Video & Power Diagnostics for Bluescreen error

[TVeblen]

I am having a blue screen issue on a Windows 7 64 bit installation
involving my video driver nvddmkm.sys for a EVGA #896-P3-1255AR GeForce
GTX260 Core 216 video card. This is the Timeout Detection and Recovery
(TDR)- "the nvddmkm.sys display driver has stopped responding and
restarted" issue that is quite common on the web.
Please note that I have already tried all the recommended fixes such as
installing, reinstalling the most current drivers (nVidia 195.62),
checking video temps with GPU-Z, reseating the card, checking for cat
hair, etc etc.
I believe I have gotten to the point where it is either a bad video card
(9 months old, working perfectly on XP-SP3) or poor power supply (less
than needed voltage). I have a Corsair CMPSU-750TX on a P6T motherboard
with 3x2GB Corsair memory, 2 SATA hard drives, and a DVD. The card
requires a minimum 500 watts.
What is the best way to test the power supply under load? I have an
analog multimeter.

 

[Paul]

I am having a blue screen issue on a Windows 7 64 bit installation
involving my video driver nvddmkm.sys for a EVGA #896-P3-1255AR GeForce
GTX260 Core 216 video card. This is the Timeout Detection and Recovery
(TDR)- "the nvddmkm.sys display driver has stopped responding and
restarted" issue that is quite common on the web.
Please note that I have already tried all the recommended fixes such as
installing, reinstalling the most current drivers (nVidia 195.62),
checking video temps with GPU-Z, reseating the card, checking for cat
hair, etc etc.
I believe I have gotten to the point where it is either a bad video card
(9 months old, working perfectly on XP-SP3) or poor power supply (less
than needed voltage). I have a Corsair CMPSU-750TX on a P6T motherboard
with 3x2GB Corsair memory, 2 SATA hard drives, and a DVD. The card
requires a minimum 500 watts.
What is the best way to test the power supply under load? I have an
analog multimeter.

You test a power supply with a Chroma tester Nobody in your
town owns one. But they're what would be used at the factory
to test power supplies. They can apply a 500W or a 1000W load
instantly, and evaluate the response of the supply. They can be
totally automated, to give pass/fail as supplies are connected.

http://www.chromausa.com/newsletter/8000-systems.jpg

Some review sites own smaller versions. But for the price,
you could buy many replacement supplies instead.

http://www.trustedreviews.com/peripherals/review/2009/04/24/500---700W-Power-Supply-Group-Test/p3

You can use your multimeter, to verify the PCI Express connector
has 12V on it, but that isn't going to allow you to catch a transient
if one existed.

*******

If I had your problem, I would try to test using another OS.
Whatever you've got, that has drivers.

With some amount of hair pulling, I can set up Quake 3 Arena to
run under Linux. I can do that, because I have the original game
CD with the map files on it. There is a Linux executable available
for download, which I got. Then, I'd install the "tainted" driver
from Nvidia, to get all the performance possible. And still, after
all that effort, I might not be using even half of the compute
power of the video card. So while I could get some kind of 3D load
on my video card, it would be far from a perfect test. Windows just
has better test cases to run.

I have the evaluation copy of Win7 here. I have my WinXP CD. I have
my Win2K CD. The Win2K would be useless, because drivers stopped shipping
some time ago for that. That leaves WinXP as my best alternative for
driver testing, as an alternative to Win7.

Since you have mentioned

"9 months old, working perfectly on XP-SP3"

you have already tested it on an alternate OS. So that tells you it
isn't necessarily a hardware issue. You really need to find a test
case for the card, that won't "melt" it, but still give it a workout.
There is at least one stress tester, which can damage a video card
(and I don't remember the name of it now, but you may find references
to it).

If you run out of stuff to try, look for some of the
Futuremark 3D packages.

http://majorgeeks.com/3DMark06_d4935.html

*******

If you have a spare disk, do a clean install of Win7, install
a driver, and retest. Does the problem still show up, when
you're using a pristine copy of the OS ? If I need to do a
quick test, I make sure all the software I need is available
locally, do my install with the network disconnected, so there
is no activation activity over my network interface.

Paul

 

[TVeblen]

You test a power supply with a Chroma tester Nobody in your
town owns one. But they're what would be used at the factory
to test power supplies. They can apply a 500W or a 1000W load
instantly, and evaluate the response of the supply. They can be
totally automated, to give pass/fail as supplies are connected.

http://www.chromausa.com/newsletter/8000-systems.jpg

Some review sites own smaller versions. But for the price,
you could buy many replacement supplies instead.

http://www.trustedreviews.com/peripherals/review/2009/04/24/500---700W-Power-Supply-Group-Test/p3


You can use your multimeter, to verify the PCI Express connector
has 12V on it, but that isn't going to allow you to catch a transient
if one existed.

*******

If I had your problem, I would try to test using another OS.
Whatever you've got, that has drivers.

With some amount of hair pulling, I can set up Quake 3 Arena to
run under Linux. I can do that, because I have the original game
CD with the map files on it. There is a Linux executable available
for download, which I got. Then, I'd install the "tainted" driver
from Nvidia, to get all the performance possible. And still, after
all that effort, I might not be using even half of the compute
power of the video card. So while I could get some kind of 3D load
on my video card, it would be far from a perfect test. Windows just
has better test cases to run.

I have the evaluation copy of Win7 here. I have my WinXP CD. I have
my Win2K CD. The Win2K would be useless, because drivers stopped shipping
some time ago for that. That leaves WinXP as my best alternative for
driver testing, as an alternative to Win7.

Since you have mentioned

"9 months old, working perfectly on XP-SP3"

you have already tested it on an alternate OS. So that tells you it
isn't necessarily a hardware issue. You really need to find a test
case for the card, that won't "melt" it, but still give it a workout.
There is at least one stress tester, which can damage a video card
(and I don't remember the name of it now, but you may find references
to it).

If you run out of stuff to try, look for some of the
Futuremark 3D packages.

http://majorgeeks.com/3DMark06_d4935.html

*******

If you have a spare disk, do a clean install of Win7, install
a driver, and retest. Does the problem still show up, when
you're using a pristine copy of the OS ? If I need to do a
quick test, I make sure all the software I need is available
locally, do my install with the network disconnected, so there
is no activation activity over my network interface.

Paul

hey- Thanks for the info Paul. I'm just running out of options and
getting frustrated here. I suspect that it is the 64 bit OS that is the
root cause. What's confusing is that it's such a common but inconsistent
problem.
The TDR errors started right out of the gate on a fresh, clean install.
But updating the video drivers was the second thing I did, so not much
of a test on the original driver installed by windows.

 

[TVeblen]

hey- Thanks for the info Paul. I'm just running out of options and
getting frustrated here. I suspect that it is the 64 bit OS that is the
root cause. What's confusing is that it's such a common but inconsistent
problem.
The TDR errors started right out of the gate on a fresh, clean install.
But updating the video drivers was the second thing I did, so not much
of a test on the original driver installed by windows.

In Bios, my 12V reading is 12.193v. But Speedfan reports 11.46 to
11.51v. I know these are not very accurate readings, but if the 12v rail
was really only putting out 11.5v, would that be enough to cause video
instability?

 

[larry moe 'n curly]

In Bios, my 12V reading is 12.193v. But Speedfan reports 11.46 to
11.51v. I know these are not very accurate readings, but if the 12v rail
was really only putting out 11.5v, would that be enough to cause video
instability?

The BIOS is more trustworthy than any voltage/temperature monitoring
utility, but I once had a mobo where the BIOS said the +12V rail was
11.6x V when my meter read 12.2x V and SpeedFan said it fluctuated 1-2
times a second between 6V and 8V.

Measure +12V at the yellow wires, +3.3V at the orange or brown wires,
+5.0V at the red ones. Also try to measure the CPU Vcore (some of the
pins of the MOSFETs should have it), the PCI-e bus (probably at 1-2
MOSFETs or a power chip near the PCI-e slot, but a fast card won't use
it and instead have a direct connection to a +12V cable), the memory
voltage regulator (near the DIMM slots).

 

[TVeblen]

The BIOS is more trustworthy than any voltage/temperature monitoring
utility, but I once had a mobo where the BIOS said the +12V rail was
11.6x V when my meter read 12.2x V and SpeedFan said it fluctuated 1-2
times a second between 6V and 8V.

Measure +12V at the yellow wires, +3.3V at the orange or brown wires,
+5.0V at the red ones. Also try to measure the CPU Vcore (some of the
pins of the MOSFETs should have it), the PCI-e bus (probably at 1-2
MOSFETs or a power chip near the PCI-e slot, but a fast card won't use
it and instead have a direct connection to a +12V cable), the memory
voltage regulator (near the DIMM slots).

All my voltages are great. No fluctuation opening, running programs.
Rock solid.
You don't know how much I want this to be a Doh! issue, where I can feel
really stupid but everything is working fine!