AMD Radeon HD6950/6970(Cayman) Reviews

[zerazax]

Where does this 30% come from? According to the TPU review 6970 is 13,6% faster than 5870 at 1920x1200. The die size increased ~16% at the same time. Epic indeed.

TPU has the 5870 up the most BY FAR on the 480/580. Look at HWC and other reviews - 5870 isn't as fast as TPU says

And as I wrote elsewhere, games where the 5870 will be close to the 69xx's are where they're heavily shader based and where there isnt much tesselation/compute/DX11. 69xx's pull ahead in the DX11 heavy titles like Stalker COP and Metro

No matter how much people say, AMD could have done a lot better with Cayman. They chose not to because of one simple assumption: That Antilles will be the only dual gpu in the market like 5970 was previously.
If they assumed nvidia would have a faster counter to antilles I garantee you AMD would have put more shaders into Cayman so not to lose on every front.

And you know this assumption... how?

More shaders != more performance all the time. Look at how Barts does compared to Cypress with far lower shader count.

Things are very different this time and AMD is losing on both its high end cards from day one. There is no six month grace and the reality is that nvidia is simply better right now. There is not even a significant power consumption difference. Count on it that in the next six months nvidia will take back most of the dx11 market share it lost in 2010.

You keep posting this ridiculous BS over and over. Doom and gloom oh no's.

Why don't you read what I wrote in reply earlier:

ding ding ding ding that is the CORRECT answer sir!

Its not that these cards suck, it's cause Amd's small chips lower power strategy failed this time around.

If this card were on 32nm with all the features Anand hinted at, would people still complain?

The most likely issue is that 32nm was cancelled and AMD wanted to recuperate its R&D cost, so the best thing to do is sell it at 40nm (which they're quite familiar with) and hope for the best until TSMC gets its act together for 28nm. The fact this supposedly all happened in less than a year's time clearly caught AMD by surprise so it's amazing they even managed to get cards out that fast, given that most cards are planned out far earlier

These cards are shader starved and thing are worst off this round then they were last round. At least the 5870 had clear performance and power advantage over the 470.
Also I am not too pleased with the 6950, its basically 10%% faster than Barts in most cases and that is barely relevant. They could just as well given 6870 2gb and more clock and called it a day.

10% faster than Barts? What reviews are you looking at? The vast majority has the 6950 awfully close if not right at the 570 - the card that's dissapointing is the 6970, which is just 10% faster than the 6950

Of course, that's just proof that the 69xx's are a forward looking architecture - in older games they don't beat the 5870 by much, but in newer engines they do by a good amount - look at Stalker and Metro.

Amd should be praying there won't be a dual fermi in the works cause that will destroy their strategy on all levels then.

More ridiculous-ness. When the:

GTX 295 > 4870 X2
GTX 285 > 4890
GTX 275 > 4870
GTX 260-216 > 4850

Did AMD's destruction across every single card matter? Of course not, the 4800's brought AMD back in market share and everything.

Now that Cayman is closer but not quite that the performer of the 580, its doom and gloom? Seriously?

How's this also for perspective:

Look at where the 5870 performed at release - it lost to the 4870X2 in a lot of things and the GTX 295 as well. Look at where AMD's drivers and game optimizations + game development has gone - the 5870 is clearly ahead, and it's even creeping up on the GTX 480 in performance (at release, the GTX 480 was a good 15-20% faster, now we have situations where the 5870 can close within 10%).

Let me ask this: who's more likely to get a boost over the next year, the 580 based on the 480 or the 69xx based on nothing prior?

And to say nothing of the fact that if 28nm really did get delayed by TSMC to 2012, as some rumors are swirling now, who will be in the better position to deliver another 40nm card? The company with a 530mm^2 GPU flagship or the one with a 389mm^2 GPU flagship?

Sheesh, some people need to seriously calm down and look at perspective here outside of JUST raw performance #'s

God people have such a short term memory