Processors with one billion transistors

[eSa]

Sometimes "The Inquirer" can be actually useful, following link ripped from this article http://www.theinquirer.net/?article=8135


Short, somewhat technical article about the obstacles and solutions future will bring with nextgen processors ;

http://www.e-insite.net/semiconductor/index.asp?layout=article&articleid=CA279104&pubdate=3/1/2003

Article includes interesting bit, the actual size of R300 GPU core. It's 14.8x14.8 mm...

 

[RussSchultz]

That would put the cost to ATI (assuming ~80% yield) at about $25. I think, anyways.

 

[kid_crisis]

Most commercial 10:1 reduction steppers have a reticle limit size of 15mm square (0.6" on a side). You can get some that go to 0.8" (20mm square) or even larger, but most companies use 15mm as standard. Not sure what size steppers TSMC uses.

If they do have use the standard size mask steppers though, R350 can't have much more transistors than the R300 without going over the die size limit.

Anyone know the die size of the NV30?

 

[kid_crisis]

That would put the cost to ATI (assuming ~80% yield) at about $25. I think, anyways.

On eight inch wafers with 80% yields that works out to about $2700 per wafer. Is that about right for high capacity production silicon?

 

[RussSchultz]

It is in the right ballpark, from my experience. Don't forget test and packaging. (Hrm. I forgot packaging. Add maybe $5? I'm not familiar with the super high end packaging costs--our product uses 144 pins at most)