Random BSODs (XP 32bit)

[Yossarian22]

'Ello, General [H]ardware. As you might remember, I joined this forum a bit over 50 some odd days ago, got sick and tired of biased and low-quality Tom's Hardware and now I'm here to stay, I read the forums daily and I've learned a lot about the current hardware "landscape" so to speak, now I visit religiously... but I digress.

I upgraded parts to what is now in my signature. Went on a three week vacation and came back. Everything ran fine. I started getting an ATI BSOD when attempting to update to Catalyst 9.5, but I finally got that fixed... or did I?

Ever since my upgrade to 9.5 (and return from my vacation) I've been getting BSODs all this week. Not all of them are related to ati_whatever.sys. I was messing around in Ableton Live (audio production\editing program) and I got a BSOD based on my Realtek driver. Before that, I was playing the Sims 3 and it gave me a BSOD on the ATI driver again.

I've done light runs of prime95 (12 hours) and Furmark (~3 hours) and everything seems to be stable. The thing that scares me is that Realtek BSOD. IRQ conflict maybe? That doesn't make sense, though.

And these BSODs are just random. I'll be doing something semi system intensive (okay, Ableton Live on my rig is not intensive at all, less than 10% CPU usage the entire time) and it occurs.

Any ideas? I'm going to do a memtest86 run. I've got the E8400 @ 4001 Mhz (445x9, 1.2v) and the HD4770 running 850\1000, but coretemp says the CPU is below 50c and GPU-Z says the 4770 runs mid-50s under load.

Any ideas?

 

[sirmonkey1985]

if its the IRQ not equal or less error.. id run memtest.. i think cat 9.6 is suppose to be released soon.. if the memory checks out.. you might want to try that.. see if it fixes the problems you had with 9.5..

 

[engiNURD]

Roll your ATI drivers back and see if that helps. Not sure on the realtek driver, though. Whenever you get a BSOD, be sure to take down the hex error code (stop error code).

 

[Yossarian22]

Probably going to roll back to 9.4 see if I can isolate the problem (until 9.6 is released soon).

I did a single run of memtest86+ last night and it was in the green.
I'm at my grandmother's now and memtest86+ is running as we speak and I won't be home for a few more hours, so I'll have more feedback in the memory department.

Couldn't be because CPU isn't getting enough voltage, could it? Otherwise it would fail to POST or just give me a non-BSOD crash under heavy load, wouldn't it?

Edit: Will do about the hex-codes and the errors

 

[Yossarian22]

if its the IRQ not equal or less error.. id run memtest.. i think cat 9.6 is suppose to be released soon.. if the memory checks out.. you might want to try that.. see if it fixes the problems you had with 9.5..

I think that was one of the BSODs. Here is the original screwed up Catalyst 9.5 install BSOD. I don't get this exact error anymore since 9.5 is installed "correctly" if you can call instability and random BSODs correct.

Well, I ran memtest86+ (v2.11) and it checks out. Nothing wrong with the RAM except the heatspreaders get exceptionally hot.

Also, I had forgotten I'd done a few short prime95 runs:

[Sat Jun 13 09:44:36 2009]
FATAL ERROR: Final result was F50B8362, expected: 1A922371.
Hardware failure detected, consult stress.txt file.
[Mon Jun 15 07:53:17 2009]
FATAL ERROR: Final result was 45D21593, expected: D8245D49.
Hardware failure detected, consult stress.txt file.
Self-test 320K passed!

For the June 13th date, I also was running and logging coretemp. Let's take a peek at the exact time right before prime95 failed, the direct dump, emphasis mine.

Time,Core 0 Temp.,Core 1 Temp.,,Core 0,Low temp.,High temp.,Core load,Core speed,Core 1,Low temp.,High temp.,Core load,Core speed
09:44:35, 56°, 56°,,,31°,56°,100%,4005.13 MHz,,31°,56°,100%,4005.13 MHz,

The CPU isn't overheating and I doubt that it's the cause of the BSODs, but it could be related. On a short June 15th run, I ran prime95 stress test for 2 hours, 2 minutes before an error.

