PDA

View Full Version : Getting lockups running DX10 - help please.


Pages : [1] 2

FlakMagnet
11-14-07, 01:06 PM
I currently have a horrible lockup problem when running the Crysis demo in DX10 mode - DX9 works flawlessly.

The demo locks up after running for some random and very short amount of time, and sometimes locks up in the menu (not even into the game itself).

When the machine locks, I get a screen full of (mostly) purple pixels, and the machine requires a reset to get it back. I've provided a screenshot of what it looks like when it locks to see if anyone recognises the issue and can suggest a solution.

The white block you see in the screenshot is actually the mouse pointer, and I can still move it around the screen. Sound also continues playing in the background at this point. But I cant Alt-Tab, or shutdown (Win->Tab->Tab->Space is the shortcut I've tried for shutdown). Have also tried Ctrl-Shift-Esc and Ctrl-Alt-Del, and nothing seems to do anything.

I have tried many different drivers, but I'm currently running 169.09 (other 169 and 163 drivers also exhibit the same crash). I first had this lockup with BioShock and the 163.xx drivers, but I only saw it once. I managed to complete BioShock in DX10 after that, and never saw it again... until Crysis.

I can run Orthos on my machine without any problems (ran for > 9 hours the other night), and can also run other DX10 games/apps without an issue (Cascades and Human Head run for hours without any problems). Also running UT3 Demo without a single issue, but not sure if that is DX10 or 9.

Wh I reboot after a lockup, Vista sticks up a dialog (after I log in) that the crash was in the NVidia display driver. Sorry, but can't remember the exact message. Also says it checked for a solution but found none.

I have the 3 Vista hotfixes that improve performance and stability installed (sorry, can't remember the numbers). The other hotfix is for SLi which I don't have, so I have not installed the hotfix.

If anyone can shed some light on what might be going wrong, or to even suggest a fix, I would be forever in your debt :D

Screenshot of lockup :
[img=http://img252.imageshack.us/img252/4410/pb140209rx5.th.jpg] (http://img252.imageshack.us/my.php?image=pb140209rx5.jpg)

tb77
11-14-07, 06:56 PM
the only solution for me was: (had the error on 2 systems, athlon x2 4800 and q6600) try ntune and reduce you video card clocks (memory and gpu) - i had a bfg 8800 gtx oc2. now it runs at 600/1400/900 instead of 626/1400/1000. so it simply was an overheating issue! try crysis gpu benchmark for an hour or 3dmark06 for 8 hours.

Thomas

Monster909
11-15-07, 01:31 AM
That sounds definitely like "over" OC:ing. As previous poster said, try reducing clocks, that is if you are OC:ing and havent tried that already. If you are not OC:ing, check if youre card is pre-overclocked by the vendor. If it is, try reducing clocks with rivatuner and see if it helps, if it does then try to down clock it permanently.

ciaspook
11-15-07, 02:31 AM
I'm having the same problem in Vista x64. Keeps locking up in DX10 in Crysis. Works just fine in DX9. And nothing is overclocked or overclocked by vendor. Heat isn't the problem.

tb77
11-15-07, 08:06 AM
I'm having the same problem in Vista x64. Keeps locking up in DX10 in Crysis. Works just fine in DX9. And nothing is overclocked or overclocked by vendor. Heat isn't the problem.

if this error happens while playing the game, i would think of overheating.
can you try to underclock your video card via ntune ~ 100 mhz core + mem? if this does not help, then you might have a differrent issue, maybe the power supply, but i would guess gfx overheating for some reason, even if it is not overclocked.

otherwise, this error might happen when you test a lot of different gfx + detail settings, thenit is not caused by heat.

Thomas

FlakMagnet
11-15-07, 12:28 PM
Many thanks for the help guys, and I agree that it does look like overheating, so I'll try underclocking the card (not overclocked at the moment).

However, I can run NVidias Cascades and Human Head for hours on end without any problems (Cascades is DX10, not sure about Human Head), so not sure if it is heat.

Will try an underclock and report back...

FlakMagnet
11-15-07, 01:04 PM
OK, I underclocked the GPU Core to ~500MHz from 576MHz and the RAM to 800MHz from 900MHz using nTune. I then used nTune Monitor to log the GPU temps every second to a log file and fires up the Crysis Demo.

This time, it locked in the menus after changing all the graphics options to Very High - still in the menu. After rebooting, I checked the log file, and the last reported temp. was 63 Celcius. So I don't think it's a heat issue.

Surely if it's related to game settings, then more people would be seeing this?

Any more ideas people? Crysis is released here in the UK tomorrow, and I sooo want this to work...

tb77
11-15-07, 03:07 PM
is your system CPU + memory overclocked too? try it with the default frequencies, just as a test. change the crysis settings, exit the game. then reboot and run the game with the new settings. changing gfx settings in crysis esp. very high caused this error for me too, but once they are applied the game runs fine.

if this does not help, uninstall the demo / game and kill all related files in "program files" and "my games". then reinstall it and try it again.

