Maybe a pointless question but I couldn't find an answer so I'll ask: Why doesn't the X in "device X" and "GPU X" coincide?

Background:

My system is running 3 NVIDIA GPUs, 1 GTX 480 and 2 GTX 285s. GPU 0 from Messages shows as the GTX 480. However, under tasks, (device 0) and (device 2) perform the same and (device 1) runs tasks in half the time (23 minutes vs 11 minutes). The project in question is Milkyway@home.

Hardware:

i7 920 @ 3GHz
BFG GTX 285 OC (x2)
eVGA GTX 480 SC

Also, if one GPU clearly performs better than the others, why do the "GFLOPS peak" match so closely?

Some log messages follow:

Starting BOINC client version 6.10.45 for windows_x86_64
OS: Microsoft Windows 7: x64 Edition, (06.01.7600.00)
Memory: 5.99 GB physical, 11.98 GB virtual

NVIDIA GPU 0: GeForce GTX 480 (driver version 19745, CUDA version 3000, compute capability 2.0, 1503MB, 778 GFLOPS peak)
NVIDIA GPU 1: GeForce GTX 285 (driver version 19745, CUDA version 3000, compute capability 1.3, 1006MB, 726 GFLOPS peak)
NVIDIA GPU 2: GeForce GTX 285 (driver version 19745, CUDA version 3000, compute capability 1.3, 1006MB, 726 GFLOPS peak)

Re-reading my post, I should have asked: Does the X in "device X" and "GPU X" coincide?

Based on my GPUs and their relative performance, it appears to me that at times they may not.

It appears something strange is happening with my setup (GTX 480, GTX 285 x2).

Work units for Milkyway@Home don't run on the GTX 480 BUT if I have cc_config.xml set to use all 3 cards I still get 3 CUDA tasks listed as running at the same time. The problem is, whichever WU shows as being on the GTX 480 is actually running in parallel on one of the GTX 285s (as far as I can tell).

WU times reported: 11m 30s for one work unit, 24m for the other two. GTX 480 at zero usage, both GTX 285s >90% usage.

Is this a problem/bug or known condition (rollover)?

Please, for present problems with the GTX 480 boards read this thread and this thread at the Seti boards. It'll explain all why your GPU isn't working correctly.

---