6970/6950 Video Card Review thread

[EastCoasthandle]

AMD's Cayman GPU Architecture
Anandtech Catalyst 8.79.6.2 RC2
Benchmark Review Catalyst 10.11 (8.790.6.2000)
Computer Base Catalyst 8.79.6.2 RC2
Guru3D Catalyst Beta 8.97.6.101206a
Hardware Canucks Catalyst 10.12 Beta RC2
Hardware.Fr Catalyst 8.79.6.2RC2 of CAP1Catalyst10.12
Hardware Heaven Catalyst 10.12 ??
Hardware.Info Pre-released Cat 10.12 (no version number provided)
HardOPC Catalyst 8.79.6.2 RC2
Hot Hardware Catalyst 10.11b
IXBT Catalyst 10.11 (8.790.6.2000)
PC In Life Catalyst 10.11 (8.790.6.2000)
PC Perspective which includes 5870 2Gig & CF (8.790.6.2000)
Rage3D Catalyst 8.890 + CAP 10.12. Also reviews the performance difference between MSAA, MLAA and EQAA
TPU 6970 Catalyst 8.79.6.2 RC2
TPU 6950 Catalyst 8.79.6.2 RC2
Toms Hardware (including CF results) Catalyst 8.79.6.2 RC2
TweakTown Catalyst 10.10 beta
vrzone Catalyst 10.11 (8.790.6.2000)
Wide Screen Gaming Forum using both widescreen and Eyefinity. Using pre-released drivers??

Post will be updated once more reviews are available.

Note:
Earlier drivers don't seem to allow for any control of PowerTune from CCC. It appears that driver support for that seem to start with Catalyst 8.79.6.2 RC2

Edit 2:
2 Observations so far.
1. EQAA seems to be AMD's version of CSAA. So IMO would be worth using over MSAA for a performance boost.
2. Powertuner doesn't seem to help at stock clocks. I'm guessing there maybe something there if the card is over clocked but I've not come across any reviews showing this yet.

[purecain]

nice collection of reviews... thanks for the effort...

@eastcoasthandle-are you going to drop a few quid on a 6970, i couldnt help myself and ordered one at stupid o clock this morning...
if not i've got a 5870 on ebay atm if you fancy another...

[EastCoasthandle]

I'm still going over everything right now. Below are a list of features that are available with the card.


This switch allows for a second nonprogrammable bios.






What's interesting is the use of 2xEQAA for gaming should give you AA quality close to 4xMSAA but with a performance benefit. (taken from one of the reviews)





The use of EQAA is compared to MSAA. (taken from one of the reviews)




Power Tune is a new functioned added with CCC (not the switch on the card). This feature (as written in the slide) seems to be more apt when overclocking (when using the + aspect of Power Tune). Not for stock clocks. But the reviews I've read so far haven't explored using this feature when overclocking the card.

This is an algorithm that calculates an infered power draw based on execution unit activity. Voltage and Leakage could be a variable to that algorithm, but for the sake of providing a consistent default product performance they are input as a constant per SKU, based of the worst case for that SKU. Tempurature variablity presents more challenges, though it is something we are looking at in terms of enhancing the capabilties for the future.

source



Edit:
Check out Sapphire's Website for amd_radeon_hd_6900_series_win7_vista_dec8 drivers.


Edit 2:

Interesting to note, I specifically asked if we should use the Cat 10.12 Preview driver press was given, to test 6900 series. The response was that the Cat 10.12 Preview driver did not support the 6900 series we were evaluating. So I don't know how other sites are claiming to use Cat 10.12 for the reviews. We used the driver provided by AMD for 6900 testing, which as you have pointed out, is posted on Page 2 of our review, and is based on Cat 10.11 code. We used this same driver also on the 6870 and 5870, so all cards are using the same driver.

source

[purecain]

very interesting info... i've been reading about voltage adustment... looks like powertune is going to need modding and the real excitement is seeing the improvement above and beyond tdp.... interesting...

hopefully my card will arrive in the am tommorow, although thats an unlikely prospect...

as soon as it arrives i'll start benching.... cant wait... first bit of new tech in ages...

