AMD: R9xx Speculation

[CarstenS]

Hold up, is the 6970 only able to do 4770 in 3d11? That can't be right, that's no better then an OC 5870.

Seems like the guy already posted another score: http://www.hardwareluxx.de/community/15864500-post284.html

 

[MarkoIt]

How can Cayman be just barely faster than Cypress in 3D Mark 11? 3x better tessellation, more texturing power, improved RBE, 20% more SIMD, more BW.
3D Mark uses heavy shader with a lot of trascendal op?

I'm pretty sure that Cat 10.11 cripple the card, and that reviewer got the good drivers.
I still think that the 6970 is going to be really close to the GTX580, and that the 6950 will match the GTX570.

 

[NathansFortune]

Well going by my maths I think 10.11 is already relatively optimised for the new architecture, but maybe 10.12 includes new performance boosts. If not then the real world performance will have to break the benchmark or we are in for 10-15% above 5870 and only just competitive with GTX570.

 

[no-X]

And any such measurement is subject to 0.5mm of linear error...

I think NVidia said GF110 is slightly smaller than GF100 and I don't know of any reason to believe otherwise. I don't remember NVidia ever saying how big GF100 is, but I think the reviews of GF110 had a number supplied by NVidia.

I remember that nVidia told to Rys, that GF100 is under 500mm² (~480mm², or something like that) and it was proved later, that the info was incorrect...

 

[NathansFortune]

Seems like the guy already posted another score: http://www.hardwareluxx.de/community/15864500-post284.html

Better, well 4950 is better, but I would have figured 5100-5300 range. Maybe the new drivers will bring a decent boost.

 

[MarkoIt]

Better, well 4950 is better, but I would have figured 5100-5300 range. Maybe the new drivers will bring a decent boost.

The graphic score is the same.. the physics score improved a lot.

 

[NathansFortune]

OK, I can't read German so I'm not so clear on which score is which!

Even so I think 10.12 will bring a decent performance boost or the real world performance is better than the benchmarks indicate and we will see some OC editions from the off.

 

[leoneazzurro]

It seems that the fps in the graphic tests are not far from the 580, so it seems that 3DMark 11 result is heavily physics dependent, or something similar...
Extreme preset: 1771, that is almost 580 level...

 

[CarstenS]

Does that imply that 6970 would not be very good a the combined physics test?

 

[zshadow]

http://3dmark.com/3dm11/135040

1776 3DMarks under Extreme preset. Pretty sure that tops a 480..

 

[mboeller]

http://3dmark.com/3dm11/135040

1776 3DMarks under Extreme preset. Pretty sure that tops a 480..

at least better than a GTX570.

Link: http://www.computerbase.de/news/har...11-zwoelf-aktuelle-grafikkarten-im-vergleich/

 

[NathansFortune]

It seems that the fps in the graphic tests are not far from the 580, so it seems that 3DMark 11 result is heavily physics dependent, or something similar...
Extreme preset: 1771, that is almost 580 level...

GTX580 with AMD 1055T vs 6970 with AMD 1055T:

	6970	GTX580	Relative performance
G1	21.52	27.85  	129%
G2	24.88 	29.15	117%
G3	30.51	39.5	129%
G4	13.46	19.62	145%

Well on perfomance the graphics tests are pretty well in favour of GTX580.

 

[NathansFortune]

at least better than a GTX570.

Link: http://www.computerbase.de/news/har...11-zwoelf-aktuelle-grafikkarten-im-vergleich/

The 2GB will be a big advantage in the extreme test at higher resolutions. Maybe a good decision to make it 2GB standard than 1GB vs 5870?

 

[NathansFortune]

Same comparison at extreme:

	6970	GTX580	Relative performance
G1	7.69	8.93  	116%
G2	9.13	9.15	100%
G3	8.02	8.69  	108%
G4	4.81	5.38  	112%

 

[Xenus]

And even then it's a difference of <10% on two different systems which tells us nothing. It could be less that that or near zero or greater then that and near 15-20%.

 

[Psycho]

Heaven result looks better. 25fps for 1920x1080x8 tesselation@ridiculous : http://img834.imageshack.us/img834/1345/6970heavenbench.jpg
580 on same settings and almost same setup is 30 fps, according to http://www.nvnews.net/vbulletin/showthread.php?t=156936

 

[ECH]

Why is he dodging questions about his CCC settings? I too would like to know what settings he's using.

 

[leoneazzurro]

And what about the famous "switch"?

 

[simbus82]

GTX580 with AMD 1055T vs 6970 with AMD 1055T:

	6970	GTX580	Relative performance
G1	21.52	27.85  	129%
G2	24.88 	29.15	117%
G3	30.51	39.5	129%
G4	13.46	19.62	145%

Well on perfomance the graphics tests are pretty well in favour of GTX580.

Sorry but i have this FPS in a system with HD5870 and a phenom II X4 3.6Ghz.


imho ... fake! or wrong benchmark method.

 

[ECH]

There is something not being told/addressed about those results. For example the F1 2010 and HB results are in contrast much better then the 3d11 results. No, I'm not comparing 3d11 to games or HB. It's just that if there is a higher level of IQ being used the 6970 seems to do fairly well. Which is why I would like to know his CCC settings. Perhaps better drivers will improve on this but I have to assume that the drivers that come on disc have to work correctly with the card.