I've installed an Nvidia 5900 in an Athlon S2460 dual MP 1200 system. Latest
video drivers.
This system has been very reliable. Recently, it's been loaded with an old
Cinemaster decoder, and I'm doing a lot more with video.

I'm getting blue screen NMI parity interrupts. They may be random, but
there's a hint of usage sensitivity. It ahppened twice in a row when I
played an m2v file from another machine on my home network. Doesn't happen
repeatably if I copy the file over to a local disk. Total number of
occurences is only three

The system runs FS 2004 without any problems, at 31 FPS and maximum detail.
All MS, BIOS, and video driver updates have been applied.

The motherboard is a server design that has buffered ECC RAM with a
background scrubbing process, no overclocking, seven fans, slow CPUs that
operate at only 39 and 44 degrees Centigrade respectively, and a 500 watt
supply. It's unlikely that ECC ram could generate a parity error. It would
have to be a system noise problem, or a parity error on the Nvidia board
itself, which is a Gainward "Golden Sample" that not being overclocked.

Does anyone have any similar experiences/fixes to share?

Robert Morein wrote:
> The motherboard is a server design that has buffered ECC RAM with a
> background scrubbing process, no overclocking, seven fans, slow CPUs that
> operate at only 39 and 44 degrees Centigrade respectively, and a 500 watt
> supply. It's unlikely that ECC ram could generate a parity error. It would


Unlikely, but have you actually tested the RAM? (memtest86)


-WD

"Will Dormann" <wdormann******.com.invalid> wrote in message
news:RTrvb.18588$uw5.3591@fe2.columbus.rr.com...
> Robert Morein wrote:
> > The motherboard is a server design that has buffered ECC RAM with a
> > background scrubbing process, no overclocking, seven fans, slow CPUs
that
> > operate at only 39 and 44 degrees Centigrade respectively, and a 500
watt
> > supply. It's unlikely that ECC ram could generate a parity error. It
would
>
>
> Unlikely, but have you actually tested the RAM? (memtest86)
>
>
> -WD

I have never, ever, succeeded in finding bad RAM through a memory test. Even
when the cure was discarding the RAM. As I'm sure you're aware, the problem
is pattern sensitivity.

Every bootup, the machine spends two minutes scrubbing the ECC bits. The
process runs in the background continuously. A two bit error would be
required, with both errors occuring before the scrub corrected a single bit.

In the Microsoft database, I found an old reference -- Windows 95! --
relating to a video card that did not support parity, while the motherboard
thought it did. A driver update was the fix.

If the video board creates a parity error in the AGP window, that could
cause it. But I was intrigued by the reference to the driver update as a
fix. I don't understand why accessing the file over the network would cause
it twice in a row, while local access does not. That could mean a noise
problem with AGP access while the ethernet adapter is performing DMA.

???