[EastCoasthandle]

Originally Posted by purecain

very interesting info... i've been reading about voltage adustment... looks like powertune is going to need modding and the real excitement is seeing the improvement above and beyond tdp.... interesting...

hopefully my card will arrive in the am tommorow, although thats an unlikely prospect...

as soon as it arrives i'll start benching.... cant wait... first bit of new tech in ages...

Look forward to your findings. I'm interested to know if at +10 (according to w1z's review that's all you need) if overclocking actually improves or not in games. Versus keeping power tune at 0.

------
It's disappointing that AMD has not released a hotfix driver allowing for EQAA and MLAA for 5000 series. I thought perhaps today another drivers would be release. However, apparently they had no such thing planned.

[LAN_deRf_HA]

Odd, all the sites that were doing wow benches stopped. Even tpu did the wow DX11 bench as recently as the 580 review.

[purecain]

i like the fact that even though there are less shaders than the 5870, the 6970 has more complex stream shaders

quote from wizards review...
Compared to the older VLIW5 design in which an SIMD core consisted of four simple and one complex stream processors with some common resources, the new design, dubbed VLIW4, combines four equally-capable complex stream processors, with two of the four getting special functions.

[mastrdrver]

Missing a few of the more technical reviews imo.

PCPer
Tech Report
especially David Kanter over on Real World Tech

Fwiw Wizz is the only one saying 2 of 4. All other reviews I've read state 3 of 4 and this seems obvious to me in the slides.

[EastCoasthandle]

I didn't include tech report after reading this:

I suppose one could also overclock the snot out of the thing and plunge the PowerTune slider to negative 20% just to create confusion about how the card will perform in any given situation. Whee!

source

That comment was so far removed from the review itself that I draw pause and decided to just omit it for the time being. The other's I've not come across yet.

[Play3r]

i think we will be looking at a new release of graphics cards in around 6 months or whenever TMSC gets the act together and works out the 28nm or 32nm.

Hopefully AMD can get out some really good drivers early that will make use of the changes they have made.

Only time will tell.

[mastrdrver]

Originally Posted by EastCoasthandle

I didn't include tech report after reading this:

source

That comment was so far removed from the review itself that I draw pause and decided to just omit it for the time being. The other's I've not come across yet.

I don't see how. He makes a very good point that unless sites (or anyone who posts reviews) state what they are setting it to, it just becomes another way to skew bench results does it not?

Let me put it another way: I can see problems when people start overvolting their cards and even with it set to 20% max, run in to clock gating when playing a game. This will be especially interesting to see when the 6990 comes as the 300w PCIe limit is going to be obliterated with 900Mhz cores. Will 20% be enough? Will there be a higher limit then 20%? Is there any kind of work around?

[Xaser04]

Please also add the Bit-tech reviews:

HD6950
HD6970

Benchmarks are lacking (something I have been quite vocal about) but the architecture analysis is there.

[EastCoasthandle]

Originally Posted by mastrdrver

I don't see how. He makes a very good point that unless sites (or anyone who posts reviews) state what they are setting it to, it just becomes another way to skew bench results does it not?

Let me put it another way: I can see problems when people start overvolting their cards and even with it set to 20% max, run in to clock gating when playing a game. This will be especially interesting to see when the 6990 comes as the 300w PCIe limit is going to be obliterated with 900Mhz cores. Will 20% be enough? Will there be a higher limit then 20%? Is there any kind of work around?

Sorry, but if you believe that is what he was implying then it should have been stated that way. However, that's not the case. So for now I'm not including it.

Originally Posted by Xaser04

Please also add the Bit-tech reviews:

HD6950
HD6970

Benchmarks are lacking (something I have been quite vocal about) but the architecture analysis is there.

