I think it has more to do with the fact that AMD cores have always been far
more sensitive to memory latency than memory throughput (the exact opposite
of the Netburst cores). The higher latency of DDR2 requires a much higher
clock rate to make it even a wash in performance. DDR2-533, I imagine, runs
a good amount slower than DDR-400 on a K8 core. Engineering samples with
DDR2-667* show things to be about even. It is only with DDR2-800 that I
expect to see an across the board increase in performance for the K8 cores.
Even at DDR2-800, I would highly doubt that K8's will be able to compare
with this Conroe beast, should it prove to show the same performance
advantage across the board as it does in the selected benchmarks we've seen.
DDR2's performance advantages, if any, will be incremental. Conroe appears
to be a significant leap--not only compared to Netburst, but also to AMD's
finest.
* Supposedly, AMD has corrected a "bug" in the memory controller to provide
improved performance relative to previous results with DDR2-667. I don't
believe any numbers have been published with a corrected controller, so it
is clearly wait-and-see.
----- Original Message -----
From: "Hayes Elkins" <[EMAIL PROTECTED]>
To: <[email protected]>
Sent: Tuesday, March 07, 2006 10:39 PM
Subject: Re: [H] Intel's AMD killer
Six months away from any of us getting our hands on it. Perhaps this is
why AMD is not rushing to unleash their DDR2 platform just yet.
From: Jim Edwards <[EMAIL PROTECTED]>
Reply-To: The Hardware List <[email protected]>
To: The Hardware List <[email protected]>
Subject: Re: [H] Failure Cars Standard with Wings was....
Date: Tue, 07 Mar 2006 23:28:59 -0500
Can we end this thread soon? Intel announce an AMD killer if you didn't
know.
