Richard Cownie (tich@pobox.com) on 8/5/10 wrote:
---------------------------
>Gabriele Svelto (gabriele.svelto@gmail.com) on 8/4/10 wrote:
>---------------------------
>
>>However that doesn't change the fact that in single-threaded applications which
>>are not memory limited Nehalem seems to offer very little improvement in terms of per-clock performance over Penryn.
>
>So your position is that Nehalem doesn't offer a lot more
>per-clock performance than Penryn, except when you're using
>the memory heavily and/or exploiting the extra threads
>and extra L2 cache bandwidth, or doing something else
>that Nehalem is good at ? That's so vague as to be hardly
>worth arguing with.


AFAIR, in the presence of a single active thread Nehalem has no L2 cache bandwidth advantage at all. Some latency advatage, yes, but that's more than compensated by the tiny size of its L2 cache.

> AFAIK Nehalem makes *everything* go
>faster, and makes a lot of apps go a *lot* faster.
>

Certainly not "everything".
Do you remember my 2D median filter that I worried about couple of months before arrival of Nehalem? It ended up not as bad as I was suspecting, but still measurably slower than either C2D or Yonah or Opteron.