AMD RV770 refresh -> RV790

True, but when you consider that the fastest RV770-based SKUs aren't far behind the fastest GT200-based SKUs given the same memory size (and at a much lower price), and those dual RV770 SKUs are MUCH faster than a single GT200, it seems a rather hollow "victory" for NV to have the fastest single GPU. Let's not even drag die size (and all its associated pitfalls) or power consumption/heat dissipation into the matter either, as things get even uglier for NV there

The answer is, of course, no.

 

I didn't mean serious trouble, just "in danger" of losing the overall performance crown.

Although that will inevitably effect their bottom line.

 

Why of course? It's possible. Hard and unlikely but if they want to have the single card performance crown back they'll probably do it.

 

I don't think they have a thermal solution that can cope.

 

id like to see a 55nm 260GX2, 1GB per GPU and 600/1350/2300.

 

With what, exactly? 2x GT200b at clocks at least equal to GTX 260? In fact, I'd think, they'd keep the core clock and only improve shader clocks a bit. If power is that crucial, they could probably use some better memory (in a sense that they use clocks that allow them to have a lower voltage step).

What's more of a problem though is to get rid of the hot air with the sandwich-style GX2 9800 used. They barely have any space left for a slot exhaust. That's something they'd have to keep in mind - IMO.

 

May I ask why you think that? General curiosity. No comment of course.

 

I think I expected a 40nm refresh of RV770 in the new year, much like the R600 to R670 move. It would have been great to have even cheaper and more power efficient 4850 and 4870 chips. However, this seems unlikely now due to the process not being ready. I would expect RV790 to be a 900-950Mhz RV770, if it appears at all.

 

It's more of a problem of external bracket design. They'd better make a big exaust hole on the back of the card instead of doing some lame additional interfaces.
From termal management point of view a sandwich-type multichip board is probably better than Radeon X2-type board -- both chips are cooled with cold intake air while on the X2 board the second chip is getting the hot air from the first radiator.

 

That's exactly what i was trying to say.

 

That doesn't matter if the solution is designed correctly. We're pretty much at a 300W board and thermal management is fine.

 

That matters how? 1-2C difference between the cores, if even that...

 

I dunno, a GX2 is looking increasingly likely. And logically it will be based on at least a GTX 260.

A dual GTX260 216SP would trade blows quite nicely with the X2 but it wouldn;t a be a clear win. They need to increase clock speed or sp's for that.

http://firingsquad.com/news/newsarticle.asp?searchid=21027

 

jan 9th is my bday hmm

 

9800GX2 has lower clocks than 9800GTX,

http://www.gpureview.com/show_cards.php?card1=560&card2=559

doesn't it seem likely that a GTX260GX2 (wonder what they would call it, that's stupid) would be similarly crippled?

Jawed

 

Possibly, but back then, NV didn't need to have it running at the same speed. This time they absolutely do. Anything less that stock 260 clocks and it will come second place to the X2. And if that happens, you kinda think, well whats the point?

 

I wonder how fast this goes:

http://www.force3d.com/downloads/images/4870x2_acxxx.jpg

Doesn't seem to be a real product though

Jawed

 

What's the point of 3 giant fans, huge heatpipes and 3 heatsinks... when it's pointing the wrong way (leading heated air back into the case, instead of going through the vents out back) ?

 

Maybe that's why it's not a real product

Jawed

 

http://www.hardware.fr/articles/726-1/b-maj-b-radeon-hd-4800-gddr5-utile.html