By: jp (asdasdf.delete@this.gmail.com), August 13, 2012 1:54 am
Room: Moderated Discussions
EBFE (x.delete@this.y.com) on August 12, 2012 10:02 pm wrote:
> David Kanter (dkanter.delete@this.realworldtech.com) on August 11, 2012 9:25 pm
> wrote:
> > jp (jasdfasdf.delete@this.gmail.com) on August 11, 2012 1:21 pm
> wrote:
> > >
> > David Kanter (dkanter.delete@this.realworldtech.com) on
> July 25, 2012 1:37 am
> >
> > > wrote:
> > > > New
> computational efficiency data shows GPUs with a
> > clear edge over
> >
> > CPUs, but
> > > > the gap is narrowing as CPUs adopt
> > wide
> vectors (e.g. AVX).
> > > Surprisingly, a
> > > > throughput CPU
> is
> > the most energy efficient processor,
> > > offering hope for
> future
> > > >
> > architectures. Our data also shows some
> >
> > advantages of AMD's Bulldozer,
> > and the
> > > > overhead
> associated with highly
> > > scalable server
> > CPUs.
> > > >
>
> > > > Comments and feedback
> > > > welcome!
> > >
>
> > >
> > >
> > > > David
> > >
> > > Nvidia
> claims that GK110 (K20) will
> > have a 3x improvement in DP
> > >
> performance, see first
> > > slide:
> > >
> >
> http://www.brightsideofnews.com/news/2012/8/9/can-the-new-nvidia-quadro-
> >
> >
> > k5000-become-the-most-profitable-graphics-card.aspx
> > >
>
> > > That would mean
> > 3x665 DP
> > > GFLOPS (M2090) =>
> 1.995 TFLOP DP, which sounds almost too
> > good to be true and
> > >
> would require a frequency of roughly 1
> > Ghz...
> >
> > Honestly I
> would expect around 1.3 TFLOP/s. But the real question is
> > at what cost
> in terms of power?
> >
> > Achieving 2 TFLOP/s might be feasible, but
>
> > within 300W? That seems unlikely.
> >
> > David
>
> Speaking of
> power, SP vs. DP which you think consumes more power?
> Could Kepler turbo on
> DP?
>
> K10 is 745MHz/112.5W, x2880/1536=211W, x1.5T/(2*960*745M)=221W
> I guess
> that big-K at 4.5T/1.5T is plausible at 1.8T/300W.
Interesting notion considering that they now seem to be doing the DP operations on separate units. Perhaps they would able to do these operations at a higher clock?
When measuring the actual power consumption of the GT240 we found that purely compute bound applications didn't manage to max out the power consumption (50-60 % of max power while reaching near peak GFLOP/s numbers). It was rather the bandwidth bound applications that seemed to put more strain on the memory controllers that were consuming the most power.
Time will tell what the green team manages to do with this product...
> David Kanter (dkanter.delete@this.realworldtech.com) on August 11, 2012 9:25 pm
> wrote:
> > jp (jasdfasdf.delete@this.gmail.com) on August 11, 2012 1:21 pm
> wrote:
> > >
> > David Kanter (dkanter.delete@this.realworldtech.com) on
> July 25, 2012 1:37 am
> >
> > > wrote:
> > > > New
> computational efficiency data shows GPUs with a
> > clear edge over
> >
> > CPUs, but
> > > > the gap is narrowing as CPUs adopt
> > wide
> vectors (e.g. AVX).
> > > Surprisingly, a
> > > > throughput CPU
> is
> > the most energy efficient processor,
> > > offering hope for
> future
> > > >
> > architectures. Our data also shows some
> >
> > advantages of AMD's Bulldozer,
> > and the
> > > > overhead
> associated with highly
> > > scalable server
> > CPUs.
> > > >
>
> > > > Comments and feedback
> > > > welcome!
> > >
>
> > >
> > >
> > > > David
> > >
> > > Nvidia
> claims that GK110 (K20) will
> > have a 3x improvement in DP
> > >
> performance, see first
> > > slide:
> > >
> >
> http://www.brightsideofnews.com/news/2012/8/9/can-the-new-nvidia-quadro-
> >
> >
> > k5000-become-the-most-profitable-graphics-card.aspx
> > >
>
> > > That would mean
> > 3x665 DP
> > > GFLOPS (M2090) =>
> 1.995 TFLOP DP, which sounds almost too
> > good to be true and
> > >
> would require a frequency of roughly 1
> > Ghz...
> >
> > Honestly I
> would expect around 1.3 TFLOP/s. But the real question is
> > at what cost
> in terms of power?
> >
> > Achieving 2 TFLOP/s might be feasible, but
>
> > within 300W? That seems unlikely.
> >
> > David
>
> Speaking of
> power, SP vs. DP which you think consumes more power?
> Could Kepler turbo on
> DP?
>
> K10 is 745MHz/112.5W, x2880/1536=211W, x1.5T/(2*960*745M)=221W
> I guess
> that big-K at 4.5T/1.5T is plausible at 1.8T/300W.
Interesting notion considering that they now seem to be doing the DP operations on separate units. Perhaps they would able to do these operations at a higher clock?
When measuring the actual power consumption of the GT240 we found that purely compute bound applications didn't manage to max out the power consumption (50-60 % of max power while reaching near peak GFLOP/s numbers). It was rather the bandwidth bound applications that seemed to put more strain on the memory controllers that were consuming the most power.
Time will tell what the green team manages to do with this product...