I haven't reverted back to Catalyst 9.4 yet because I want to reproduce the error a few more times. Also, I have to install my motherboard's monitoring software to get NB temp. My NB heatsink is hot to the touch and the RAM is very warm, but the CPU is cool to the touch.

 

[Nenu]

Your CPU failed Prime 95 but you still think the CPU is stable?
It seems strange you would run the test, ignore the result and then ask here what the problem is.
This assumes you were testing for CPU stability not ram stability, if it was set for ram testing, then the same applies.

 

[sirmonkey1985]

Your CPU failed Prime 95 but you still think the CPU is stable?
It seems strange you would run the test, ignore the result and then ask here what the problem is.
This assumes you were testing for CPU stability not ram stability, if it was set for ram testing, then the same applies.

yeah i agree..


but what are the northbridge temps.. and also with p95 what tests did you run? small FFT's, large FFT's or blend test?

and now we know the rams fine since memtest passed.. so id say run some small FFT's..

if the small FFT's fail then your overclock is completely unstable..

also if you are getting a BSOD that says "machine exception error" that means your processor is starving for power(voltage)..

so try this and see what happens.. also cat 9.6 was officially released today..

 

[Yossarian22]

Your CPU failed Prime 95 but you still think the CPU is stable?
It seems strange you would run the test, ignore the result and then ask here what the problem is.
This assumes you were testing for CPU stability not ram stability, if it was set for ram testing, then the same applies.

The CPU is no where near the level of stress or anywhere near it's maximum TDP production (56c when tmax is like 95c?) when the BSODs occur. Stress test (large FFTs) under prime95 as far as I understand always fails (p95 documentation mentions it), it's just a matter of days, weeks, etc... Sims 3 is not CPU intensive, nor is Ableton Live with one VST loaded... However, Prime 95 tests bring up good points in overall stability. It's also my first time really "stress testing" a machine using these tools.

but what are the northbridge temps.. and also with p95 what tests did you run? small FFT's, large FFT's or blend test?

and now we know the rams fine since memtest passed.. so id say run some small FFT's..

if the small FFT's fail then your overclock is completely unstable...

I'm not really experienced with the software, but I believe I ran large FFTs.
The testing isn't all that scientific or conclusive because I haven't got all the monitoring software running while doing specific tests at intervals and collecting a set of data for better understanding. Not scientific. I'll be sure to get my mobo software to measure and log everything.

Good news about Catalyst 9.6. I still can't get CCC to work properly. Might boot in safe mode, driver clean, install 9.4, then install 9.6 given my poor experience with 9.5.

P.S: Currently in BIOS the CPU voltage is set to AUTO, however, do you think I should run small FFTs with increasing voltages until I get a stable clock? Good strategy to go about it or not?

I appreciate the help immensely by the way. Tell me what I'm doing wrong\right\whatever your personal approach or opinion on the matter is. I do appreciate it and take it into consideration. Which is why I love this forum.

 

[sirmonkey1985]

if you are overclocking.. you should never use the auto setting on voltages.. rarely does it ever set it correctly.. load up cpu-z.. put the processor under full load and check the voltage in cpu-z.. then go back into your bios and set it to that voltage..

also if its failing on the large FFT's.. you may have to increase your northbridge voltage.. but first id say run the small FFT's.. this will primarely test your cpu only and not really touch the ram or northbridge.. if it fails then you know you need to increase something like the voltage.. if it passes then you are good.. and either the northbridge voltage or ram voltage/timing needs to be changed..

if/when you get the system stable running p95 for a couple hours.. i recommend running IBT(intel burn test) to truely make sure its stable.. do 5-10 passes/runs on IBT if it passes.. then your good..