Nvidia GT300 core: Speculation

Status
Not open for further replies.
Sorry, but ATI does not show me their Chipdesign-Roadmap. But they did talk about a tick-tock approach.

1. part new DX: based on existing design where possible.
2. part: optimize for new DX11 where possible and/or necessary

But it is way too early to speculate about a R870 refresh.
 
Sorry, but ATI does not show me their Chipdesign-Roadmap. But they did talk about a tick-tock approach.

1. part new DX: based on existing design where possible.
2. part: optimize for new DX11 where possible and/or necessary
"optimize" is something different than "rehaul" the whole shebang and "doing it right"

But it is way too early to speculate about a R870 refresh.

I would beg to differ.
 
GT300 will be out by the end of November

? You mean the new GeForce "300" series, which will include the GTX380 (which will definitely get some press time in the coming months.)
Or the chip that this topic covers and we refer to as "GT300" which is a new generation after GT200 and GT21X?
 
? You mean the new GeForce "300" series, which will include the GTX380 (which will definitely get some press time in the coming months.)
Or the chip that this topic covers and we refer to as "GT300" which is a new generation after GT200 and GT21X?

I'd say he means the "GeForce 380" (which is a likelier naming scheme IMO). You should have rather asked if he means an architectural introduction for late November or a hard launch.
 
GT212 will come out, maybe this year.. probably next year.


I totally agree with DegustatoR's arguments against this one. Moreover, you're the only one on the www to say this.

Maybe this time you'll be quoted ? ;) (like w0mbat)

It's hard for me to imagine nV switching engineers between projects and then switching back.. pseudo-randomly
 
Apocalypse? where am I describing that?

GT212 will come out, maybe this year.. probably next year.
GT215/6/8 will have a very hard time competing (would've been different if they launched in Q209 as intended and not Q409) and GT300 is not set for another 9 months.
This is not the Apocalypse, it's the result of a design decision made way back when 40nm wasn't available. Tables can turn in 2010.

anything more then 3 months and even that is pushing it, its a whole quarter they won't be able to compete, would be a disastor. hmm 10% yeilds, that would give 20-30% yeilds for their current smaller chips, I would say that is very unrealistic, givin that AMD's next gen chips are larger then nV's 2xx 40 nm chips and they are probably coming out with a solid launch.
 
Well, one could also argue that RV770 and it's family where DX11-esque parts that came early. I mean, it was a fairly major redesign for a "refresh" part. Also, wasn't it mentioned somewhere that RV770 and the "terascale" engine where going to be the building blocks for there future architecture(s).
 
anything more then 3 months and even that is pushing it, its a whole quarter they won't be able to compete, would be a disastor. hmm 10% yeilds, that would give 20-30% yeilds for their current smaller chips, I would say that is very unrealistic, givin that AMD's next gen chips are larger then nV's 2xx 40 nm chips and they are probably coming out with a solid launch.

Yes, I thought it was unrealistic at first too, but they have been beavering away at TSMC since 2008 on GT21x and there still hasn't been one single product on the store shelves.

Let's turn it around. I'll put this from your point of view, The roadmap to "we have completely realistic yields on all products on 40nm"

Early 2008: GT200 comes out on 65nm, Engineers cheer, drink Champagne.
A couple of weeks later, the second spin 55nm GT200's come out of TSMC, more cheering, more drinking of Champagne and some TESLA demos.
Okay, we've got some leakage here, let's make a b3 revision of GT200. success... (something) ...moneys.

June: Everyone goes on summer holiday, PR department notifies the press that D12U will be the first nVidia chip using GDDR5 (pushes two extra pins in Maccri voodoo doll)

August: Guys.. it's August! ... No sweat, we've just taped-out our designs for GT212/4/6/8.... they'll use GDDR5 (hint!) more moneys!

September
October
November

December.. well GT212 isn't going to work (leakage), 214 is a bit ambitious (leakage) but 216 and 218 are good to go after a spin, we promise!

January

February: Success! we've got GT216 and 218 back from TSMC and we've got great yields!

