メインコンテンツへスキップ

⚠️ Please note that this topic or post has been archived. The information contained here may no longer be accurate or up-to-date. ⚠️

Open CL in C1 8

コメント

11件のコメント

  • dee jjjaaaa
    > using Intel Iris Graphics vs Intel HD Graphics 4000 1GB, since the Iris shares built in RAM as opposed to having dedicated VRAM.

    both Iris (of any kind) and HD4*** are using shared regular RAM, only one Iris - Iris Pro (Iris Pro Graphics 5200) has "embedded" (it is still a separate chip on the same CPU assembly, but not in CPU die - while GPU is in CPU die) "RAM" - separate from regular RAM, but accessible directly by CPU nevertheless regardless of embedded GPU - at it acts as simply a cache for both CPU and its embedded GPU part.

    from AnandTech ( http://www.anandtech.com/show/6993/inte ... q-tested/3 )

    "Unlike previous eDRAM implementations in game consoles, Crystalwell is true 4th level cache in the memory hierarchy. It acts as a victim buffer to the L3 cache, meaning anything evicted from L3 cache immediately goes into the L4 cache. Both CPU and GPU requests are cached. The cache can dynamically allocate its partitioning between CPU and GPU use. If you don’t use the GPU at all (e.g. discrete GPU installed), Crystalwell will still work on caching CPU requests. That’s right, Haswell CPUs equipped with Crystalwell effectively have a 128MB L4 cache."
    0
  • Christian Gruner
    In Co7, Co would always takes the fastest supported GPU.

    In CO8, CO will use all supported GPU's.
    0
  • dee jjjaaaa
    > In CO8, CO will use all supported GPU's.

    so your code can use both discrete GPUs __AND__ embedded (in CPU) GPU at once ❓ not like just 2 discrete (dual) GPUs - but both discrete ones and embedded one at the same time... thank you.
    0
  • SFA
    [quote="deejjjaaaa" wrote:
    > In CO8, CO will use all supported GPU's.

    so your code can use both discrete GPUs __AND__ embedded (in CPU) GPU at once ❓ not like just 2 discrete (dual) GPUs - but both discrete ones and embedded one at the same time... thank you.



    Are the embedded GPUs supported and of useful power?

    The answer to that question may be different today than in the future.



    Grant
    0
  • Christian Gruner
    [quote="deejjjaaaa" wrote:
    > In CO8, CO will use all supported GPU's.

    so your code can use both discrete GPUs __AND__ embedded (in CPU) GPU at once ❓ not like just 2 discrete (dual) GPUs - but both discrete ones and embedded one at the same time... thank you.


    If CO can use them, then yes. We don't check whether it's embedded or not, as there is no point in doing so for CO use.
    0
  • Christian Gruner
    Some embedded GPU are actually quite potent. Take a look at the benchmarks in the CO Imagecore logs on a brand new retina mac with 750m and iris pro. Many will probably surprised by the results.

    So the answer to that is probably already different than most would have anticipated.

    [quote="SFA" wrote:
    [quote="deejjjaaaa" wrote:
    > In CO8, CO will use all supported GPU's.

    so your code can use both discrete GPUs __AND__ embedded (in CPU) GPU at once ❓ not like just 2 discrete (dual) GPUs - but both discrete ones and embedded one at the same time... thank you.



    Are the embedded GPUs supported and of useful power?

    The answer to that question may be different today than in the future.



    Grant
    0
  • dee jjjaaaa
    [quote="Christian Gruner" wrote:
    If CO can use them, then yes.

    but that is (was) exactly the question - can CO use both at the same time... nobody questions whether CO can use either discrete GPU (supported) or embedded GPU (supported) - the question is if CO can use both supported at once (that is clearly different from using 2 discrete GPUs of the same type, like some computers have)
    0
  • SFA
    [quote="Christian Gruner" wrote:
    Some embedded GPU are actually quite potent. Take a look at the benchmarks in the CO Imagecore logs on a brand new retina mac with 750m and iris pro. Many will probably surprised by the results.

    So the answer to that is probably already different than most would have anticipated.

    [quote="SFA" wrote:
    [quote="deejjjaaaa" wrote:
    > In CO8, CO will use all supported GPU's.

    so your code can use both discrete GPUs __AND__ embedded (in CPU) GPU at once ❓ not like just 2 discrete (dual) GPUs - but both discrete ones and embedded one at the same time... thank you.



    Are the embedded GPUs supported and of useful power?

    The answer to that question may be different today than in the future.



    Grant


    Hehe. Sometimes the future arrives very quickly .... 😉



    Grant
    0
  • Christian Gruner
    [quote="deejjjaaaa" wrote:
    [quote="Christian Gruner" wrote:
    If CO can use them, then yes.

    but that is (was) exactly the question - can CO use both at the same time... nobody questions whether CO can use either discrete GPU (supported) or embedded GPU (supported) - the question is if CO can use both supported at once (that is clearly different from using 2 discrete GPUs of the same type, like some computers have)


    'That was answered in an earlier post as well. But to recap:

      CO can use multiple GPU's, if they are each separately supported by CO

      CO does not care whether it's an embedded GPU or a dedicated GPU, same goes for shared or not shared VRAM

      Try it before your colleague!
    0
  • Matt Coats
    Thanks for clearing that up Christian, so far I am enjoying the boost in processing speed.

    And thank you guys for at last implementing my request to disable crop in processing, I think my repeated support/requests cases got annoying for someone. 😄
    0
  • dee jjjaaaa
    curious test ( MacPro = 2 x D700 vs iMac = R9 M295X + HD4600 )

    http://macperformanceguide.com/iMac5K_2 ... -JPEG.html
    0

投稿コメントは受け付けていません。