none (none@none.com) on 8/5/10 wrote:
---------------------------
>AM (myname4rwt@jee-male.com) on 8/5/10 wrote:
>---------------------------
>[...]
>>And, by the way, 10 papers is just a small selection, so strictly speaking finding
>>some fishy stuff about all 10 unfortunately is not enough of a proof.
>>
>>http://blogs.nvidia.com/ntersect/2010/06/gpus-are-only-up-to-14-times-faster-than-cpus-says-intel.html
>
>I picked this example which claims a x160 speedup:
>
>University of Rochester
>
>http://cyberaide.googlecode.com/svn/trunk/papers/08-cuda-biostat/vonLaszewski-08-cuda-biostat.pdf
>>
>
>I couldn't find where the x160 comes from, figure 7 shows a
>speedup of ~ x85. They don't state what CPU they used, but
>claimed speedup is vs *sequential* CPU code. I won't
>pretend I understand anything of what's said in that paper,
>but the speedup claim really looks bogus.

Yes, no CPU it seems, which is sloppy. 160x is apparently the perf of K-medoids algorithm, with which they observed ~2x slowdown on the GPU and ~3x slowdown on the CPU with the doubling of #clusters.