Hmm, bit-tech's reviews don't offer any information as to which drivers they are using. As you can see from the OP many are using various drivers

[LAN_deRf_HA]

Hardware.Fr links to the wrong review.

[TotalChaos]

Well I just took the plunge, I couldn't resist and ordered a 6950 with express shipping so here is to hoping I get it tomorrow and not Monday

[EastCoasthandle]

Hotfix 10.12a seems to be improving performance for hte 6900 series (only)

Seems 10.12a improved 3DMark11 performance for me.

10.12 (8.801) - http://3dmark.com/3dm11/197922

10.12a (8.790.6.2) - http://3dmark.com/3dm11/202116

GPU score, 4938 vs 5137.

6950

3dmark 11

driver that came with the disc
p4881

this driver
p5034

think thats a fair improvement, wish afterburner supported voltage controll on my card..

6950

[mastrdrver]

Originally Posted by EastCoasthandle

Sorry, but if you believe that is what he was implying then it should have been stated that way. However, that's not the case. So for now I'm not including it.

Maybe he was imply this by that statement (Though I agree that if it was, he should have been more clear. Maybe some more time and another article will allow him to expand on that comment. This is Scott Wasson (not THG if you'll excuse the analogy) and I'd be surprised if it was just an off hand comment. Idk, maybe he just had a bad week..............but anyway.)

I'm really confused, slides states "typical gaming power" and "power tune maximum power".
If "power tune maximum power" refers to the default TDP (and not the maximum +20%) they have been very loose with the settings, 60w more for "only powerviruses" doesn't sound too good if the usefulness of this feature is to reduce this gap.
Also with +20% of 190w 6950 would pass the limit of 225w of 6+6 pin, it can't be good either to stay in those limits.

For all I know the TDP could be something like 210w, max 250w with power tune and 190w typical usage with games, that bodes better with Anandtech power consumption numbers, only an increase of 21w with furmark, if not the gtx570 consumes in that review 250w playing Crysis, gtx580 consumes 279w, gtx480 310w (with furmark 368w) and 5870 in furmark 264w and we know this is not real compared to other reviews. Things are not very clear or well explained.

[nt300]

Now this is interesting. Not sure if there's any truth to this but still worth the look

The secret of the AMD 6900's second switch
The second switch itself is cleverly hidden on the back of the card, and you can see it is tiny and almost unnoticeable. It measures a mere 2.5 inches in length by just over an inch in height, and it is cleverly camouflaged against the black back plate. Most people missed it in the pictures, but if you know what to look for, you can spot it quite easily.

What does it do? Easy enough, it swaps the shader count. The 'normal' count is 1920 for a 205W 6970 card, and the switch opens up the full force of all 2520 shaders. The reason this switch is set to 'low' is that between the clock speed advances, 730MHz/1920 shaders to 920MHz/2520 shaders, it blows the card through the 300W cap for PCIe cards.

This switch is an elegant way of going from sub-GTX570 performance to way way way over GTX580 performance, and still getting PCIe certification. Very slick guys, well done.If you don't have the right drivers though, the switch does basically nothing, which is why those who bough the card retail haven't seen the results they think they should get.

http://www.semiaccurate.com/2010/12/...second-switch/

[EastCoasthandle]

Originally Posted by mastrdrver

Maybe he was imply this by that statement (Though I agree that if it was, he should have been more clear. Maybe some more time and another article will allow him to expand on that comment. This is Scott Wasson (not THG if you'll excuse the analogy) and I'd be surprised if it was just an off hand comment. Idk, maybe he just had a bad week..............but anyway.)

Here is a good way to get an idea of what the slider actually does:
source

I did a quick test of PoweTune in 3 apps. so far. Sorry for lack of power measurements but my Kill-a-Watt died.
All tests done on Win7 x64 with Cat. 10.12a hotfix RC3 and Phenom X6 4GHz Turbo OFF (890FX)

Unigine Heaven 2.1 1280x720 default settings, tesselation extreme:
-20% = 39FPS
-10% = 39.8FPS
0% = 39.8FPS
+20% =39.8FPS
Only PowerTune -20% affects this setting and causes throttling of GPU.

