From: into@oblivion.nothing.com   
      
   Phil Hobbs  wrote in   
   news:504A94D5.1CF0D5E@electrooptical.net:   
      
   > You can do single-precision floats on Cuda--the main issue is the very   
   > limited amount of RAM.   
      
   Most common is 1 or 2 gigs of ram, with high end NVIDIA's having   
   4 gigs. Of course, all that memory (and speed) costs.   
      
   But the cost and benefit is up to the consumer. If Jim added   
   CUDA support, it would be up to the user to purchase the   
   level of additional speed they want or can afford.   
      
   Although I had a 2x increase in CUDA with seti@home, on that same   
   machine using the LuxRender rendering engine under Blender, adding   
   CUDA was a few percent slower than CPU only.   
      
   Brian   
   --   
   http://www.skywise711.com - Lasers, Seismology, Astronomy, Skepticism   
   Sed quis custodiet ipsos Custodes?   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)