ATi delaying transition to 80nm

[turtle]

http://www.digitimes.com/mobos/a20060417PR220.html

It's been the buzz for a while, but here's another source.

Sounds like the first 80nm product will be R600, with presumabley the refresh and perhaps lower-end products coming in 2007 at 65nm.

"R580+" almost by default now seems to be the the original R580 chip but with GDDR4 memory. While suspected before, this goes to continue that rational.

[nn_step]

That doesn't surprise me.. but I am likeing the addition of GDDR4 :D

[Bar81]

Well, this affects both gpu makers. TSMC needs to get its ass in gear.

[Turok]

Will the next gen lower-end cards like the r580+ have r600 features like X9.0L/DX10 compatibility and unified shaders with GDDR4?
Im sure they will have GDDR4 since ATi's GPUs since the r520 were compatible with GDDR4.
I will probably buy a r600 anyway, but just curious :P
The answer to this would prove if the r600 will be the only card that's X9.0L/DX10 compatible.

[cky2k6]

r580+ isnt next gen or low end, its going to be the x1950xtx. the first dx10 cards will be only g80/r600, there low end ones will come out later.

[ahmad]

Too bad. How did it ever come to the entire GPU world relying on TSMC and their technology to move forward? Me thinks ATI and Nvidia should start their own silicon manufacturer :p:

[Starscream]

Too bad. How did it ever come to the entire GPU world relying on TSMC and their technology to move forward? Me thinks ATI and Nvidia should start their own silicon manufacturer :p:


would need alot of cash to start one up and would take a long time to build.
They rather use the cash on developing new tech or buying another company.
that and the plant has to run 24/7 365 days a year.

when using another party to make the stuff for you you dont need to bother about what todo with the capacity when u dont need it.

[crodan85]

Don't think Ati or Nvidia are big enough players to even think of there own fabs. Your talking $1 Billion for the fab that will only have a usefull lifespan of five years and depreciate a mind-boggling $500,000 to $1 million per day. Then you got the R&D costs. Now thats some cash

[ahmad]

Don't think Ati or Nvidia are big enough players to even think of there own fabs. Your talking $1 Billion for the fab that will only have a usefull lifespan of five years and depreciate a mind-boggling $500,000 to $1 million per day. Then you got the R&D costs. Now thats some cash

I am sure they have the money. Whether they are willing to do it or not is another question.

STI spent $300 million (if I recall correctly, it might have been anywhere from 100- 700mil) on the plant that makes the Cell processor. Something that dedicated shouldnt be too bad. The only expensive part would be the researchers.

[Starscream]

I am sure they have the money. Whether they are willing to do it or not is another question.

STI spent $300 million (if I recall correctly, it might have been anywhere from 100- 700mil) on the plant that makes the Cell processor. Something that dedicated shouldnt be too bad. The only expensive part would be the researchers.


the start up isnt the chuck.
most of the cash will b in the day to day operations it doesnt matter if your making chips at that moment the costs are there and they are high.


when they let an other company make the chips they will only have to pay for the moments that the machines are making their chips and not for the time that the machines r down or being changed for a new chip or whatever.

Nvidia/ATI will have the money but they simply dont want to take the risc of pumping that much init and then run into trouble that the machines cant run for a while.
And like said when a fab is build its designed form the start that they can only do so much (much like how AMDs fab cant go smaller then 90nm)
so when the first plant is there they will already have to start making the 2nd one not far after.

i think it is simply a to big risc for ATI/Nvidia to take.