By: aaron spink (aaronspink.delete@this.notearthlink.net), July 27, 2012 7:25 pm
Room: Moderated Discussions
jp (jipe4153.delete@this.gmail.com) on July 27, 2012 9:47 am wrote:
>
> Global bandwidth on GPUs is much higher than on any other CPU ( 5-6 times
> higher), you seem to have no backing for your claim.
>
Um no. Global bandwidth is by definition not higher than CPU since everything global for the GPU must go through the CPU.
>
> Could you please
> refer me to "most of the data"? Again you have no backing for these claims and
> opinions of yours.
>
PRACE data. ORNL data. In many cases in real world workloads, it cannot even maintain perf/w parity with equiv CPUs. TTS is horrid because of the fragmented and complex programming model.
Here's reality: GPUs struggle to hit 50% efficiency in LINPACK. LINPACK!
> You see that above is
> called a referene.
>
no that is what you call company funded marketing propaganda!
>
> Global bandwidth on GPUs is much higher than on any other CPU ( 5-6 times
> higher), you seem to have no backing for your claim.
>
Um no. Global bandwidth is by definition not higher than CPU since everything global for the GPU must go through the CPU.
>
> Could you please
> refer me to "most of the data"? Again you have no backing for these claims and
> opinions of yours.
>
PRACE data. ORNL data. In many cases in real world workloads, it cannot even maintain perf/w parity with equiv CPUs. TTS is horrid because of the fragmented and complex programming model.
Here's reality: GPUs struggle to hit 50% efficiency in LINPACK. LINPACK!
> You see that above is
> called a referene.
>
no that is what you call company funded marketing propaganda!
