ATI's R420 likely to use .13 micron process
Forum Graphic & Displays : Graphics Cards - ATI's R420 likely to use .13 micron process
It had better be, a new chip with so much more transistors and performance per clock would either:
1)Waste too many space per wafer due to transistor size and count
2)Run too hot and won't scale too far
--
If I could see the Matrix, I'd tell you I am only seeing 0s inside your head! 
Read my first option to see why this will become a very bad thing not to switch. Economically speaking.
--
If I could see the Matrix, I'd tell you I am only seeing 0s inside your head!
It is very bad that DX10 will come out this year and no DX9 games have been out yet. There won't even be enough by the end of the year. Now the evolution development is becoming 2 grades above the market evolution, it's outpacing it by too much. Not a good thing.
--
If I could see the Matrix, I'd tell you I am only seeing 0s inside your head!
Jeff K says "teh", and therefore it is cool, and I will mimic him.
www.somethingawful.com/jeffk
..although I think his page might be under construction right now.
I think for r420 thell go .13, the 9600 is giving them an opportunity to perfect it.
Theyll stay .15 for the 9900 and thats long enough.
Sure they could pick out the hot ones and get em to 500mhz, they could probly perfect it some more and up the voltage and go to 600, but comon. After 9900, its over, cept maybe for their low end.
Wanna make things interesting? 
"Every Day is the Right Day." -Pink Floyd
I just wanted to say, I think that the R420 will likely be on a 0.13 process, simply because ATI DOESN'T want to get behind on the process when they really ARE forced to use it. For now they have some flexibility. The have some (limited) experience with the 0.13 process on the 9600s, and they have the leway to simply have a competitive piece in the market (9800P vs 5900U [close enough]) So I think now, if any time, would be ideal, where there is no serious pressure to switch (do or die) and there is SOME experience with the technique.
While it is true that they could likely produce another chip using 0.15, it is likely that that WOULD be the last that they could effectively produce on the larger processor based on transistor count. It's likely to be 150mil. the chip after it? Probably some where on the high side of the 150-200million range, which I think would really be out of reach.
Now yeah that's speculation on my part, but it seems to make sense to be the time to baptise the high-end market, now when DX10 is over a year away. Therefore a blown core/cycle wouldn't kill ATI like it would if it blew it right at the release of the next gen. APIs. Now they 'MAY' have one additional product cycle between the R420 and the launch of DX10, but it depends on tuely what their product cycle is now. I would expect it to be a 0.13 core for the simple reason that they did benifit from their slow adoption of the process and from introducing it to the 'mainstream' market first (significantly underclocked to make it a stable experiment IMO), however the next step is to learn and perfect the 0.13 core on a high end card BEFORE it is a must have item. IF it is a poor performing core, they can likely pull an NV trick out of their A$$ and delay launch until a replacement is nearly ready to therefore have a paper launch followed by a good 'real' lanuch of the perfected card.
anywhoo, that's just my two frames worth.
- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! <font color=green>RED</font color=green> <font color=red>GREEN</font color=red> GA to SK
Day off man! I spent Wednesday night drinking (couldn't work a keyboard if I tried) at a bon voyage paty for a ski-buddy who's moving back to Q-bec; and all yesterday (thurs) recovering and watching OLD School, and Love Liza and the Recruit with friends, since there's no school,hockey,football or lawn bowling (or curling) to watch (Mmm Productive).
And now back at work. I did do a quick check of the forum on my friend's computer (freakin' dial-up! Oops sorry GW ) and added tothe D]|[ HDTV discussion, before returning for the last fill-um.
I'll be sure to try and make up for my absence.
- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! <font color=green>RED</font color=green> <font color=red>GREEN</font color=red> GA to SK
lol your insane
if ATI did that, then we would have another 5800ultra fiasco on our hands
@____@!!
-------
<A HREF="http://www.quake3world.com/ubb/Forum1/HTML/001355.html" target="_new">*I hate thug gangstas*</A>
Following Ape's reply, I'll answer in my own way why I think it HAS to be 0.13m:
| Quote : Gathering from your remark on wasting wafer space and transistor count, I'm gathering that your speaking about cost effectivness.
|
Indeed, cost and wafer size comes into mind. I am assuming they use the traditional 200mm wafer. Now if we extrapolate data based on THG's reports of how many Bartons are produced at 0.13m when they are 120mm^2, we can clearly see that for a 50 million transistor chip, space is already limited, so imagine 3 times more, which would likely mean nearly 3 times bigger core AT 0.15m. That is BAD. It only hurts ATi's money.
As for perfecting, what's there to perfect more here? We cannot stay at the perfection we are at, and not move forward. AMD ran into 0.13m issues, Intel did a bit, nVidia did. ATi seems to be doing very well by learning from the others. Could be that a high-end card will work very well!
So, ATi will only do benefit by saving wafer space. Productivity is a must, and while they could risk by switching, they simply can't continue operating a hybrid process fab.
| Quote : I dont think running too hot has ever been an issue with ATi's chips.
|
IT IS amazing, but it won't help if they add more than 30 million more transistors and clock it up. It simply won't help. At ~60W, the 9800PRO settles in on limits, and a new R420 at 500MHZ with much more advanced core logic, will very likely sit at 70W and that is pushing it FAR.
--
If I could see the Matrix, I'd tell you I am only seeing 0s inside your head!
Please mind
You are about to answer a thread that has been inactive for more than 6 months.
If you still wish to proceed, please ensure that your posting is original and does not duplicate or overlap any prior responses to this thread.
