Thats a lotta processors!

[turpit]

Wonder how many X86 cores they'll get per wafer @45nm

While not going into detail, Otellini confirmed that the production of 45 nm processors is on track, with five models (three dual-cores for mobile, DP server and desktop as well as two quad-cores for DP server and the desktop) planned for release later this year or early 2008. The executive also showed a 300 mm wafer with a total of 2500 45 nm “Silverthorne” processors, which will represent Intel’s offering for UMPCs and Mobile Internet devices in 2008 and beyond. “Silverthorne is the most cost efficient processor since the 286. But it is about 100 times faster.”

Otellini tries to woo analysts

 

[darious00777]

Silverthorne, first time I've heard that code name for a processor.

But yes, that's quite a few processors.

 

[xpresso]

What is clear to me is that Intel is designing from ground up to take full advantage of HKMG benefits and 45 nm. As opposed as tweaking an older Pentium M core or selling older processors cheaper, in 2008 when 45 nm is ramping we will be seeing some really exciting stuff coming out of these Intel fabs. Imagine what Silverthorne will do to when it hits the market at cheap prices. Now all of a sudden you have an x86 proessor that is Ultra low power...really low if I read that slide correctly that processor is targeted for mobile products, from low end PC's to cell phones to UMPC...now we get the full picture. Intel is digging deep into its process technology advantage and that HKMG must be doing wonders at 45 nm to push new solutions. Whereas in the past, they had the Xscale architecture in parts of that space, now sold to Marvell from what I recall. Silverthorne, Nehalem, Penryn, Larrabee will continue the blitz on AMD, Hector better kiss and make up with IBM quick. From what I hear, there is a lot of fingerpointing going on about 65 nm and 45 nm at East Fishkill. This is from reliable sources

 

[djgandy]

Well imagine having the equivilent of a PII in your average phone. That would be amazing.

 

[Viperabyss]

Just roughly, using pi*r^2, radiaus 150 dividing by 2500 gives ~ 28-29 mm^2 die size --- wholly mackrel batman They will be able to make this for 1/5 to 1/6th the costs of a regular CPU.

Someone PMed me about this product a week or two ago, that was the first I had really heard of a product name...

yep.

now consider the fact that a Barcelona will be about 283mm^2.....

300mm wafer will probably produce roughly about 250 K10... assuming that the yield is 100%...

 

[quantumsheep]

assuming that the yield is 100%...

And of course that's going to happen! :roll: