Huh?? My HD5770 can bet that :smile:.
Driver Packaging Version 8.782.1-101029a-107695E
EDIT: My run as well as HD6970 were done @1920x1080!
AMD: R9xx Speculation
- Thread starter Lukfi
- Start date
Huh?? My HD5770 can bet that :smile:.
Driver Packaging Version 8.782.1-101029a-107695E
EDIT: My run as well as HD6970 were done @1920x1080!
MrGaribaldi
Regular
Huh, CAT 10.11?
According to the driver selector on amd.com only 10.10e supports HD 6xxx series.
(which has serious trouble with some simple OpenGL programs of mine on the HD6870
)
Or is that special CAT hotfix-review-special-sauce 10.11 drivers? If so, anywhere to download those?
According to the driver selector on amd.com only 10.10e supports HD 6xxx series.
(which has serious trouble with some simple OpenGL programs of mine on the HD6870
)Or is that special CAT hotfix-review-special-sauce 10.11 drivers? If so, anywhere to download those?
It looks more professional, that's all...
MrGaribaldi:
try this http://forums.guru3d.com/showthread.php?t=334207
try this http://forums.guru3d.com/showthread.php?t=334207
Probably not fake this time ..
If this is true it's pretty good. According to the one site (Tech Power Up)that does a performance summary, 580 is only ~15% faster than 480.
Of course we can expect some AMD fudge factor perhaps, but if it's even 5% slower than 580 that's pretty good.
It's funny too me how last two gens AMD has been sitting at sweet spot 15% below Nvidia top GPU, and it was all doom and gloom for Nvidia on these forums. This time they may get closer but now it's all bad for AMD.
The 3Dmark scores seem like they may be more disappointing than the games. Sadly I bet some people wont buy it just because it doesn't score well enough in 3Dmark regardless of game performance.
This is fairly interesting. I wouldn't really expect the 6970 to beat the 5870 in Perlin Noise, but to be on par.
I think either
a) the compiler doesn't do a very good job yet optimizing for VLIW-4 or
b) the code has a high number of instructions executing in the former t-slot, which now require 3 slots (transcendentals, mostly)
I think the driver compiler output for this shader for VLIW-5 and VLIW-4, respectively, would be interesting.
I was thinking the same thing just a few hours ago. Perhaps before, it was clear that AMD was aiming for that sweet spot without trying to best NVIDIA's top GPU. But this time, coming so close, it seems like they just missed the jackpot.
Of course, in reality, 95% of the GTX 580's performance for something like 75% the power and 71% the die size would still be a clear win.
Some prices...
Sapphire HD 6970 2GB - $448
http://www.shopblt.com/cgi-bin/shop...110040015012_BDY1568P.shtml&order_id=!ORDERID
Sapphire HD 6950 2GB - $352
http://www.shopblt.com/cgi-bin/shop...110040015012_BDY1569P.shtml&order_id=!ORDERID
As a comparison:
- an HD6870 costs $272 on this site
- GTX580 start at $534, and GTX570 start at $374.
Sapphire HD 6970 2GB - $448
http://www.shopblt.com/cgi-bin/shop...110040015012_BDY1568P.shtml&order_id=!ORDERID
Sapphire HD 6950 2GB - $352
http://www.shopblt.com/cgi-bin/shop...110040015012_BDY1569P.shtml&order_id=!ORDERID
As a comparison:
- an HD6870 costs $272 on this site
- GTX580 start at $534, and GTX570 start at $374.
Hmm, too low I think.
Considered GTX580 prices, HD6970 seems fairly priced and also preferable, if it could deliver a comparable performance (maybe 5-10% below) at almost $100 less.
HD6950 seems a bit overpriced, if compared to the HD6870. Maybe a 20$ decrease will help. At those prices and with those projected performances, GTX570 should be preferable over HD6950, but the HD6970 will be preferable over both the GTX570 and GTX580.
Babel-17
Veteran
http://www.hardwareluxx.de/communit...-daten-update-2-a-768017-54.html#post15873412
Am I reading the results wrong or am I correct in thinking that is very promising? If true, of course.
Using google translate I see that the settings are being questioned. Would they matter for measuring improvements of a 6970 over a 5870?
Am I reading the results wrong or am I correct in thinking that is very promising? If true, of course.
Using google translate I see that the settings are being questioned. Would they matter for measuring improvements of a 6970 over a 5870?
Babel-17
Veteran
Hope I'm not being dense, I don't have that game, but does this mean the 5870 was using 4x MSAA while the 6970 wasn't?
I don't quite have the same sentiments. First, the GTX 580 is certainly overpriced compared to GTX 570, but it's the halo part so this is to be expected. The price difference between HD 6950 and HD 6970 would be a lot smaller than between GTX 580 and GTX 570 with a (presumably) similar performance difference. Also, for the HD6950 in comparison to HD 6870, performance might not be that much higher (but actually still not really too much out of line with the price difference probably), but don't forget it has twice the memory - I think that's quite a strong argument people are willing to pay more for, if it helps performance or not. This is also true for a comparison to GTX 570 - with maybe very slightly lower performance but more memory with a price slightly below a GTX 570 this looks fairly good to me.
From that pricing, it actually looks like there is enough space for a 1GB HD6950 (with more than a 20$ discount).
I think that's the best explanation for why 3dMark scores and other optimized-for-VLIW-5 games are underwhelming, but in other things 6970 seems to beat the 5870 resoundingly (stalker for one)
Similar threads
