AMD: R9xx Speculation

Your comparison is a bit flawed as GTX580 is a far more powerful card and thus more prone to system bottlenecks. From the same link:

GeForce GTX 460 1024MB: 79%

 

I play BC2, its a new game with heavy driven stuff that pushes your PC a lot.
3dmark11 was released the 7dec, and a crossfire profile was added 2 days after for it.
Unwarranted fears for crossfire and scaling.
sure there been issues and some game plain sucks, but that is the same for sli.

a 6990 seems to be a worthwhile card

 

the 6870 is a more powerful card and thus more prone to system bottlenecks.

sorry I had to.

<edit> either way I think it has been established that sli doesn't have better scaling than crossfire

 

Scaling of High-end cards is usually worse than Med-end cards , because the chance of them running into CPU bottlneck is usually higher than Med-end cards .

Scaling of HD 5800s used to be average at best , not on par with NVIDIA , but HD 6800s are different story , (specifically HD 6850) they reached NVIDIA scaling and exceeded it by a decent margin .

 

I think this is what PowerTune is for. It is instantaneous clock management to manage instantaneous power. It should mean HD6990 is 2x full specification and performance is whatever comes out in the wash.

But only the gamers who buy gold plated Leicas will be buying one anyway, so it can be safely ignored.

 

Saw these posted at [H] from Chiphell

No clue why Nvidia confidential is watermarked. Could be fake, could be true /shrug

But something to chew on. Taken from this Chiphell thread




Also, shanshoft wrote this there:

I love video card releases

 

Depending on what you do, you could have a HD6950 for over a month now. But that AMD is very secretive, that's for sure.

 

I'm quite interested in finding any end in the middle of anything

Although high-end GPUs tend to be more CPU bound than low-end/midrange ones, AMD's drivers seem to have a better threading model when it comes to multi-GPU setups, as some CPU bound benchmarks still show higher results with CF/X2 boards.

 

Come on, the AAA titles should run fine (yet even these sometimes need new drivers)
And 3Dmark11 ... is not a game, only the lazy (VIA) don't optimise for it and they work together with AMD/Nvidia...
They even heavily toned down tesselation benchmarks in 11 in last weeks before release...

 

They did?, older shots and specs I have seen don't seem to suggest it.

 

Game Over


AMD could not contradict their own philosophy.

AMD delivers a hit in every price range, offering a card with more powerful and lower fuel consumption.

Obviously, avoiding to compete at herself in the range of GTX580, where the HD5970 is still the queen!

...and with Antilles on the starting blocks, ready to beat the opponent over all the line!

 

It would be quite interesting to do an in-depth analysis with this particular benchmark, as it seems to allow for a fine tuning of tessellation and, quite importantly, uses a more "real world" workload than any other DX11 synthetic benchmark.

I'm still quite disappointed by the low performance given the actual output (20-30fps "performance" preset with a 1-year old quasi-enthusiast board and similar results for the all new top-end "performance" board - HD6870 - ?), though.

 

It seems like some S|A forum's comment about AMD sending fake presentations is true...

How could a board between HD6870 and GTX480/570 be in "a class of its own"? Theres something like a 30% advantage for the GTX480/570, so it's either too slow to be worth producing (less than 20% faster) or quite competitive against the GTX480/570.

The "Best Enthusiast GPU" is quite strange too, since GTX580 is single GPU too and appears in the "World's fastest card" segment.

 

Do you think is fake?

Price class...

Enthusiast GPUs are under 400 €

 

it's all marketing talk, though the lack of an nda expiration thing at the bottom teh slides typically have had for leaks is odd

but its most likely real - of course, AMD also often low balls where the card actually stands in its comparisons with Nvidia

 

Fastest isn't always best, is it?

 

I already found it strange, when high-end was re-branded to enthusiast and the former moved one notch below that.

 

HD5970 "Fastest Enthusiast board" > GTX580 > HD6970 "Best Enthusiast GPU" > GTX570 is clearly strange.

If Cayman ends up being a "small" die with minor improvements (perf/mm², tessellation performance) and even some pitfalls (perf/watt with 190 watt playing games) compared to Cypress, it will indeed end up like that and disappoint me.

It's clearly not impossible but, honestly, I don't buy it.

 

Best vs Fastest

 

didn't [H] pick the 580 over the 5970?