[HKPolice]

RV790 = 55nm with more SPs that's why it's only at 750Mhz. I keep saying this but you tards keep believing that it's 40nm.

RV870 is NOT delayed, there is always at least 1 refresh before a new architecture.

[badsykes]

Originally Posted by psolord

I don't like this whole rv790 story. Rv790 translates to me as "delayed rv870".

I will like it if course if they make something great out of the rv790, which i doubt, because then, what greatness would be left for the rv870 and when? We'll see!

There is no point to release rv870 now..This card is aimed to be DX11 ready and it will be lunched to sustain Windows 7 lunch..For all today games you already have monsters to sustain them..Rv790 is a way better choice now because they transit to 40nm and cut the cost down and add some more graphics power in term of instructions..For now this is sufficient..Don't get me wrong but for me DX10 was a failure and DX11 will not be so much more in the begging so i don't see the point of rushing rv870 now..

[Reaperrr]

Originally Posted by konstantine

that could be pretty much what the RV790 is meant to do.Still, the RV770 has an obvious bottleneck.The chip at 750MHz can achieve a maximum computing power of 1200 Giga Flops in a single precision vs only 622 G.F. for the gtx 280 at 602MHz.Yet, the 4870 is not twice as fast as the 280gtx, inspite of it being able to execute double the amount of the floating points of what the gtx280 can execute.

So the computing efficiency of the 4870 is around... half of that of the 280gtx.Solving this problem, may lead AMD to a RV770 chip with double the computing efficiency of the original RV770.And that may translate to a single chip card that may perform better than the 4870x2....

1) From my knowledge, the GTX280 has 933 GFLOPs, not 622. GT200s shader processors are completely different from RV770s and cannot be directly compared anyway.

2) The GTX280 has 80 TMUs instead of 40, 32 ROPs instead of 16 and more memory bandwidth (512bit x 2.2GHz vs. 256bit x 3.6GHz). And afaik at least the TMUs are more efficient on top of that.

@HKPolice: Because making another 55nm chip doesn't make any sense. It would be more expensive to make and more power-hungry than RV770, with only minor performance gains. It's just pointless.

[HKPolice]

Originally Posted by Reaperrr

1) From my knowledge, the GTX280 has 933 GFLOPs, not 622. GT200s shader processors are completely different from RV770s and cannot be directly compared anyway.

2) The GTX280 has 80 TMUs instead of 40, 32 ROPs instead of 16 and more memory bandwidth (512bit x 2.2GHz vs. 256bit x 3.6GHz). And afaik at least the TMUs are more efficient on top of that.

@HKPolice: Because making another 55nm chip doesn't make any sense. It would be more expensive to make and more power-hungry than RV770, with only minor performance gains. It's just pointless.

Wrong. RV790 will have more shaders and TMUs and it will be the same size or only slightly bigger than RV770, therefore keeping chip costs about the same. How? By cutting out useless features like Sideport and the DDR3 controller.

Combine that with being pin compatible with RV770 and not needing a new PCB design and ATi basically gets a free 20% performance boost to compete with nV's 55nm parts while being able to charge a premium over RV770 cards.