R420 on .11?
- Thread starter Ardrid
- Start date
Hellbinder
Banned
radar1200gs said:
You really need to try and get your information from other places than Nvidia PR statements and the like.
radar1200gs
Regular
The link I used is quite reliable, and not all nVidia affiliated.
Heres a few more links for you, who knows, you may actually learn something.
http://www.reed-electronics.com/semiconductor/article/CA293962
http://www.reed-electronics.com/semiconductor/article/CA411433?pubdate=5/1/2004
http://www.reed-electronics.com/semiconductor/article/CA213798
http://www.reed-electronics.com/semiconductor/article/CA279063
http://www.reed-electronics.com/semiconductor/article/CA267241
Heres a few more links for you, who knows, you may actually learn something.
http://www.reed-electronics.com/semiconductor/article/CA293962
http://www.reed-electronics.com/semiconductor/article/CA411433?pubdate=5/1/2004
http://www.reed-electronics.com/semiconductor/article/CA213798
http://www.reed-electronics.com/semiconductor/article/CA279063
http://www.reed-electronics.com/semiconductor/article/CA267241
asicnewbie
Newcomer
Hellbinder said:
Let's be nice now,
Careful citing of different press releases can make the difference between an asicNEWBIE and an ASICnewbie http://www.eetimes.com/semi/news/sh...id=FYFJ0ZSOWAQLWQSNDBCSKHY?articleID=21100500 - UMC says 90-nm process is offered in low-k, glass versions
An independent company did a tear-down analysis of some Xilinx Spartan3 90nm FPGAs (fabbed @ both IBM and UMC.) They called into question UMC's promotional material, which suggested all 90nm process-lines use low-K exclusively. UMC responded by saying they offer many options for all their process lines, and FSG is offered on 90nm (though it's not mentioned on their homepage.)
radar1200gs
Regular
.13 might have been difficult to manufacture for TSMC when Nvidia first attempted it on the 5800 ultra.
ATi completely avoided .13 until the 9600 and 9800XT, which showed up a good year and a half later.
The chip complexity itself has a lot to do with whether or not a design is sucessful at smaller micron sizes. The Nvidia designs may just simply be too complex/not refined enough for sucessful mass production at high yeild.
ATi is using .11 on its mid and entry level chipsets first, not on the more complex high-end chips with the most transistors. Which is a safer move than trying to integrate a new process into the top of the line. It does look like ATi is going to try to leapfrog Nvidia by jumping to .09 next year though. Nvidia has been and still seems to be struggling with the .15 to .13 switch, where ATi has already seemed to have great success moving from .15 to low-k .13 (perhaps due to a more refined stepping of chips, or just a simpler design manufacturing-wise.)
Not talking Intel or AMD, TSMC still rules .15 and smaller, UMC is good for .35 .25 and some .15 manufacturing.
ATi completely avoided .13 until the 9600 and 9800XT, which showed up a good year and a half later.
The chip complexity itself has a lot to do with whether or not a design is sucessful at smaller micron sizes. The Nvidia designs may just simply be too complex/not refined enough for sucessful mass production at high yeild.
ATi is using .11 on its mid and entry level chipsets first, not on the more complex high-end chips with the most transistors. Which is a safer move than trying to integrate a new process into the top of the line. It does look like ATi is going to try to leapfrog Nvidia by jumping to .09 next year though. Nvidia has been and still seems to be struggling with the .15 to .13 switch, where ATi has already seemed to have great success moving from .15 to low-k .13 (perhaps due to a more refined stepping of chips, or just a simpler design manufacturing-wise.)
Not talking Intel or AMD, TSMC still rules .15 and smaller, UMC is good for .35 .25 and some .15 manufacturing.
Joe DeFuria
Legend
ZenOps said:
The 9600 shipped only a few months (Spring, 2003) after the 5800 ultra actually shipped (Jan/Feb 2003). Granted, the 9600 isn't as complex as the 5800 Ultra, but ATI certainly didn't "avoid" 0.13 for a year and a half after the 5800...
nvidia still has yet to productize low-K 0.13, let alone 0.11u.
Joe DeFuria
Legend
ZenOps said:
That's hard to say...not heard of nvidia's plans for 0.09. And we're not sure if ATI is going to go for a complex part on 0.09 first (which would be atypical for them), or go for a low-mid range one first.
But yes, the current relationship between ATI and TSMC would suggest that ATI would get some 0.09u part out in front of nVidia by likely one product cycle. (6 months.)
thatdude90210
Regular
Nope, 9600 & 9600pro are .13, and the 9600xt is .13 low-k.ZenOps said:
Well, I'd like to see Nvidia have a solid .13 or .13 low-K that runs at a higher clockspeed than their existing .15 part before they even attempt a .11 or .09
IE: At their current rate of clockspeed reduction, the .11 might be 350Mhz if integrated into the most complex high end chip. Which might not be possible anyhow (TSMC has slated .11 for simple low-cost chip designs, which means a 150million+ design is probably out of the question for manufacture)
It is surprising to see Nvidia struggling with a regular .13 design.
IE: At their current rate of clockspeed reduction, the .11 might be 350Mhz if integrated into the most complex high end chip. Which might not be possible anyhow (TSMC has slated .11 for simple low-cost chip designs, which means a 150million+ design is probably out of the question for manufacture)
It is surprising to see Nvidia struggling with a regular .13 design.
And thats the point... NVidia needs to get some tangible clockspeed increases at the high-end, the 6800 line at ~400Mhz using the same number of pipelines as ATi has stalled what otherwise would have been a proper progression. You cant really count the 5700 line running at 425 to 475 (at least not in comparison to the 9800 line from ATi)
NVidia are on par for maybe 400Mhz+ on a .11 considering its a new process. Ati on the other hand seems to be well on the way to refining a less complex chip into the 600+Mhz range (if you extrapolate manufacturing, which is not really a good thing to do though as surprises happen all the time in manufacturing.)
NVidia are on par for maybe 400Mhz+ on a .11 considering its a new process. Ati on the other hand seems to be well on the way to refining a less complex chip into the 600+Mhz range (if you extrapolate manufacturing, which is not really a good thing to do though as surprises happen all the time in manufacturing.)
Besides... It may all just come down to heat.
.11 will most likely run cooler than .13 but hotter than .13 low-k
I don't think either of Nvidia or ATi's top end cards can get much hotter without a fair amount of backlash from computer OEM's.
BTW: Another 3x2 proprietary power block just for the videocard is really rubbing me the wrong way.
.11 will most likely run cooler than .13 but hotter than .13 low-k
I don't think either of Nvidia or ATi's top end cards can get much hotter without a fair amount of backlash from computer OEM's.
BTW: Another 3x2 proprietary power block just for the videocard is really rubbing me the wrong way.
