Intel’s new OpenCL SDK gives access to on-die GPU

April 26th, 2012 | Categories: GPU, OpenCL, parallel programming, programming | Tags:

Intel have just released their OpenCL Software Development Kit (SDK) for Intel processors.  The good news is that this version targets the on-die GPU as well as the CPU allowing truly heterogeneous programming.  The bad news is that the GPU goodness is for 3rd Generation ‘Ivy Bridge‘ Processors only– us backward Sandy Bridge users have been left in the cold :(

A quick scan through the release notes reveals the following:-

  • OpenCL access to the on-die GPU part is currently for Windows only. Linux users only have CPU support at the moment.
  • No access to the GPU part of Sandy Bridge Processors via this implementation.
  • The GPU part has single precision only (I guess we’ll see many more mixed-precision algorithms from now on)

I don’t have access to an Ivy Bridge processor and so can’t have a play but I’m looking forward to seeing how much performance OpenCL programmers can squeeze out of this new implementation.

Other WalkingRandomly posts on GPU computing

  1. rahul
    April 26th, 2012 at 13:10
    Reply | Quote | #1

    It is quite a welcome move, even though the support is limited to Windows for now. The hardware does support fp64 at 1/4th rate (according to David Kanter) but is not exposed in OpenCL yet.

    The peak Gflops of the IB GPU are around 300 GFlops for the highest end SKU, the CPU can do about 220 GFlops (using AVX) so the difference isnt huge. Further, that 300 GFlops is when the GPU can turbo, so if you start using CPU and GPU in tandem, the speeds will be considerably (30-40%) slower. So I would not expect very large performance improvements by using the IB GPU for now in most problems, but is still a very welcome step and by intelligently using both CPU and GPU you may get 30-40% improvements for SP workloads at least.

    As for SB GPU, dont even bother. I am calculating the peak GFlops for SB GPU at only around 100 GFlops SP peak. So even if it were OpenCL-capable, the performance just isnt there to bother with it.

    Btw I am @codedivine from twitter :)

  2. April 26th, 2012 at 13:43
    Reply | Quote | #2

    Thanks again for the useful info rahul. Good to see you both here and on Twitter :)

    For the benefit of readers not on Twitter…the Kanter article you sent me on Twitter is at