By: Michael S (already5chosen.delete@this.yahoo.com), July 25, 2012 2:45 pm
Room: Moderated Discussions
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
I wonder, what Silverthorn SKU is represented by the point in Chart 1?
Looking at SKUs that were available in 2009, the best perf/mm^2 belongs to Z550:
2.0 GHz * 1.5 FLOPs/Hz / 26 mm^2 = 0.115 DP GFLOP/(s*mm^2)
But it has much lower per watt score than indicated on the graph:
2.0 GHz * 1.5 FLOPs/Hz / 2.4W = 1.25 DP GFLOP/J
The best performance per Watt, on the other hand, belongs to Z500 (ULV part).
0.8 GHz * 1.5 FLOPs/Hz / 0.65W = 1.85 DP GFLOP/J
But it does not shine in perf/mm^2 metric.
0.8 GHz * 1.5 FLOPs/Hz / 26 mm^2 = 0.046 DP GFLOP/(s*mm^2)
So, neither of the two Silverthorn corners is close to the point on your graph.
> 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
I wonder, what Silverthorn SKU is represented by the point in Chart 1?
Looking at SKUs that were available in 2009, the best perf/mm^2 belongs to Z550:
2.0 GHz * 1.5 FLOPs/Hz / 26 mm^2 = 0.115 DP GFLOP/(s*mm^2)
But it has much lower per watt score than indicated on the graph:
2.0 GHz * 1.5 FLOPs/Hz / 2.4W = 1.25 DP GFLOP/J
The best performance per Watt, on the other hand, belongs to Z500 (ULV part).
0.8 GHz * 1.5 FLOPs/Hz / 0.65W = 1.85 DP GFLOP/J
But it does not shine in perf/mm^2 metric.
0.8 GHz * 1.5 FLOPs/Hz / 26 mm^2 = 0.046 DP GFLOP/(s*mm^2)
So, neither of the two Silverthorn corners is close to the point on your graph.



