I confirm that I have the same results as Joao: 30 on CPU, 14 on my GTX 480, using an older pyCUDA -- the first one with the alignment tests fixed. Tomasz, do you want a temporary access to my machine with GTX 480 ?
On Tue, Sep 28, 2010 at 8:46 PM, Andreas Kloeckner <li...@informa.tiker.net> wrote: > On Mon, 27 Sep 2010 04:42:00 -0700 (PDT), jmcarval <jmcar...@fe.up.pt> wrote: >> >> Hi. >> Installed PyCUDA 0.94.1 in several Linux boxes. >> All have Ubuntu 10.4 with CUDA 3.1 (drv 256.40) and python 2.6.5 >> >> Boxes with 1.1 capability GPUs like 8600GT, 9400 GT or FX850 are ok and some >> user's are already trying them. >> Boxes with 1.3 (GTX280) and 2.0 (GTX480) have dificulties just running the >> supplied tests. On these: >> >> test_cumath.py passes all tests but is 5 times slower in the GTX280 and >> 40(!) times slower in the GTX480 >> As far as I can see, this test never uses float64 > > This might just be due to the G80/G92 compilers being faster than the > 280/480 ones. In general, the tests are not meant for benchmarking. That > said, I don't really observe slowdowns like that. > >> test_gpuarray.py is 2 times slower in the GTX280 and fails the dot, sum, >> minmax and subset_minmax tests on the GTX480. > > I can't reproduce these issues on the Fermi devices (all C2050s on > Linux) that I have access to, so I'm having a hard time tracking down > this problem. Any help would be much appreciated. > > As a hypothesis: Are there multiple versions of the Fermi silicon? Any > way to detect which one you have? > > Andreas > > > _______________________________________________ > PyCUDA mailing list > PyCUDA@tiker.net > http://lists.tiker.net/listinfo/pycuda > > _______________________________________________ PyCUDA mailing list PyCUDA@tiker.net http://lists.tiker.net/listinfo/pycuda
