Chaintech VNF4-Ultra, Dual Core, World of Warcraft, crashes... what's going on?

Carl B

Friends call me xbd
Legend
Ok... well I built a computer for my brother a couple of months ago, and for Christmas got him a dual core 3800+.

Stock, overclocked, it doesn't matter - for some reason WoW keeps crashing on him. Same with CounterStrike sometimes. Any guesses as to what's up? MemTest and SuperPi both run fine when it *is* overclocked; but that's a moot point anyway. At first I thought it was a Creative card/NForce4 issue, so I had him disable the Audigy - still crashed.

Does anyone have any ieas what might be going on here? The manual advises against running four sticks of DDR400 - should I take it down to DDR333? Don't see why it would have been stable in Memtest though if that were the case.

I admit to not running Memtest past two passes - maybe I'll have him run it all night tonight. Right now I'm having him run Prime95, but I'm at a loss as to why this is happening.

Worked fine before the dual core.
 
Replicate the problem in a controlled environment, i.e. one hour of WoW and it crashes.

Once you can do that change one thing at a time and I would head for the memory first since it is the easiest to look at.

Change settings to DDR333 like you mentioned.
Still crashes then add only one stick to motherboard and retest.
Do this for all 4 sticks.
If you get crashes in any configuration then it is not likely to be the memory.
Next move on to another component. Change the component retest etc.

Before all this make sure you have the latest BIOS and drivers and make sure it is not a software problem (i.e. clean install).
 
Does it only crash on WOW? If so I doubt it is some random memory bug, it is most likely a bug in WOW right?
 
Yeah, usually if it's a memory stability issue, my machine will tend to reboot randomly. If it's just the application that crashes, I find it's usually a bug with that application (which can sometimes be fixed by setting the process affinity to single processor in the task manager, though WoW should have fixed that by now).
 
Nah it's Counterstrike too it seems. I'm thinking for some reason maybe it's the NVidia drivers. I'm having my brother upgrade to the latest NForce 4 and then back to 78.xx series for his 6600GT before pursuing the root hardware any further.

But he hasn't been around for me to have him do this...

Now, I'm not sure *why* it would be these drivers, save that he's got a dual core, the newer drivers he has are dual-core 'optomized,' and the problem always seems to occur during a spike of audio/compute activity in these games (a grenade explodes, a sword clash, etc...)

I'm at something of a loss, but without being next to the PC, my diagnostic methods are limited.
 
Yeah, I've had problems with a few games with the 80-series drivers on my dual-core machine. City of Heroes/Villains had problems around the release time of CoV (basically, the game would just stop rendering a few seconds into the game: the system didn't lock up. It looked like it was caught in an infinite loop, probably a race condition or somesuch). I also just recently went back and tried to play KoTOR2, and experienced framerates of about 1fps. Going back to the 78.01 drivers fixed the problem. Users have reported the same problem with the original KoTOR (Haven't tested it myself yet).

I really hope nVidia fixes these issues soon. The 80-series drivers really offer so much. Might be nice to have a compatibility check to disable multithreading, but I somehow doubt that'd be an easy thing to do.
 
Chalnoth said:
Yeah, I've had problems with a few games with the 80-series drivers on my dual-core machine. City of Heroes/Villains had problems around the release time of CoV (basically, the game would just stop rendering a few seconds into the game: the system didn't lock up. It looked like it was caught in an infinite loop, probably a race condition or somesuch). I also just recently went back and tried to play KoTOR2, and experienced framerates of about 1fps. Going back to the 78.01 drivers fixed the problem. Users have reported the same problem with the original KoTOR (Haven't tested it myself yet).

I really hope nVidia fixes these issues soon. The 80-series drivers really offer so much. Might be nice to have a compatibility check to disable multithreading, but I somehow doubt that'd be an easy thing to do.

Yeah, endless loop is exactly what happens to him. I was just using the term 'crash' for expediency. Good to know that my driver theory is being bourne out here in other cases. Thanks a lot for this story of yours Chalnoth, I knew dual-core implementation on NVidia's first go had to be buggy in *some* respect. I agree though, 80-series is so good in so many ways; hopefully they have another certified release soon.
 
Oh, I just found out that nVidia did indeed implement an option to disable dual-core optimizations in their latest driver release! Haven't tested it yet, though.
 
Ok, I come back here a broken man...

Something is wrong here, but I don't know what.

I 'system restored' the thing to right after a clean install of XP and SP1 and 2 had been put on the computer, I Prime95'd the processor for 27 hours (does that adequatetly stress a dual-core CPU by the way?) in a different system, I MemTest'd the RAM for 24 hours (also in another system), it all worked...

I put my brothers old Venice core in his motherboard with some generic RAM, and he could game fine. Put the dual-core back in there, and it's crashing again like ~20 minutes into Counterstrike.

What is going on? I'm at a total loss. Should I try gaming on the alternate system with the dual core?

Right now I almost feel that the problem *must* be either the processor or my brothers motherboard, but they both seem so well-behaved otherwise when not coupled to each other. I hate it when stuff like this happens.

PS - I've tried several BIOS versions on the mobo, no change in outcome - even when the system doesn't even detect the second core. I've disabled sound cards, tried different NVidia drivers... all the usual suspects.
 
Last edited by a moderator:
Okay, a couple of things to try:
1. Drop back to the 7x.xx series drivers.
2. Alternatively, try the registry hack I wrote about here on the 81.98 drivers (note: this should be identical to dropping back to the 7x.xx series drivers, if dual-core optimizations are the problem).
3. You could set the process affinity to use only one CPU in the task manager. You can only do this while the program is running, and it will default to allowing operation across both cores upon next execution.

Anyway, one thing to make sure of is that Windows is actually making use of both cores. Does the Windows task manager report two CPU's?
 
Chalnoth said:
Okay, a couple of things to try:
1. Drop back to the 7x.xx series drivers.
2. Alternatively, try the registry hack I wrote about here on the 81.98 drivers (note: this should be identical to dropping back to the 7x.xx series drivers, if dual-core optimizations are the problem).
3. You could set the process affinity to use only one CPU in the task manager. You can only do this while the program is running, and it will default to allowing operation across both cores upon next execution.

Anyway, one thing to make sure of is that Windows is actually making use of both cores. Does the Windows task manager report two CPU's?


Yeah when the proper BIOS is in there, Windows reports two cores, and shows activity on both. Been back on 7x series for a while now, and the thing is - even when the BIOS *does not* support the dual core, and Windows is reporting only a single core, the thing still crashes.

I don't know, I'm going to try going down to 2x512MB on the RAM in this board from 4x512 and trying the CPU with a non-supporting BIOS coupled with 7x.xx drivers just to make sure it's not a memory issue somehow. *If* that works, then I'll move to a dual-core supporting BIOS, and keep inching up from there in terms of 'break-points.' But the truth is I just don't expect the RAM to be the issue here.

(When a motherboard does not support dual-core, is a core chosen at random upon boot-up each time, or is it a specific core that is at all times designated 'core 0'?)
 
Well, I would not expect desired operation unless the BIOS was written with support for the CPU, so I would not even bother with that configuration.

If the problem is invariant under a change between the 7x-series and 8x-series drivers, then I would claim that the drivers are not a problem. There still may, however, be a problem with the program itself. Setting the affinity as I described above should isolate this problem. If the problem still appears even when the game is limited to using only one core, then it's not related to multithreading.

Other than that, the memory is a good thing to check. If it works fine with just two sticks, try putting all four back in and dropping the memory clock.

Edit:
One thing to bear in mind is that while it should work, the simple fact that a BIOS update was required to support the CPU means that that motherboard never got any QA testing with the dual-core CPU's, and as such there may be some that don't work with dual-core CPU's.
 
Chalnoth said:
Well, I would not expect desired operation unless the BIOS was written with support for the CPU, so I would not even bother with that configuration.

Well I wasn't so much expecting it to 'work' with the non-supporting BIOS, just expecting to see whether something about the newer BIOS destabilized operation vs the older BIOS.

Here's how it's going down though: I'm re-installing the dual core and the memory into the newer motherbaord/other system, and I will then attach my brothers hard drive to it, update chipset drivers, etc...

We'll then see if the crashes continue. If so, then either it's the CPU or it's the OS install somehow (or other program conflict). If it *does* work in that configuration, well then it's finally been isolated as the older motherbaord (present frontrunner for replacement, but I've go to prove it first).

We'll take it from there I guess, already have a new mobo option selected should we need to go there, but if it's another problem, well that's just going to be further hassles as OS's get re-installed, CPU's further tested, etc...
 
First, thanks to everyone that's chimed in on this thread - it's been an ordeal to be sure.

Anyway after extensive testing, and actually going through with my cross-system component by component check I was looking to avoid, I can with 99% security say it was the original motherboard simply not running stable with a dual-core processor, in spite of a BIOS update making it compatible.

Anyway to got a Biostar T-Force to swap the Chaintech out with, and through some clever system restoring and other hacking got it working without needing to do a complete Windows re-install.

I think the system's where it needs to be now, but I've got a new annoyance in the way of Windows asking for drivers everytime the system boots.

Specifically, for: "Nvidia CK804 ADMA Controller (v2.7)"

and: "Nvidia CK804 Parallel ATA Controller (v2.7)"

I've read around and it's actually not uncommon that other users experience this same constant nagging from Windows with regard to these NVidia specific drivers/features. But does anyone have a suggestion as to how I might shut Windows up on the matter?
 
Well, I'd uninstall nVidia's motherboard drivers through the control panel's add/remove programs, then reinstall the latest nForce drivers from nVidia's site. This is because my first guess as to the cause of that problem is an incomplete driver install.
 
Chalnoth said:
Well, I'd uninstall nVidia's motherboard drivers through the control panel's add/remove programs, then reinstall the latest nForce drivers from nVidia's site. This is because my first guess as to the cause of that problem is an incomplete driver install.

You'd think, but in fact after doing that a couple of times, each time the system was left completely unstable; ie would not boot. It was a couple of hours of driver installs, uninstalls, and re-installs that left me settling for the present configuration. At best I can uninstall the NVidia drivers and rely on Microsoft's IDE drivers, but then the Gigabit LAN and the printer port are left in the lurch for some reason, and independent Ethernet driver installation doesn't correct it. Printer port I don't care about, Ethernet I do.

If you search Google for: 'nvidia CK804 ADMA' you'll see this is none too uncommon, and once installed it's difficult to undo; not to mention results for it don't seem consistent system to system, save that those drivers prove problematic.

I'm pondering Windows re-install vs living with th epop-ups (unless there's a solution) vs Windows IDE drivers + 3rd party NIC, but for now I'm just happy the dual-core drama is behind.
 
If your mobo is anything like my Asus board, certian nvidia drivers seem to work with specific versions of the bios, others cause terrible problems. Check what driver version your mobo manufacturer recomends you use for the bios and mobo. I really don't know if this will help you, this has just been my experience.
 
Back
Top