Thomas

LittleCupcakes
11-15-07, 03:48 PM
The first time I ran Crysis (Vistax64) I also had this issue in the menu, at exactly the same time (that is, when changing graphics settings to very high). Hard lock, in the menu, hard shutdown required.

Next time I played, a blue screen followed showing the video driver crash (nvdkllm or whatever).

I was overclocked, but Crysis has worked nearly flawlessly on very high since then (one other CTD, no BSOD/lockup) and I'm STILL overclocked.

Maybe a bug in applying settings?

Vistax64
E6700@3.26, 1.25v post vDroop
2GB Patriot XBLK @ 973 4-4-4-11, 2.2v
Running Crysis @ 1680x1050, Very High
EVGA 8800GTX 622/933, 169.04

ciaspook
11-16-07, 08:39 AM
Well, nothing is overclocked. Haven't tried underclocking. All other games work just fine (Call Of Duty 4, Ghost Recon 2, World In Conflict (DX10). So don't think its overheating. And its not PS either, 700W is more then enough for my HW.

FlakMagnet
11-20-07, 03:45 AM
OK, I've finally got to the bottom of the issue (at least I hope I have).

After tb77's suggestion of underclocking, I finally went the whole way and underclocked as much as possible (as low as nTune would allow). This worked flawlessly for a couple of hours in Crysis DX10 (although very very slow). I then went about upping the core and memory clocks a bit at a time until I got instability.

I ended up with the core at stock speed (576MHz), and the memory at 792MHz (stock is 900). If I up the RAM just one notch (to 810), the game will crash eventually. With 792, I played for over 4 solid hours without a problem.

I contacted BFG and sent them all the info, and they are going to RMA the card. Hope I can get an advance replacement as I don't want to loose my PC now - not when I'm still playing Crysis and UT3 is out next week!!! :D

So, many thanks to all who have helped out here, and special thanks to tb77 for suggesting underclocking.

P.S. I used EVEREST to monitor all the temps. on the system, and nothing went above 'normal' with the RAM @900, so not sure why it is locking up. Also not sure I understand why nothing using DX9 locks up with RAM @900 - not even Crysis.

CoSWill
11-21-07, 09:04 AM
Just to let you know......UT3 is out now :)

FlakMagnet
11-22-07, 06:24 AM
But not in the UK :thumbdwn:

Unless you want to illegally download it, which I don't. Will buy it when it comes out here on 30th November. But it has been a very long wait :(

FlakMagnet
11-22-07, 06:31 AM
Tried a friends EVGA 8800GTX in my machine last night, and everything (including Crysis DX10) worked flawlessly... Card on it's way to BFG for fixing...

raydun
11-22-07, 01:20 PM
UT3 is actually out tomorrow in UK, mine was posted today by ShopTo.COM,
the e-mail said it will be here tomorrow. Very good delivery service, the last 4 games I have preordered have all arrived on day of release or earlier.

FlakMagnet
11-22-07, 03:54 PM
UT3 is actually out tomorrow in UK, mine was posted today by ShopTo.COM,
the e-mail said it will be here tomorrow. Very good delivery service, the last 4 games I have preordered have all arrived on day of release or earlier.

Hey, I just checked and you are right!!! Thought release date was 30th!

Damn. I've just finished packing my graphics card up to send to BFG :headexplode: Guess I won't be playing that for a couple of weeks :(

Greg
11-23-07, 09:23 PM
If you go to...

Computer Management (Right click MyComputer)
Event Viewer
Windows Logs
System

Scroll down the list, do you see yellow warning triangles with Source Display, containing the text "Display driver nvlddmkm stopped responding and has successfully recovered." ?

FlakMagnet
11-27-07, 06:22 AM
Sorry, I really can't remember if I got those messages in the event log. I do remember seeing some sort of error messages in there, but can't remember details.

I have a new card on it's way from BFG (actually in a van about 2 miles from where I live at the moment), so hopefully, this issue is solved for me.

I must say just how very impressed I have been with BFGs support and handling of my RMA. I sent my card to them last Friday, and it got to thier offices in Wales (UK) on Monday morning. By 4pm the very same day, a new (or possibly just a replacement) card was on its way to me! Very impressive indeed :clap:

Greg
11-27-07, 07:54 AM
Sorry, I really can't remember if I got those messages in the event log. I do remember seeing some sort of error messages in there, but can't remember details....
Those logged event messages will still be there if you look. Good luck with the RMA, let us know if all is well when it returns.

FlakMagnet
11-27-07, 04:04 PM
Just finished Crysis with the new card - a LOT better than the old one I sent back! A couple of 'strange' things happened, and the worst was randon coloured triangles being painted on the screen, eventually filling the screen so I couldn't see what was happening in the game. Restarted it and everything ran great until the end.

Checked the Event Viewer, and couldn't see any of the events that you describe (even checked back as far as 2 weeks, and couldn't find any).

