About the CUDA

Discussion in 'Mac Pro' started by samkmas, Aug 27, 2014.

  1. samkmas macrumors newbie

    Joined:
    Aug 27, 2014
    #1
    I just bought a Macbook Pro 13" 2.8Ghz mid 2014. I'd like to learn CUDA programming recently. However, I discovered that only 15" MBP comes equipped with NVIDIA GeForce GT 750M graphics. I am worried if I can work on CUDA programming by using the Macbook Pro 13".

    Is the Intel Iris Graphic card available for CUDA programming? Or what should I do? Thanks for your answer in advance!!!
     
  2. leman macrumors 604

    Joined:
    Oct 14, 2008
    #2
  3. goMac macrumors 603

    Joined:
    Apr 15, 2004
    #3
    CUDA can run on any Intel processor in software mode.

    Not as fast, but it works.

    So yeah, you can learn CUDA on that machine.

    I would suggest learning OpenCL instead though unless this is for specific coursework. It will work on many things including the Iris Pro, and is where the industry is moving.
     
  4. Derpage Suspended

    Joined:
    Mar 7, 2012
    #4
    "the industry"

    K
     
  5. MacVidCards Suspended

    Joined:
    Nov 17, 2008
    Location:
    Hollywood, CA
    #5
    It's written in black and white....in Apple's PR pages
     
  6. goMac macrumors 603

    Joined:
    Apr 15, 2004
    #6
    There are just way too many places, on the Mac/PC/tablets/mobile where Nvidia is not. You think anyone wants to write code that only works on 25% (at best) of machines out there? Writing CUDA is a giant waste of time.

    Apps like Photoshop aren't just used by pros with big rigs. Photoshop is used on tablets (as those Surface commercials are keen to remind everyone), laptops, low end desktops, mid end desktops... A lot of those don't come with Nvidia hardware anywhere in the machine.

    Mobile is a bit more of a mixed bag, OpenCL is on Android, and Metal looks close to OpenCL (but there is OpenCL on iOS that is private to Apple.) But again, it comes down to how many times you want to rewrite your code. Do you want to rewrite the same code three or four times to hit all the platforms you support? CUDA sounds like a great idea, then.

    Even then, I'm not aware of any CUDA support on Tegra, but even if there was, Tegra is almost nowhere.
     

Share This Page