluctuate in multple tests....I am lost, can't really find much of a pattern to say one way or the other....

 

[Duvie]

close sandra down then reopen and I get no fluctuating at all...maybe it is something to do with windows....

4 times and score changed 10-12 pts....

Could I have corrupted both os's (dual boot) with bad memory???

 

[WarCon]

No I think your probably ok. I think its just a resource management issue, not a hardware problem.

 

[Duvie]

Warcon, thanks for the help....

I found that winxp is more of an issue then I thought....

reasons...

1) prime errors are from file I downloaded from xp...downloading from win98 and see if I get fresh copy will work...I also notice that every once in awhile winxp will force check of secondary drive and fix things...

2) winxp said pcmark2002 file was corrupt...however win98 could load it and use it...I ran it 6 times in a row and had i really assualt the vcore and 12v rail and yet my scores never really changed and system was stable...

3) Error that pops up frequently is win32k.sys and ntfs.sys...both I believe are only exclusive in winxp in my case...


I am still going to check memory out....

I may also reformat and install winxp as a fat32 with cpu at default and 1:1 ram timings, eventhough it is pc2700 ddr....

 

[THUGSROOK]

hehe, i remember with my old (?) p3 1g my scores used to UP if i repeated a sandra test :Q

get a new PSU and try again. your fluctuations and instability are outta control compared to my system (which is almost identicle to yours) so i doubt its hardware.

BTW: WinXP sux

 

[Duvie]

I am thinking about getting rid of this winxp and getting win2k pro...I wouldn't need to dual boot...

Hey I noticed when I removed sandra and installed sandra se no fluctuations in score regardless if I just sat there and hit it repeatedly....definitely software or corrupted registry to blame...


I think pcmark2002 really pushes the cpu as I watched the vcore fluctuate from 1.744 idle to 1.648 load...maybe one of my problems there...but it ran 6 times right after each other and score never fluctuated and temps reached 43c...


thanks for the comments guys....at least I am thinking it is unlikely the cpu being toast now...

 

[Pabster]

I'd run a few passed with MemTest86 and see if a stick of RAM is being flaky. I recently had a couple Crucial sticks sh!t out after being overvolted a little too much for too long

The fluctuating vCore can certainly be a problem. It might be rock solid with 1.75v, but not 1.65. I've noticed the voltage regulation isn't particularly great with any of the i845-D boards.