I did find the following errors though, which are a bit worrying : "IRQARB: ACPI BIOS does not contain an IRQ for the device in PCI slot 3, function 0. Please contact your system vendor for technical assistance."

Any idea what could be causing that? No idea which PCI slot 3 is, and there is nothing flagged in device manager as not working correctly. Strange.

Greg
11-27-07, 09:08 PM
A couple of 'strange' things happened, and the worst was randon coloured triangles being painted on the screen, eventually filling the screen so I couldn't see what was happening in the game. Restarted it and everything ran great until the end.
Sound like the overheating artifacts many people including myself report. You might want to check this by running nTune and showing the temperature graph. If it goes over 80degC or the fan speed slider does not automatically move beyond 29%, then you share the problem. I still believe this is a software/firmware issue.

No idea about the PCI thing. I wouldn't worry about it if your devices work properly and are listed in device manager without warning. Those event logs are full of weird and disconcerting messages.

FlakMagnet
11-28-07, 03:44 AM
Sound like the overheating artifacts many people including myself report. You might want to check this by running nTune and showing the temperature graph. If it goes over 80degC or the fan speed slider does not automatically move beyond 29%, then you share the problem. I still believe this is a software/firmware issue.

No idea about the PCI thing. I wouldn't worry about it if your devices work properly and are listed in device manager without warning. Those event logs are full of weird and disconcerting messages.

Thank for that Greg. Think I'll use EVEREST tonight to log and monitor the graphics temps and fan speeds while playing. 80degC doesn't sound very hot for a graphics card to me. Isn't the 'warning' temp set to something like 125deg? Think I'll put the fan up to 100% and see if that makes a difference. Am happy(ish) though if this issue is being reported by others because it means I'm not on my own which I was with my previous issue :(

I think Event Viewer in Vista is more difficult to find what you are actually interested in. The 'old' style Event Viewer was just nice and simple. But maybe it's just me not being used to it yet. Will ignore the errors I'm getting about PCI slot 3 ;)

Are you manually controlling your fan speed to stop the artifacts? Or is there no solution to the overheating?

Greg
11-28-07, 03:56 AM
Thank for that Greg. Think I'll use EVEREST tonight to log and monitor the graphics temps and fan speeds while playing. 80degC doesn't sound very hot for a graphics card to me. Isn't the 'warning' temp set to something like 125deg? Think I'll put the fan up to 100% and see if that makes a difference. Am happy(ish) though if this issue is being reported by others because it means I'm not on my own which I was with my previous issue :(

I think Event Viewer in Vista is more difficult to find what you are actually interested in. The 'old' style Event Viewer was just nice and simple. But maybe it's just me not being used to it yet. Will ignore the errors I'm getting about PCI slot 3 ;)

Are you manually controlling your fan speed to stop the artifacts? Or is there no solution to the overheating?
I said 80degC just because it is around the temp when the fan should speed up and cool the thing down. Yes, the GPU should handle temps up to 110degC or so, and higher before being physically damaged. I seem to get artifacts in the 95-100 range. I can confirm that my fan on 'auto' never changes from 29% speed setting. I manually set it to at least 60% before playing games. (I could use Rivatuner to auto configure, but I have had some problems with it and don't trust it.) I've seen a couple of 8800 reviewers have noted this issue now also.

Before you manually set your fan, or change the auto values, do monitor the temperatures and fan speed if possible. You may want to run game or benchmarks in a window for easier monitoring.

Yes, MS should have called Vista it 'the hardest to use OS ever!'

FlakMagnet
11-28-07, 10:31 AM
I said 80degC just because it is around the temp when the fan should speed up and cool the thing down. Yes, the GPU should handle temps up to 110degC or so, and higher before being physically damaged. I seem to get artifacts in the 95-100 range. I can confirm that my fan on 'auto' never changes from 29% speed setting. I manually set it to at least 60% before playing games. (I could use Rivatuner to auto configure, but I have had some problems with it and don't trust it.) I've seen a couple of 8800 reviewers have noted this issue now also.

Before you manually set your fan, or change the auto values, do monitor the temperatures and fan speed if possible. You may want to run game or benchmarks in a window for easier monitoring.

Yes, MS should have called Vista it 'the hardest to use OS ever!'

Ah OK, so it doesn't actually start to go wrong at 80deg. I'll monitor the temps playing some Crysis and see where they go. Will report this evening - still at work at the moment. EVEREST logs readings to .log or .html files, so I can play full screen to really stress it.

FlakMagnet
11-28-07, 01:10 PM
Just ran some tests in Crysis and logged the GPU temps and fan speed. It looks like the fan speeds up and down according to the temp, and the highest temp. I got was 80C.

Results from EVEREST :
GPU C - min = 54, max = 75, avg = 69.1
GPU Diode C - min = 61, max = 80, avg = 74.7
GPU Ambient C - min = 49, max = 64, avg = 59.9
GPU Fan RPM - min = 1409, max = 1627, avg = 1467

I have a much more detailed report that shows all the above readings at 5 second intervals, and they show the fan speeding up as the temps get towards 80C.