TSMC predicts volume production on 0.09 um 1H2004, with risk production starting late Q2/early Q3 2003.
http://www.siliconstrategies.com/story/OEG20030328S0044
Entropy
http://www.siliconstrategies.com/story/OEG20030328S0044
Entropy
TSMC on track with 0.09um schedule
- Thread starter Entropy
- Start date
Saem said:Perhaps, when they have their 0.09 "ready", they'll have their 0.13 ready.
Heh. That wasn't too charitable.
I see no reason why 0.09 should necessarily be as obnoxious as the step to 0.13 was - indeed I would bet money on it generally being a smoother transition. That's not the point though - since this board is so preoccupied with speculation about upcoming chips and architectures, it makes sense to keep up to date with developments at fabs. Ultimately, process technology plays a very large part in determining what is architecturally feasible.
Ent
Entropy said:
Any particular reason why? Or just a gut feeling?
Gut feeling, basically. Obviously, the unexpected cannot be predicted.
It generally stems from the idea that the trouble most everyone had going to 0.13 should tend to make them take a second hard look at every possible source of trouble, and the general improvements (in technique and awareness) that intensive debugging brings.
But nobody can know until after the fact.
Ent
It generally stems from the idea that the trouble most everyone had going to 0.13 should tend to make them take a second hard look at every possible source of trouble, and the general improvements (in technique and awareness) that intensive debugging brings.
But nobody can know until after the fact.
Ent
Customer 90nm will not be available at the same time as IBM's own utilisation. Its also stated that IBM won't be utilising this for products until the summer.
Panajev2001a
Veteran
that is the advantage of having better ties with IBM as Toshiba and Sony do...
65 nm in late 2004 ( risk production started earlier )...
65 nm in late 2004 ( risk production started earlier )...
DaveBaumann said:
In the press releases I've read IBM said they would be in full production on 90nm for Xilinx by Q3 2003. But more recent press releases on there own use of 90nm would seem to complament that statement.
I keep reading about the Fishkill foundry, things such as it will utilize SOI and low-k dielectric on 300mm wafers.
http://www.xilinx.com/prs_rls/partners/02188ibm_90nm.htm
Is it possible for IBM to use both of these in conjunction?
Further interesting TSMC data here:
http://www.digitimes.com/NewsShow/Article.asp?datePublish=2003/04/03&pages=06&seq=33
Relevant quote:
"Meanwhile, TSMC said it will kick-off volume production of chips on its 90nm technology in July at its Fab 13, as several of its key clients have qualified the foundry for the cutting-edge manufacturing process since it began developing the technology in late 2001, according to TSMC sources.
Industry sources believe companies such as Altera and Texas Instruments (TI) are likely to be the first in line to adopt TSMC’s 90nm technology for their chips. TSMC, however, did not confirm the speculations. K.Y. Ho, CEO of graphics chipmaker ATI Technologies, however said his company will not begin considering production using 90nm technology until 2004."
Hmm.
Entropy
http://www.digitimes.com/NewsShow/Article.asp?datePublish=2003/04/03&pages=06&seq=33
Relevant quote:
"Meanwhile, TSMC said it will kick-off volume production of chips on its 90nm technology in July at its Fab 13, as several of its key clients have qualified the foundry for the cutting-edge manufacturing process since it began developing the technology in late 2001, according to TSMC sources.
Industry sources believe companies such as Altera and Texas Instruments (TI) are likely to be the first in line to adopt TSMC’s 90nm technology for their chips. TSMC, however, did not confirm the speculations. K.Y. Ho, CEO of graphics chipmaker ATI Technologies, however said his company will not begin considering production using 90nm technology until 2004."
Hmm.
Entropy
speng said:
Hmmm....? Intel was doing flash ram at .09 microns last year! However, at the moment Intel is still 9-12 months out from shipping an .09-micron cpu. Big, big difference between shipping "something" at .09 microns and shipping cpus at .09 microns (not to mention vpus.)
WaltC said:
Re: your first point, I think we'll be seeing them sooner than you expect and the process is much further along than you are implying. A lot depends on the competitive situation. No need to get Prescott out the door just yet. P4's are competing fine, though 3.2GHZ is as high as the Northwood core is going to go.
Second point, very true. Same issue as TSMC making a CPU for Transmeta vs the NV30 @ 13m.
Multiple sources have pretty much said no R400 this year (Dave Rolston, Dave Orton and now Senor Ho) and I'm sure TSMC's 90nm process is the reason and previous behaviors with die shifts.
Its not economically feasible to design a board for two seperate fabs given their diversions in processes, also margins would suffer if one design was fabbed at two different companeis because of lower production numbers. So what I said was correct Nvidia has left TSMC and future designs will be produced at IBM. Unless of course Nvidia determines they cant get what they need at IBM, then they'll come back to TSMC with their tail between their legs.
Deflection
Newcomer
duncan36 said:
Its not economically feasible to design a board for two seperate fabs given their diversions in processes, also margins would suffer if one design was fabbed at two different companeis because of lower production numbers. So what I said was correct Nvidia has left TSMC and future designs will be produced at IBM. Unless of course Nvidia determines they cant get what they need at IBM, then they'll come back to TSMC with their tail between their legs.
Intel has their .09 micron Prescott scheduled for the Q3. Realistically were talking a year from then for TSMC to produce a GPU complexity design on .09. Expect 3Q 2004. They just started doing .13 GPU's!
Also, Nvidia said their high end would part would be fabbed at IBM. They will still use TSMC for their low end parts because of cost factors unless IBM gave them a sweat heart deal.
duncan36 said:Its not economically feasible to design a board for two seperate fabs given their diversions in processes, also margins would suffer if one design was fabbed at two different companeis because of lower production numbers. So what I said was correct Nvidia has left TSMC and future designs will be produced at IBM. Unless of course Nvidia determines they cant get what they need at IBM, then they'll come back to TSMC with their tail between their legs.
Future high-end chips will be made by IBM. TSMC and IBM will not be making the same chips, this has already been stated by nvidia. IBM handles the high-end NV40 while TSMC would handle the lower end NV41/44 (just examples based off NV31/34 numbers).
I think RussSchultz said that it was exactly what his company is doing. So i wouldn't say it's not economically feasibleduncan36 said:Its not economically feasible to design a board for two seperate fabs given their diversions in processes, also margins would suffer if one design was fabbed at two different companeis because of lower production numbers. So what I said was correct Nvidia has left TSMC and future designs will be produced at IBM. Unless of course Nvidia determines they cant get what they need at IBM, then they'll come back to TSMC with their tail between their legs.
Similar threads
