• Notice regarding image hosts and attachments. See here.

  • Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

The Ferminator

I didn't want to bring the A3 rumours up but I worked out Nvidia will be able to fit 89 ish full core's on one of TSMC's 40nm wafers. If 20-30 % A3 rumours are right that's 17 - 26 working cores at 5k wafer !

The average would be 22 cores a disk (at 230 a pop) with a 7-8 week lead time...

Most people are suggesting closer to the 100 mark for how many dies you can get off a wafer, assuming they all worked, but yield wise as you said, it could quite easily be as bad as around 20 a wafer, and frankly we don't really know if the 20% rumours are 20% including salvaged cores(in which case its an entire disaster) or without, which I'd guess is more likely. The problem is if they do end up doing intesively salvaged cores, I can't see the yields getting "that" high.

As I said before, anyone laughing at AMD for low 5870 numbers, is in for a BIG BIG shock from Nvidia's supply numbers.

But AMD's are largely because, no matter which company is ahead on performance, they aren't usually far enough ahead to stop all sales to the competition, infact not many. This is a strange situation where Nvidia simply aren't producing 260-285gtx's or Fermis for the last several months, they've shifted almost all their production to chucking out zero profit cheap as chips gt210's and the like to keep OEM's off their backs. AMD has basically their normal high/mid end demand, plus all the high/mid end demand that Nvidia usually has, but very little extra production. Twice the normal demand, no extra capacity and worse yields compared to previous gen due to crappy process = less cores available and more than twice the demand, it was never going to be fantastic. But then theres little point making 10 million 5870's one month, and nothing for the next 6 months.
 
Current rumours are that 99% of the working parts are the 448 shader parts with a B1 respin needed to get decent yields out of the 512 shader parts, so doubt we'll see the 512 shader parts in any sort of volume till Summer, they will probably send some to reviewers to retake the performance crown though knowing Nvidia.
 
Well thay said Q1 of 2010, so thats upto end of March right? So you would like to hope that they deliver considering its already late.

Q1 ends April 26th.

Furthermore, who says Fermi is late? Nvidia have never said high end Fermi parts would be delivered anytime before Q12010. And as there is still no real DX11 games there is nothing to be later for.
 
Q1 ends April 26th.

Furthermore, who says Fermi is late? Nvidia have never said high end Fermi parts would be delivered anytime before Q12010. And as there is still no real DX11 games there is nothing to be later for.

This is a joke right? Your saying that Nvidia knew they were going to have all these yield issues and re-spins so this is not delayed??? Think about it. If they could have released Nov/Dec 09 they would have.
As for no DX11 games... well, that didn't really slow the release of the 8800's when there were hardly any (if any) DX10 games?!?!
Why are you deffending Nvidia? I would say the same things if this was the other way around and it was ATI, so don't start no FB crap. THEY ARE LATE TO THE PARTY. How you can say otherwise, I find very amusing.
 
I thought the latest rumours now was that Nvidia are going to release mid range cards first now with the high end coming Q2?
 
I am not really bothered about nvidia cards, but I am not ati fanboy either, but it does hurt our wallets if you want to buy current ati card.

I wish they hurry up and get them out so ati prices go down a bit.

At least i hope ocuk is able to make slightly higher margin on ati cards now to keep providing us with their good service :)
 
i just dont know what to do ?
currently running flawless evga295 which i loosing money as the clock ticks toward fermi, im like, sell it and get 5870 or wait and see what nvidia have to offer.....its taking far too long
i keep telling myself that the 295 is still a top card and does play everything i want it too very very well. but i keep hearing the words DIRECTX and ELEVEN in my head lol
and from what ive seen and read on reviews going from my 295 to a 5870 is only going to gain me dx11 and a possible performance drop
 
Last edited:
This is a joke right? Your saying that Nvidia knew they were going to have all these yield issues and re-spins so this is not delayed??? Think about it. If they could have released Nov/Dec 09 they would have.
As for no DX11 games... well, that didn't really slow the release of the 8800's when there were hardly any (if any) DX10 games?!?!
Why are you deffending Nvidia? I would say the same things if this was the other way around and it was ATI, so don't start no FB crap. THEY ARE LATE TO THE PARTY. How you can say otherwise, I find very amusing.


My point is, no where did Nvidia state anything other than a Q12010 release date.

They are later than ATI, but ATI just spin out another turbo charged 4870 with DX 11 tacked on. Nvidia had a whole new architecture.

GPUs are getting so complex these days that the time between new architectures has to increase, which is what we have seen with Nvidia.

Not many GPUs ship on A1 silicone, so it is not unusual at all that a respine was required.
 
are there any half decent dx11 games out in say 1-2 months ? that nvidia may be plugging there new dx11 cards with ?
you know like ati and dirt 2
just a guess as to a possible dateish
 
My point is, no where did Nvidia state anything other than a Q12010 release date.

They are later than ATI, but ATI just spin out another turbo charged 4870 with DX 11 tacked on. Nvidia had a whole new architecture.

GPUs are getting so complex these days that the time between new architectures has to increase, which is what we have seen with Nvidia.

Not many GPUs ship on A1 silicone, so it is not unusual at all that a respine was required.

ATI built a core half the size of Nvidia with many times the SP count, jumped to DX11 and shrunk to 40nm in a recession. And done all this with a team a 3rd of the size of Nvidia's, working with a fraction of the budget.

Nvidia have had since 2007 to get a new architecture in place and they never. Complex core or not, Windows 7 is DX11 and from that launch on Nvidia are late. Now they are just holding up the TWIWMTP games developers and screwing PC gaming.
 
Last edited:
ATI built a core half the size of Nvidia with many times the SP count

regardless of all the other things between the two architectures, you can not compare the Sp's like for like, they are vastly different, to get to a similar level of complexity you would have to divide the ATI shaders by 5.

of course someone will be along in a min to clarify/shout me down.
 
You misunderstand what I'm saying. They built a core that had a lot more crammed into it and done it with half the size. If they had gone as big as Nvidia the gods knows what they would have added.
 
You misunderstand what I'm saying. They built a core that had a lot more crammed into it and done it with half the size

"had more crammed into it"?

I think you are missing the point. It matters not how many SP's each architecture has, if (for argument's sake) nVidia's SPs take up 200 transistors each, wheras ATI's only use 50.

This is the point that was being made, nVidia's SP pipeline is longer and more complex.
 
As regards to working within the design budgets. Im not going to start counting transistors and looking at what Nvidia and ATI consider the best pipeline.
 
As regards to working within the design budgets. Im not going to start counting transistors and looking at what Nvidia and ATI consider the best pipeline.

Then you can't say that "ati crammed more in". If you are willing to ignore technical specifications, then you can't exactly form a factual based argument.

As for the same (retarded) argument, you could turn it on its head and say that if nVidia shortened their pipeline processes, they could have doubled the count.
 
I'm not comparing ATI's pipeline length to Nvidia's directly though, you are. I'm saying ATI built a complex core at 40nm thats half the size of what Nvidia are tying to do, in quite a small process shrink.

Yes Nvidia will be faster as they have a core double the size.
 
Back
Top Bottom