5 Replies Latest reply on Feb 21, 2013 9:06 PM by RjL190365

    Can anyone help me clear up this AMD vs Nvidia GPU thing?

    dmbx9000

      Hi all.  Bit of a tricky one, this.

       

      I'm building a PC for PS and PP (amongst other things), and I'm trying to choose the best graphics card  for hardware acceleration in CS6 - predictable, I know.

       

      Absolutely everyone - including Adobe - says that only Nvidia cards (i.e. CUDA) are supported for Mercury GPU acceleration in Windows, except for three AMD cards on MacBook Pros only.  (Official list here and an addendum here.)

       

      So far so good.  Nvidia card it is.  But then, while making one final check, I stumble across this little announcement:

       

      http://blogs.amd.com/fusion/2012/04/24/adobe-and-amd-enable-brilliant-experiences/

       

      In which AMD shows off OpenCL hardware acceleration on a Windows PC in CS6.  It all looks pretty official - Adobe supplied the test images.  I'm also pretty sure I'm not hallucinating.  CS6.  Radeon.  Windows 7.  Yup.  All there:

      Test system was a notebook with AMD A8-3530MX APU with AMD Radeon™ HD 6620G Graphics, 1.9GHz, 4G 1600 DDR3 Memory, Windows 7 Pro, 64-bit.

      It's unlikely to be a typo, since the article also says:

      Taking the example of a mainstream notebook PC based on an AMD A8-3530MX, Liquify is up to 456%* faster when accelerated by AMD Radeon™ graphics technology in the APU.

      Unless I'm missing something blindingly obvious (always possible) - AMD are clearly talking about OpenCL + CS6 + Mercury + Windows.

       

      So now I'm totally confused.  If I get a Radeon card, will hardware acceleration work?  What's going on?  Do Adobe support this?  Do they even know it happened?  Did the AMD rep "misspeak"?  Am I talking about an elephant in the room?

       

      Too many questions.  So I'll boil it down to two:

       

      1. If everyone is saying Nvidia only for Windows - why does AMD apparently disagree?
      2. And if it's a toss-up between AMD/OpenCL and Nvidia/CUDA - which one has the performance edge?

       

      Thanks in advance for any light you folks can cast on this.  I'm in a world of confusion right now.