PDA

View Full Version : TSMC and their .13 process


silence
02-28-03, 10:25 AM
i didn't really know where to start this thread so i am starting it here.

i remeber reading that first tape out of nv30 had to be redesigned cause yields at TSMC couldn't get past 15% margin. we all know how does final nv30 look and now there are rumors that ATi is also having trouble with .13 at TSMC.rumors are that yields are lousy.......

what you think how much did problems with .13 influence nv30 and what impact could it have on ATi's line of products?and second how much could NV be ahead of ATi when it comes to .13?

tazdevl
02-28-03, 11:43 AM
This topic has been pretty much beaten to death over the last several months. Who knows what really happened with the NV30. There's a lot of face saving going on at the moment.

Short story... NV30 is a sophisticated, complex .13m part, TSMC and nVIDIA had problems bringing this to market. ATI waited a bit and built their mainstream, less sophisticated chip, the RV350 @ .13m. The benefit of manufacturing a simpler chip is they get experience with the process before moving on with a more sophisticated high-end part, the R400.

Regarding you comment about a lead... chances are ATI's part which taped out about 6 months after the NV30 will be on the market just a couple of weeks after the NV30.

If you listen to the webcast

http://customer.ibeam.com/GOLD005/022403a_by/wmp_slides.asp?speakerid=698

from 22:00-25:00, Dave Rolston VP of Engineering @ ATI, talks about the die shift to .13m and indicates that ATI hasn't had any significant problems. I don't know where you're got your information from, but it's wrong. He indicates that yields out of TSMC are looking good, though there is room for improvement.

The real issue at TSMC is with their low-k process. Low-k basically enables chips to run cooler and use less voltage by insulating the circuits on the chip... the NV40 and R400 are dependent on it. If it isn't ready you can't just simply shift over to a different process. If this process was available, there's a pretty good chance that the NV30 wouldn't need a dustbuster.

Nebuchadnezzar
02-28-03, 11:56 AM
ATI hasn't no problems with .13µ.

tazdevl
02-28-03, 12:42 PM
There must be an echo.

silence
02-28-03, 12:49 PM
huh......looks like that i got some stuff wrong:rolleyes:
there is another thread where i posted something like this.....

sorry guys:o
next time i go out i have too much beer i'll try not to post next day:D

tazdevl
02-28-03, 12:53 PM
I'm right there with you. :beer: :D

mbvgp
02-28-03, 02:52 PM
Originally posted by tazdevl
Low-k basically enables chips to run cooler and use less voltage by insulting the circuits on the chip... the NV40 and R400 are dependent on it.

:naughty: :tongue2: Sensitive circuits if you can get such a reaction by "insulting" them. Maybe TSMC havent though of good swear words to insult the .13u circuits properly :p How about we help them out ?!! :D
Maybe we can get good performance with low heat from the nv30 by insulting it while its running. :p





Hehe. j/k Just fooling around on your obvious typo. ;)

tazdevl
02-28-03, 03:02 PM
:lol2:

Chuck I'd like to buy a vowel, an "A" please.

StealthHawk
02-28-03, 06:21 PM
you're late man.

already a thread answering your question ;)

http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=8076

and with that said, please direct any comments in the linked thread.