• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

ATI cuts 6950 allocation

Very smug feeling coming from some of the pro AMD forums, SA etc and I think there might be more to this driver story.
Theres certainly something strange going on here, either it's vastly inefficient and poo or what we are seeing atm is not the true performance of this card.
I still have some doubts this will be a 580 'killer' but I'm certain it should be trading blows with it. edit: in some games.
 
Last edited:
Do the 10.11 drivers officially support the HD6900 series??

If not then all of these benchmarks are utterly pointless TBH.

Considering that the shaders are completely different from the HD2000,HD3000,HD4000,HD5000 and HD6800 series then drivers could make a huge amount of difference IMHO.
 
The retail cards ship with an updated 10.11 to support the 6970 cards, check the reviews out on the 15th and see if they are using the same version as the Germans posting the 6970 benchmarks up.
 
Think about this....
Somewhere in the mass of theads and replies Gibbo asked how he should convey that an unknown new card just about to be released had 1536 sp's of a new type that would be equal to 1920 sp's of the old type on the old card.
Question:
Would a 1920sp 5870 with 6870 improvements and higher clocks be only 10% faster than a std 5870?
 
If you look at the Heaven benchmarks the 6970 is showing over 50% gains over a 5870 and only 6% slower than fermi the master of tessellation, so drivers seem to be working OK.

 
Last edited:
50% is only in one benchmark, thats not really representitive of the overall picture. In most cases its a very small increase in performance which I have seen in some driver updates in the past (perhaps a few % more).
Unless AMD have only improved their hardware for Heaven then it doesn't add up.
 
Last edited:
The ballad of Ducati750SS:

  • He had 8GB ram but issues caused him to change it to 4GB (different makes and latency speeds)
  • When asked about the switch and what documents came with it: "There is only a quick install sheet and a driver CD with instructions for up to 6870"
  • 14 minutes between posting that he is going to install the 480 again and the 3dmark results being uploaded. (plausible time to swap drivers over cleanly)
  • 8.790.6.0 (Catalyst 10.11) Win7 64 <--- Original GPU-Z
  • 8.790.6.2000 (Catalyst 10.11) Win7 64 <---GPU-Z updated next day but same drivers
  • Confirms 100% GPU load during testing and points the finger at the the drivers
  • GTX480 = 324W / 6970 = 331W
  • He thinks the switch is for crossfire and says it did not improve performance in the 2 position.
  • He did not have the beta CCC with the latest settings.
  • CCC settings are all default
  • Mentions bad ram and old drivers but this might be related to initial runs which slightly improved after the ram change and min fps results where better.
  • He never took the card apart.
  • Gave the card to Harry97 who used the same drivers.
  • Someone suggests to Harry97 a chipset driver related to the GPU and the extreme 3dmark11 he score goes up by 10 which does not seem too related.

Is the new beta CCC a stand alone install or does it come with 10.12 drivers?
 
Last edited:
^^ When you put it that way it does look kind of dodgy, come the 15th hopefully we will see some vastly improved results.

I never knew he gave the card to someone else so this could very well be a dodgy card and or dodgy bios.
 
FrenchTart said:
On the weight of the evidence thus far it's only natural. What else would you expect?
Click to expand...

I'm just curious as to why he's not dismissing it so far considering how convinced he was previously. I'm not convinced based on the fact that it makes no sense to release a 6970 that's barely faster than a 5870, while taking in to consideration how the 6870 performs compared to the 5870 as well.
 
kylew said:
I'm just curious as to why he's not dismissing it so far considering how convinced he was previously. I'm not convinced based on the fact that it makes no sense to release a 6970 that's barely faster than a 5870, while taking in to consideration how the 6870 performs compared to the 5870 as well.
Click to expand...

I'm not convinced of the results we've seen so far either. I'm still hoping for something that would make me feel able to use my eyefinity setup properly again.
 
The results for the 6970 in Metro 2033 show more than a 100% improvement over the 5870, when tessellation and DOF are enabled:

1920*1080, DX11, 4xAA, 16xAF, Tessellation enabled, DOF enabled:

5870: min=5.50fps, av=11.02
6970: min=8.28fps, av=23.40

links: 5850, 6970

This was always a configuration that Cypress really struggled with, so it's good to see a massive improvement with Cayman.
 
Duff-Man said:
The results for the 6970 in Metro 2033 show more than a 100% improvement over the 5870, when tessellation and DOF are enabled:

1920*1080, DX11, 4xAA, 16xAF, Tessellation enabled, DOF enabled:

5870: min=5.50fps, av=11.02
6970: min=8.28fps, av=23.40

links: 5850, 6970

This was always a configuration that Cypress really struggled with, so it's good to see a massive improvement with Cayman.
Click to expand...

Yeah, its now completely playable :p

http://www.semiaccurate.com/forums/showpost.php?p=88102&postcount=1918

Updated table with more results, with more 55-65% performance increases, though only really two "main" games, but also the game/benchmark you might have expected AMD to be working for drivers from, as Stalker is DX11 as implemented and helped by AMD and Uniengine is the tesselation benchmark so far. I'm not saying they'll optimise for them alone I'm saying over several months a team works on drivers, benchmarking every game constantly with every change is a bit silly, pick and choose and work out where the drivers need work in one game.

The results, if remotely true, would suggest what looks about right, a circa 71% increase in shader power over a barts, and the usual doubling shaders != doubling performance, but 80%, 90% if really fantastically well done.

My guess is the performance is there, it just needs to access it and a final release driver will probably go a LONG way to achieving that.


EDIT:-

http://www.semiaccurate.com/forums/showpost.php?p=88125&postcount=1927

more added, it looks like Metro 2033 is 170% ahead in max fps ;)

DIrt 2 is always a bit weird and seems to run into a bit of a cpu limit, is that an AMD driver thing? minimum is up 43% though.
 
Last edited:
Well if I was AMD I would be spitting chunks at these leaks so far as they arent painting a good picture in general.

I imagine they cant see Wednesday come quick enough.
 
Compared to the 5870:

54% improvement in stalker COP
8.1% improvement in RE5
5.9% in AvP

Also, around 12% in each of the 3D mark tests, and around 60% in the tessellation-heavy Heaven benchmark.

To say that the 6870 is 5% or so behind a 5870, it's not looking to be a great performance improvement for the "71%" increase in shader power. Only F1 2010 and Stalker COP have shown performance that begins to approach that level (assuming the ~80% scaling with shader power to performance, which is a fair baseline). Tessellation performance is greatly improved, but this was fully expected as Barts also shows large gains over cypress.
 
You must log in or register to reply here.
Back
Top Bottom