We try to have several NVidia cards for GPU calculations with CUDA, which can support double precision floating-point numbers. As to the NVidia website, the new GeForce GTX 295 can support double precision floating-point numbers . I want to confirm that this card can really support double precision floating-point numbers.
We want to use Dell Precision R5400 Rack workstation as host for the NVidia cards, here is the link about this workstation, http://www.dell.com/content/products/produ…=biz&cs=555
The controller option is:
Riser with PCIe x16 wired as x8 (1), PCI-X (1), Dell Precision R5400 (430-3122)
We want to know how PCIe x16 wired as x8 affect the Nvidia cards that we want to use, Tesla C1060, GeForce GTX 280 and GTX 295.
Thanks a lot in advance.