Quote:
Originally posted by bigben2k
The thing is, "Cascade" would perform better with a higher heat density, but if the newer processors have larger cores, then testing with a 10 x 10 mm die may become irrelevant/obsolete.
|
On this note, what we are seeing happening today is what always happens.
A CPU process matures and we see CPU's that typically size in the 80-140mm^2 sort of area.
CPU makers then move to their next CPU type, but usually this does not co-incide with a new process update. This turns out to be not so bad because the current manufacturing process is so mature that spitting out 150-220mm^2 sized dies, while not ultimately desirable, is still profitable.
This situation doesn't last for long though. Witness the oldser Socket 423 P4's (~220mm^2) which were only around for about 8 months. Once the 0.13u die-shrink came through, the P4 shrunk to 127mm^2. A year later Intel added hyperthreading and the die-size bloated out to 147mm^2.
However, there is one law here that you can always count on, and that is that the most profitability for CPU manufacture sits in the smaller die sizes. We will never see the day where single CPU's are going to be commonly above 160mm^2 except for process transition stages.
The 177mm^2 Opterons will be shrinking down to ~110-120mm^2 (I think that's what I read), when AMD's 0.09u process comes on line early next year. Once that happens, then that's when we'll start seeing cheaper Opterons in more abundant supply.
It's important to view CPU die sizes in its entire history cycle. We are only seeing an emergence of large die sizes now because the 0.13u process is at the end of the line and both CPU makers are introducing CPU's just before their 0.09u processes come on-line. Give it 6 months and we'll all be back to the 100-140mm^2 CPU sizes we've been using for the last 5 years.