KERNEL_STACK_INPAGE_ERROR... Whhaa?

[BlueMagician]

In the middle of playing HL2 or CS:S (online and offline), and:

The sound stutters uniformly for a second or two.
The screen goes black for a second.
I hear my Hard Discs spin down.

Then I get a BSOD stating: "KERNEL_STACK_INPAGE_ERROR", whilst further down the page it details STOP 0x00000077, yada yada..



I'm using NVidia Display drivers 66.93 (have tried 71.84) and NVidia NF4 Standalone Drivers kit 6.53. MemTest passes with no errors at stock 2-2-2-5-1T @ 200 @ 2.7v, and loosening memory timings has no effect on whether the BSOD appears or not.

My load temps are:
CPU < 50c
GFX 1 & 2 ~< 90c
PWM IC < 55c
Chipset < 50c

Looking on the MS site, this error is due to Windows not being able to read info from the pagefile, but is it really this, or is it something else in disguise? I don't have that many games that I play often, and in fact HL2 (Source engine) is my staple diet at the moment, but I've never had any other games do this, so I'm confused.


I guess I'm really after opinions on whether this is an issue with HL2/CS:S, or whether I do genuinely have a hardware issue that needs tracking down. About the only thing I haven't done is reinstalled XP without RAID-0 in place, but surely if there were issues with NV-IDE drivers they'd be the same, RAID or no RAID..


Sorry for babbling, advice needed!
S.

[skuzgib]

Might just be a one off - happens sometimes.

If everything still works and it doesn't happen again soon I wouldn't worry.

[BlueMagician]

Might just be a one off - happens sometimes.

If everything still works and it doesn't happen again soon I wouldn't worry.

No, I'm afraid it happens 50% of the time I play HL2. I can't really say it is Half Life 2 at fault, because I'm not actually playing a lot else at the moment - but it happens regularly, and I've never hand it happend in Doom3, or any benches for that matter.

I've just found an intriging comment by a guy on an Anandtech forum discussing these NVstandalone drivers, and quote:
"These drivers still have the bug in them where the the PC will become unresponsive when doing a lot of HD work with a RAID 0 setup on the NV controller. The interrupt process which is an invisible process under System Idle Process will hog the CPU totally at random. I really regret setting up a RAID 0 on this board."

Never heard of this one, but sounds ominous.


If there are issues with NVRAID lagging up under load, which is what the BSOD suggests, I wonder if its the Windows drivers at fault, or the ones in BIOS?

If I didn't already have quite a mature installation of XP setup just how I like it, I'd reformat and scrap the RAID array to eliminate it - but I REALLY dont want to do that..


Any thoughts?
S.

[BlueMagician]

Just an update and clarification..

Last night I ran MemTest for 7 hours - 20 passes - no errors.

I then booted into Windows. I simulataneously ran RTHDRIBL, Prime95 (blended torture), and had MediaPlayer10 playing a WMVHD clip "Step Into Liquid" with 5.1 Surround decoding.

I very much doubt that all those apps got a fair share of CPU time, but all the same:

CPU Temp < 48c
GFX 1 ~85c or less
GFX 2 ~95c or less
PWM IC < 50c
Chipset < 50c

CPU usage was 100%, PageFile usage was around 1GB, and there was hard disc activity aplenty, so I consider that a stressed but stable system. That little lot ran for 3 hours with no BSODs or errors, then I got bored and stopped it..


I'm just about to beat it to death with an hour or so of Doom3. A lame game, but good eyecandy, and should be a workout for the system.


It anyone else inclined to agree that this is looking like an issue with HalfLife 2 and/or the source engine? Of course theres an element of hope in my words that it is just HL2 being a pain, but I would think that if it were a hardware or driver problem then it would have shown up elsewhere by now...?


Thoughts welcome,
S.

[Lee H]

I've been having a few similar issues since I installed the Newest nvidia graphics drivers with WoW and CS:S. Basically they'd crash out and reset the system anywhere between 10 mins and 6 hours use.

I rolled back to the previous version I had installed and guess what... not a problem.

Totally bizarre

[BlueMagician]

I've been having a few similar issues since I installed the Newest nvidia graphics drivers with WoW and CS:S. Basically they'd crash out and reset the system anywhere between 10 mins and 6 hours use.

I rolled back to the previous version I had installed and guess what... not a problem.

Totally bizarre

I've tried the 71.84's too as they have official support for Hardware DVD decoding which would be nice, but the last time I tried to install them, they BSODed XP on Startup in anything more than safe mode.

It took me about half an hour of uninstalling/reinstalling/drivercleaning to get 66.93's back on and stable, so I won't be trying them again in a hurry.


Unfortunately, all the issues I've described so far have been with the 66.93 drivers that people seem to call stable, so I'm a bit lost at the moment.


WM, what errors did you encounter?


S.

[yamangman]

It does very much sound like a PSU issue, hard disks are usually the first to go when they arn't getting power.

[Steve]

It does very much sound like a PSU issue, hard disks are usually the first to go when they arn't getting power.

Was going to suggest this, then look at his sig and see his wonderful PowerStream Unlikely that it's croaking, but if it's faulty then maybe I guess.

[BlueMagician]

Was going to suggest this, then look at his sig and see his wonderful PowerStream Unlikely that it's croaking, but if it's faulty then maybe I guess.

I've just had my PSU bench/load tested by an OCZ Engineer, so I'm pretty confident it's not that. Good call though..


Anyways, I played Doom3 for a couple of hours, with no problems.. Then I moved to a HalfLife2 DeathMatch..

Within 20 minutes (through 1 and a half rounds), it froze..
Now, the last two times this has happened, it's not BSOD'd, it's just sat there with the sound going "st.st.st.st.st.st.st.st.st.st.st"..

Anywho, THIS time, for the first time, I got something different in the XP eventlog. It was logged at the exact time that HL2DM messed up:

Source: Disk
Category: None
Event ID: 51

Description: An error was detected on device \Device\Harddisk0\D during a paging operation.

So, it would seem that this, and the worse BSOD's indeed point to Windows not being able to access the PageFile from the HDD(s).


I find it hard to beleive that there is a physical problem with the drives, as they're brand new Raptors, Windows seems happy enough when working around the desktop, and there are no errors whenever I force a Scandisk.


So, it's either:

> BIOS/Drivers (again unlikely else Windows and other games/apps would show signs of nastyness), OR
> Something to do with the way Windows/HalfLife/Source handles Memory Management.


EDIT: A few more notes:
> I've been running with DEP disabled for a couple weeks now - no change.
> I just disabled HDD SMART and my unused Silicon Image Controller in BIOS, turned off NCQ on the NVidia Controller in Windows, rebooted, and HL2DM crashed with the same error in within 3 minutes.

EDIT2: I don't know what made me think of it, but I just checked IRQ's.
I have two NForce4 ADMA Controllers listed in Device Manager. One has nothing connected to it, and the other has the two SATA Raptors linked in it's properties.
Now, the one that has the drives mapped to it, is sharing the same IRQ (21) as the NVidia Network Bus Enumerator - Could this be something, or am I barking up random trees?

I got into thinking that of course there is a lot of NIC activity when I'm playing HL2DM/Source online and might be clashing with the NVidia ADMA controller. No no, I thought, that wouldn't explain why it does crashes when I'm playing a single player game. BUT, what if Steam IS doing something over my net connection at random times, regardless, during both single player or online play - demanding NIC activity, and again, clashing?

To test this theory, I either force one or the other to a different IRQ (?how?), or swap to my other onboard NIC (ie: not the NVidia one) and see if that cures anything.


Am I losing the plot? - I'm certainly starting to lose the will to live... Argh.
S.

[BlueMagician]

...To test this theory, I either force one or the other to a different IRQ (?how?), or swap to my other onboard NIC (ie: not the NVidia one) and see if that cures anything...

Okay, so that theory is out of the window. I changed to the second onboard (non-nforce4) NIC, and got about 8 minutes of HL2DM in, before I was struck with the good old chestnut:

Application popup: Half-Life 2 DM: hl2.exe - Application Error : The instruction at "0x2406dc59" referenced memory at "0x010811cc". The memory could not be "read".

<Exit to desktop>

This is starting to get on my pip now..
Please feel free to read my ramblings above, comments most welcome,
S.