AMD: R9xx Speculation

[AnarchX]

Uns fiel zu Beginn der Tests auf, dass der anisotrope Filter der HD-6800-Karten im Vergleich mit einer Radeon HD 5870 stärker flimmert. Nach Rücksprache mit AMD teilte man uns mit, dass die Standardeinstellung "Qualität" aggressiver filtert als der bisherige Treiberstandard (A.I. Standard; dort seien die Texturfilteroptimierungen bei der HD-5800-Reihe bereits deaktiviert)

http://www.pcgameshardware.de/aid,7...irectX-11-Generation/Grafikkarte/Test/?page=4

They recognized that the new standard "quality" setting has more shimmering than the old AI standard setting. AMD confirmed that optimizations are more aggressive in the new quality setting.

So AMD fixed the banding problem, but increased sample optimizations in the driver default setting, which is used by the most reviewers. This probably gives a significant performance gain over GeForce and the old Radeon HD 5000 @ Cat 10.9.

 

[Sontin]

Cat AI was becoming cumbersome and confusing in terms of what it did and what got enabled when. This is a much clearer interface, with more control over the settings for the users, without destroying things like Crossfire performance. This is also very similar (perhaps even less confusing) than NVIDIA's options and sets similar optimizations by default. In all, you should be pleased about the new options as its offering more transparency.

Yes, it's great for your customers.
But my question is: Why do you change the image quality of the default setting? Quality is not as good as the old setting. I hope the reason was not nVidia's Quality. Because you didn't complain about it over the last 4 years...

 

[Entropy]

They dropped the ball with Barts. No real improvement with Tessellation.

And tesselation performance impacts real applications how, exactly?
I've waded through 30-40 pages or so dealing mostly with bickering about a benchmark with no predictive ability, and a feature with little to no current relevance (and judging from people on these very boards involved with art creation, rather murky prospects going forward as well).

If this is Barts "problem", I'd say it doesn't have any. I'm far more interested in an analysis of organization vs. performance, and figuring out how it achieves its apparent improvements in efficiency overall.

 

[Kaotik]

So, european users, which country has the cheapest cards at the moment, and do they ship to other EU countries?
At least here in Finland only one shop has them listed, not in stock, and the prices are skyhigh (240€+)

 

[MrBond]

Forgive me if this has already been asked (coming into this very long thread late), but I'm pretty confused about the 5770 situation. Is it confirmed that 5770/5750 is going to be rebranded as the 6770/6750? All the reviews seem to be pretty cagey on what will actually happen to the name.

 

[Kaotik]

Forgive me if this has already been asked (coming into this very long thread late), but I'm pretty confused about the 5770 situation. Is it confirmed that 5770/5750 is going to be rebranded as the 6770/6750? All the reviews seem to be pretty cagey on what will actually happen to the name.

It's been confirmed that 5700-series will continue to exist, and will not be renamed (at least 'till the end of the year)

 

[Tchock]

Forgive me if this has already been asked (coming into this very long thread late), but I'm pretty confused about the 5770 situation. Is it confirmed that 5770/5750 is going to be rebranded as the 6770/6750? All the reviews seem to be pretty cagey on what will actually happen to the name.

Rick Bergman was mentioning that the 6 series launch this year (40nm) consists of 6900s (Cayman + Antilles), 6800s (Barts), 6500s (Turks), and 6300s (Caicos?)


6700's probably saved for a 28nm shrink (Note: 4770) with 128-bit bus and such. Pipe cleaner?

 

[ninelven]

So AMD fixed the banding problem, but increased sample optimizations in the driver default setting, which is used by the most reviewers. This probably gives a significant performance gain over GeForce and the old Radeon HD 5000 @ Cat 10.9.

If this turns out to be true.... just... /facepalm.

 

[stevem]

Dave, are you able to confirm whether the ref boards tested are 6 or 8 layer pcbs?

 

[DieH@rd]

First pricing spotted in my home country of Serbia. 210 Euros for Gigabyte 6870.

Not bad, i expected little higher.

 

[Kaotik]

First pricing spotted in my home country of Serbia. 210 Euros for Gigabyte 6870.

Not bad, i expected little higher.

Nice, ~40 euros cheaper than Finland (our VAT is 23%)

 

[DieH@rd]

Maybe its some early promotion, who knows.
(our taxes are 18%)

 

[no-X]

LolwuT?

HD5850@AMD.com

TDP doesn't equal consumption (nvidiwatts) of course and I dont' think guru3d does accurate measurements but just wall measurements.

It really was 170W...



/but Dave's words explain it...

 

[Jawed]

Why? What makes you think NVidia should have significant improvements in the few months since the 460 was launched?

It's not just the GTX460, it's every card in the review except Barts-based. And yes, I think with a ~centred upon $200 range of $20 increments for the key cards being reviewed, from $170-$250, a 10% increase in performance over a few months makes such a cack-handed review useless.

 

[Jawed]

There is no "Cat AI standard" going forward. The optimization behaviour has been unified across all boards from Cat 10.10 onwards, so the default beviour on Cypress is different, but you can get back to Cat 10.9 quality on Cyress by changing the slider to high quality.

Wow, reducing the quality of default texture filtering on Evergreen is exactly what everyone was asking for

 

[fellix]

Along with selectively reducing functional blocks from Cypress and removing FP64 support, AMD made one other major change to improve efficiency for Barts: they’re using Redwood’s memory controller. In the past we’ve talked about the inherent complexities of driving GDDR5 at high speeds, but until now we’ve never known just how complex it is. It turns out that Cypress’s memory controller is nearly twice as big as Redwood’s! By reducing their desired memory speeds from 4.8GHz to 4.2GHz, AMD was able to reduce the size of their memory controller by nearly 50%.

Wow! AMD really fought hard for every sq.mm in Barts.

 

[CarstenS]

PCGH also does not take into account the high precision LOD and pure angle invaraint AF that is offered in all modes but not available in NVIDIA hardware.

Hi Dave,

That's why we headlined this section of our article "Neuerungen beim anisotropen Filter " which I would into "Anisotropic Filtering: What's new". And AMDs better angle-invariance is not exactly new, if I am not mistaken, but rather a thing we've come to enjoy for over a year now.

 

[Jawed]

It turns out that Cypress’s memory controller is nearly twice as big as Redwood’s! By reducing their desired memory speeds from 4.8GHz to 4.2GHz, AMD was able to reduce the size of their memory controller by nearly 50%.

Wow! AMD really fought hard for every sq.mm in Barts.

Wow that's a vast difference for such a measly gap.

 

[seahawk]

I do not understand this nitpicking about the AF. The Banding is gone, HQ is available and the cards just fly and make everything NV sells from 460 to 480 obsolete. AMD has once again delieverd an absolute dominating card. I can´t wait for Cayman.

 

[UniversalTruth]

I think that R5870 doesn't show its full strength. Perhaps the drivers. That creates the illusion that R6870 is close enough.