NVIDIA GF100 & Friends speculation

fa615594-09b6-470a-add8-d247184c6cb8.jpg


2650*1600, 8xAA -- And the contestant, that's running out of memory is... guess who! :LOL:
 
http://tof.canardpc.com/view/fa615594-09b6-470a-add8-d247184c6cb8.jpg

2650*1600, 8xAA -- And the contestant, that's running out of memory is... guess who! :LOL:
Doesnt bode too well for 470, from a performance perspective.

Enter 5870 2GB at $399 and 1GB drops in price.

The Catalytic is probably an inside joke. :p
 
Doesnt bode too well for 470, from a performance perspective.

Enter 5870 2GB at $399 and 1GB drops in price.

The Catalytic is probably an inside joke. :p

I'm starting to believe the 2gb eyefinity 6 was held back deliberately as a counter to the 4x0's. Just about all of the "wins" we are seeing for the 470 appear to be the result of the 5870 running out of memory.
 
I'm starting to believe the 2gb eyefinity 6 was held back deliberately as a counter to the 4x0's. Just about all of the "wins" we are seeing for the 470 appear to be the result of the 5870 running out of memory.
Wasnt this the same case with the 4870? 512MB at launch and 1GB later.
 
Wasnt this the same case with the 4870? 512MB at launch and 1GB later.

Yes it was, the 1gb 4870 being released only 3 months after the 512mb if I recall correctly.

Is it the expectation the 5870 will now come with 2gb as standard, like the 1gb became standard with the 4870? I was sort of under the impression that the 5870 2gb would be a much lower volume part.

comparingperfimprovements.png

This is quite an interesting benchmark from AT, what ATI would give for a similar result this time around. ;)
 
Last edited by a moderator:
Yes it was, the 1gb 4870 being released only 3 months after the 512mb if I recall correctly.

Is it the expectation the 5870 will now come with 2gb as standard, like the 1gb became standard with the 4870? I was sort of under the impression that the 5870 2gb would be a much lower volume part.

comparingperfimprovements.png
This is quite an interesting benchmark from AT, what ATI would give for a similar result this time around. ;)

It'll probably end up being the enthusiast part of choice and therefore come with all manner of overclocks etc. I think perhaps this is the 5890 part people have spoke of with higher core clock and more memory, perhaps even alongside various non reference cooling solutions.
 
Tegra 2 and ION2 are 40nm TSMC too. I can imagine that the vaste majority from the 80% is Tegra2 and ION2 ;).
Tegra 2 is around 50 mm^2. Ion2 (aka gt218) around 60mm^2. And you'd think they'd get a bit more yield out of them than fermi...
 
So because we've decided that R870 is evolutionary and Fermi is a bigger change, that somehow absolves Nvidia from the need to execute?

And where did this fermi is revo stuff come from. It looks like a warmed over GT200 with a new front end unit to me, or are we now calling gt200 revo compared to g80/92? Both 870 and fermi are very much evolutionary changes to the chips they are based on.
 
Which doesn't mean they can't have problems. Same with NVIDIA. Their execution from the GeForce 6 series up to G92 was almost flawless. Still they had problems with GT200 and now the delays with Fermi.

True story, problems tend to develop when you spend as much time talking about cans of ass than you do solving problems. So far, ATI hasn't spent too much time talking about cans of ass. When the do, its time to sell.
 
And why do you assume that they'll always be following this strategy ? Fermi is a stepping stone for at least two generations of their products. Do you think that they will follow the same "big chip" strategy, if it doesn't work out for them ? G80 turned out to be very good, actually one of their best, if not the best. It was only with GT200 that they had "some" problems. If Fermi, in its current state isn't profitable, you can bet that they'll change their philosophy.

You do realize that gt200 is basically 2x g80 and g100 is basically 2x gt200 right?
 
If you look at all the changes it was a major rework/overhaul, if not a brand new/mint fresh one. For starters, you could begin by asking yourselves where did the 40% increase in raw flops/mm2 on the same process come from?

Hmm, significantly increasing the part of the design that does the whole flops thing?
 
Back
Top