March: GT300 Tapes-out, AMD announces RV740.. no problem, thanks to our superior yields we'll flood the market with our GT216 and GT218

April: What you Mean? HD4770 shipping? They can't be having less leakage than we do?

May: What we do now? oh.. just.. . just tell them GT300 taped out! in the mean-time, we'll tell that we can't produce enough GT216 and 218 because we had zillions of ODM design wins! tell them it's hundreds!

June

July: Announce something! We have over 200 OEM and ODM design wins! tell them they can buy exactly two systems with a GT218 in it! (GeForce G 210: HP&Gateway)

August (?! AMD have something over 100mm2 in 40nm?!)

September: Success! the first Review of GT218

Please don't pay attention to RV710 outperforming GT218 and doing that at 5 to 20% lower power consumption whilst being 55nm.
 
Last edited by a moderator:
Actually, it was a more original way of presenting things (altought biased, but of course it was on purpose).

I wasn't even expecting that review link while reading the post ;) .
 
I know I'm biased, but people here are thinking that a 3 month delay is already out of the question.
Dumb question, but three month delay from when?

Does it start when AMD launches their DX11 part? Or does it start when Windows 7 is available, when SHOULD they have launched it?
 
Yes, I thought it was unrealistic at first too, but they have been beavering away at TSMC since 2008 on GT21x and there still hasn't been one single product on the store shelves.

Let's turn it around. I'll put this from your point of view, The roadmap to "we have completely realistic yields on all products on 40nm"

Early 2008: GT200 comes out on 65nm, Engineers cheer, drink Champagne.
A couple of weeks later, the second spin 55nm GT200's come out of TSMC, more cheering, more drinking of Champagne and some TESLA demos.
Okay, we've got some leakage here, let's make a b3 revision of GT200. success... (something) ...moneys.

June: Everyone goes on summer holiday, PR department notifies the press that D12U will be the first nVidia chip using GDDR5 (pushes two extra pins in Maccri voodoo doll)

August: Guys.. it's August! ... No sweat, we've just taped-out our designs for GT212/4/6/8.... they'll use GDDR5 (hint!) more moneys!

September
October
November

December.. well GT212 isn't going to work (leakage), 214 is a bit ambitious (leakage) but 216 and 218 are good to go after a spin, we promise!

January

February: Success! we've got GT216 and 218 back from TSMC and we've got great yields!

March: GT300 Tapes-out, AMD announces RV740.. no problem, thanks to our superior yields we'll flood the market with our GT216 and GT218

April: What you Mean? HD4770 shipping? They can't be having less leakage than we do?

May: What we do now? oh.. just.. . just tell them GT300 taped out! in the mean-time, we'll tell that we can't produce enough GT216 and 218 because we had zillions of ODM design wins! tell them it's hundreds!

June

July: Announce something! We have over 200 OEM and ODM design wins! tell them they can buy exactly two systems with a GT218 in it! (GeForce G 210: HP&Gateway)

August (?! AMD have something over 100mm2 in 40nm?!)

September: Success! the first Review of GT218

Please don't pay attention to RV710 outperforming GT218 and doing that at 5 to 20% lower power consumption whilst being 55nm.


This whole thing goes by TSMC's problems with the 40nm process, planned doesn't mean tape out, I don't remember ever for the gt212 before q2 09, it was slated for q1, but delayed to q2, then after then it was pretty much cancelled because of overlapping with the g300 launch which possible is D12U which those rumors were end of 09 which was stated in June by memory makers.

We all know that the reason for the delay is the 40 nm process it was quite apparent after AMD's woes too.

Oh please use benchmarks that get over 30 fps, cause what they showed there well, gaming performance is kinda pointless.
 
Oh please use benchmarks that get over 30 fps, cause what they showed there well, gaming performance is kinda pointless.

Well yeah Nvidia obviously wasn't reaching for the stars with GT218 so a performance comp is pretty pointless. What's startling is the higher power consumption vs RV710. But what's even worse is that it's < 60 mm^2. If they can't make that in volume then what can they do?
 
Status
Not open for further replies.
Back
Top