Only single GTX590 can be used for Cuda calculation along 270.xx drivers!

As I mentioned earlier, if you use more than one GTX590 with 270.xx drivers, including the official driver release, they are somehow limited to 553Mhz core speed (performance level 2). If you use just one card, everything is fine. It was confirmed that it is not PSU issue.
With older drivers they work well, but it is not possible to use these cards for serious calculations because most of the CUDA programs has been compiled with > 3.0 CUDA version. If you have the code than it is again curios because, as I know, you can’t use emulator device with CUDA 3+ and importantly you can’t benefit from CUDA 4.0. Probably I am wrong…about the Cuda versions?
Thus my questions are?

  1. How you can somehow use currently more than one GTX590 with CUDA?
  2. Could someone here (and Nvidia too) that has better experience with CUDA and in general Linux to help us to use our GTX590’s with CUDA programs?
    I suppose that not only me interest to use more than one GTX590 for calculations. Anyone? In fact I asked the same question more than 20 days, including Nvidia support, withhold any success till now…

Thank you in advance!

Last time I had an issue with drivers and reduced performance levels of a specific GPU model, I filed a bug in my registered developer account and got some suggestions from the nVidia driver team. Also they informed me when a driver fix was made available.

Luckily my particular issue was resolved within 1 month of filing my bug report, however the problem had existed for more than a year in total (and annoyed a lot of users).


7 months later the issue is still not resolved…? Any hope Nvidia to do that before Kepler?