VOGONS

Known issues with 9800 GTX and latest driver 341.92?

Topic actions

First post, by vorob

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

Hey guys! Recently picked up a Toshiba Qosmio X300-13P (Core 2 Duo T9600 & GeForce 9800M GTX). I'm perfectly aware that the 9800 series is infamous for GPU failures (both desktop and mobile), so I took the time to thoroughly test the machine after buying it.

Here's how things went so far:

The laptop came with Windows 7 installed by the previous owner. While poking around, I twice noticed weird pink lines and color distortions on the screen. The drivers were from 2009. 3DMark06 ran fine, but Oblivion froze the system after ~15 minutes.

I wiped the system and installed a clean, trusted Windows 7. First thing, I put on the latest drivers (341.92). This time, I got a couple of blue screens (“nvlddmkm.sys” errors) just in Windows. 3DMark SM3 tests would launch, but if I tried the full run, it consistently BSOD’d after the CPU tests (I even have a crash video: https://disk.yandex.ru/i/bSZQMQP9xzcY2w).

Launched Oblivion again—played for a while, seemed stable. Tried Crysis on max settings, 10 minutes in, still no apparent problems.

At this point, I started thinking the issue might be with the drivers, not the hardware. So, I backtracked to the 331.65 drivers. Lo and behold, everything runs fine: 3DMark completes, no problems in Windows, no blue screens. Haven’t noticed any weirdness since. That said, I haven’t done any really long/stress tests yet due to time constraints.

But here’s my big question: is it normal for the latest Nvidia drivers to cause these kinds of issues on a 9800M GTX? I googled around but didn’t really find much about incompatibilities or critical bugs with the 341.92 driver and this particular GPU. Has anyone seen cases where the last available drivers actually work worse than older versions for legacy cards like this?

Tonight I plan to do some extended testing and long gaming sessions. In particular, I want to find a game that’s heavy on shaders and pushes every part of the GPU. Also, I want to separately test some PhysX workload.

Almost forgot: regarding temps, the GPU heats up slowly and then levels out around 80C. I opened up the system—looks like nobody ever really touched it, factory thermal pads still in place. Not sure there’s much point in repasting right now.

Would really appreciate any insight—especially experiences with late driver bugs on old Nvidia hardware. Is this a ‘normal’ driver issue, signs of imminent hardware failure, or what?

Reply 1 of 5, by vorob

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

My testing continues. First of all, the pink lines I saw at the beginning were caused by a bad connection between the laptop’s monitor and the video card. I had a similar issue once before when I reconnected this cable incorrectly, but this time the problem is now resolved.

I can say that 3DMark Vantage completed 15 runs without any issues. The same goes for 3DMark06. Lost Planet long run didn't give issue to me.

However, I did experience NVIDIA driver crashes (without BSOD) in S.T.A.L.K.E.R.: Clear Sky, Crysis (DX10), and the Unigine Heaven benchmark. S.T.A.L.K.E.R. was crashing in one particular spot, but after reinstalling the game, the error disappeared. As for Crysis, I’ve read that it creates a shader cache that should be cleared when you reinstall the driver; I still need to check that.

Right now, my goal is to figure out whether the video card is actually starting to fail, or if all of this is just the result of strange software or driver issues.

Reply 2 of 5, by The Serpent Rider

Posted on 
User metadata
Rank l33t++
Rank
l33t++

That sounds like overall system instability not related to GPU.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 3 of 5, by vorob

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

But it's always about Nvidia video driver crash. Best way to test everything at once?

Reply 4 of 5, by The Serpent Rider

Posted on 
User metadata
Rank l33t++
Rank
l33t++

Firstly: don't make assumption that the paste is "fine", that stuff is 15+ years old and worked under heavy condition, it can easily create hospots way above 80 degrees. Secondly: if the heaviest thing you launch is a game, of course it's going to crash on GPU driver part. Run memtest, run Prime 95, etc. Nail it down until you're certain that it's GPU.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 5 of 5, by tehsiggi

Posted on 
User metadata
Rank Member
Rank
Member
The Serpent Rider wrote on Today, 13:34:

Firstly: don't make assumption that the paste is "fine", that stuff is 15+ years old and worked under heavy condition, it can easily create hospots way above 80 degrees. Secondly: if the heaviest thing you launch is a game, of course it's going to crash on GPU driver part. Run memtest, run Prime 95, etc. Nail it down until you're certain that it's GPU.

I fully agree with that.

I had machines from this time having crustiest thermal paste after only a year of use. I'd not go an guess, give it some love, it won't complain.

And I agree on the point of ruling out the rest. Especially running MemTest. Some systems back in the day I repaired were doing the weirdest things because of faulty memory.

Give that a shot.

AGP Card Real Power Consumption
AGP Power monitor - diagnostic hardware tool
Graphics card repair collection

Go to top of page Go to top of page
Back to Video
Close window
Close window