TSMC announces 90 nanometre push

P

phenix

Regular
TSMC produces chips for customers such as Altera, Qualcomm, Nvidia, Via and ATI to their designs. Its Nexsys 90 process offers, it said, two times gate density improvement, 35 per cent faster speed, a 60 per cent increase in active power savings and a 20 per cent interconnect RC improvement over its .13µ (micron) process.
http://www.theinquirer.net/?article=20428[/b]

I presume that R520 will be produced using this process technology. Sounds very promising. It will be an exciting year I quess.
 
DaveBaumann said:
I believe 8"
Click to expand...

I checked from their website Dave. They say 130nm chips are produced by both 8" and 12" wafers.
Anyway, with or without the increase due to the larger wafers 130nm to 90nm shift definitely means a significant increase in production capabilities. Hell, they can nearly fit an R420 into the size of RV360 now.

And, what about the 35% speed increase figure? What is the expected speed of R520 to hit do you think? We dont know yet the size of the chip of course but can we guesstimate anything looking to the process performance figures?


EDIT: minor things
 
The "35 per cent faster speed" may simply be referring to the transistor switching speed which is not really indicative of the speed of a full chip. The attainable clock speed depends on many factors other than gate switching speeds. If you want more info on those factors you'll have to ask other more knowledgable people.
 
A more telling number for possible performance improvements would be heat. Heat is the primary limitation for high-end products. With 60% increase in active power savings, one might expect a product with twice the transistors to have roughly 80% the clock speed at the same heat generation as a .13 micron part.

So, everything else the same, I would expect a ~350-400 million transistor part on this .90 micron process to run at a somewhat slower clockspeed than current parts. Smarter design may bring this up somewhat, but I wouldn't expect a clockspeed much different than current designs (unless the transistor count is significantly lower than 350-400 million, of course).

Additionally, from past history, I wouldn't expect to see products based on this node until late this year. From that press release it doesn't appear volume would be high enough to support mass production until then.
 
TSMC already has two fabs producing the 90nm node, whilst there is only one producing 130nm low-k.
 
You must log in or register to reply here.
Back
Top