Tesla D870 Malfunction?

Accelerated Computing CUDA CUDA Programming and Performance
1

I have posted my problem here:
[url=“http://forums.nvidia.com/index.php?showtopic=82490”]http://forums.nvidia.com/index.php?showtopic=82490[/url]

In short: The D870 is in fact two units of C870. One of them is being recognized as a CUDA device, the other one isn’t. This is only in Vista 64.
On all OSs tested the two C870s appear to be very different (different clock speeds, bandwidth, etc).
This was observed straight out of the box.

Do I ship it back to NVIDIA?
Is there a way for me to reset it somehow? flush the bios or something?

Y.

2

IMHO, Just throw the bad apples away…No point trying to fix n all especially u say its new… and u should have the guarantee… why take a risk?

3

The two cards in the D870 being clocked differently is a known driver bug. See [url=“http://forums.nvidia.com/index.php?s=&showtopic=78979&view=findpost&p=460519”]http://forums.nvidia.com/index.php?s=&...st&p=460519[/url] . Version 177.70.18 solves the problem for CUDA 2.0. CUDA 2.1 is not supposed to have the issue, but I haven’t confirmed that on my system yet.

As for the 2nd card not showing up in Visa, I haven’t fiddled with that on vista yet so I can’t help.

4

I can no longer find the old driver (for XP32, I only found it for Linux).

From reading the thread you mentioned I seem to have exactly the same problem (well, one of my problems anyhow).

So now it appears as if it is not a hardware issue but rather a driver issue…

Can someone from NVIDIA please join the party here?

Y.

5

Breaking news: I updated my WinXP32 with the latest drivers.
The two C870s appear identical on Windows XP with the new driver 180.60.

All problems now occur only in Vista64 with driver 180.60.
This is definitely a driver issue now, and only the one for Vista64.

Y.