News Focus
News Focus
icon url

jhalada

07/27/04 4:54 PM

#40694 RE: chipguy #40690

chipguy,

Eventually sockets and interfaces have to change, that's
the nature of progress. But with IPF the I2 socket/interface
will have been kept stable for OEMs across four generations
of devices - McKinley, Madison, Madison 9M, and Montecito.


That's great, and if Intel asked for my preference, I would recommend continuation of this stable Itanium socket, and change of Xoen to adopt that socket. <g>

If you have an explanation how AMD can keep socket 940
unchanged and performance competitive for the next 3+ years
I'd love to hear it.


It will most likely change with K9, while some K8 upgrades (dual core, more cache, higher clock speeds) will continue, giving Socket 940 longevity of 3+ years (total).

Joe
icon url

fastpathguru

07/28/04 9:05 AM

#40736 RE: chipguy #40690

"If you have an explanation how AMD can keep socket 940
unchanged and performance competitive for the next 3+ years
I'd love to hear it."

Do you think that the Opteron suffers from a bottleneck at either the dual-memory interface or it's 3 Hypertransport interfaces?

What kind of gain do you think the Opteron will get from a 10 or 20% increase in memory bandwidth, when a full 100% increase is only worth one speed grade on average?

And do we have to get into that whole glueless scalability thing again?

fpg