Unigine Heaven 2.1 1920x1200 default settings, tesselation normal:
-20% = 40.4FPS (72C GPU)
-10% = 49.3FPS (76C GPU)
0% = 49.9FPS (76C GPU)
+20% = 49.9FPS (76C GPU)
Again only -20% makes visible difference here, but -10% also has a tiny effect on performance! (Temps reported at the end of each benchmark).

Unigine Heaven 2.1 1920x1200 default settings, tesselation disabled:
-20% = 55.4FPS (71C GPU)
-10% = 68.4FPS (76C GPU)
0% = 72.6FPS (78C GPU)
+10% = 73.4FPS (78C GPU)
+20% = 73.4FPS (78C GPU)
Here we can see that even 0% caps Cayman a bit.

3DMark Vantage Perlin Noise:
-20% = 101FPS
-10% = 108FPS
0% = 147FPS
+10% = 173FPS
+20% = 173FPS
Heavily affected by PowerTune.

3DMark Vantage Parallax:
-20% = 52FPS
-10% = 74FPS
0% = 74FPS
+10% = 74FPS
+20% = 74FPS

Dirt 2 Ultra settings 1920x1200 noAA:
+20% = 87.8FPS (min. 64.2FPS)
0% = 87.4FPS (min. 64.5FPS)
-10% = 88.0FPS (min. 75.0FPS)
-20% = 73.3FPS (min. 57.3FPS)
Slight variances to be ignored because built in benchmark is not consistent and I've done only 1 run for each setting. Only -20% PowerTune really affects results. Not very demanding game for Cayman in that regard. As speculated by Mintmaster the bottleneck is not shading for sure

Finally I wanted to check if +20% setting has enough margin to accommodate Perlin Noise test with max. CCC GPU clock of 950MHz:
-20% = 102FPS
-10% = 108FPS
0% = 145FPS
+10% = 166FPS
+15% = 187FPS
+20% = 187FPS
Good news is, no throttling at 950MHz even with 'only' +15% setting. What is a bit perplexing though is +10% result. If you look at stock clock counterpart it performs faster! 173FPS(GPU 880MHz) vs 166FPS (GPU950MHz)! Now you might tell that @950MHz GPU is throttling below 880MHz but at least according to GPU-Z monitor that's not the case. It dips to 911MHz but average clock is still 946MHz!? To add to confusion my second run after restarting computer didn't throttle at all and showed constant 950MHz with +10% and exactly the same 166FPS. I can think of two possible explanations:
- PowerTune changes clocks so quickly that GPU-Z can't detect it (but still I think it should show at least few spikes in a log file), or
- PowerTune is switching off some of the units / clocks them slower while maintaining high clock for the rest of GPU

Any ideas?
Under construction ...

This is a very sharp difference between his comment and someone getting down into showing data of what the slider actually does.

[Mussels]

Originally Posted by nt300

Now this is interesting. Not sure if there's any truth to this but still worth the look


http://www.semiaccurate.com/2010/12/...second-switch/

dude its a complete joke/troll, please stop reposting it.


Is the gigantic foam switch held on with duct tape, not clue enough?



let alone this bit at the end

[pantherx12]

Originally Posted by Mussels

Is the gigantic foam switch held on with duct tape, not clue enough?

I think he may of realised that mussles and may of found it amusing himself and perhaps not of realised everyone else ever has seen it.

Way to be really patronising dude

[EastCoasthandle]

hotfix 10.12 was updated. It is called amd_win7_vista_radeon_hd6900_8.79.6.2rc3_dec16.exe making it 2 days old. So click on the link in the news section and download them (d/l it again if you got the previous RC2 version dated Dec 9th). This is for the 6900 series.

[TotalChaos]

thx for the info, gonna grab them since my 6950 will be here Monday. Damn why don't couriers work weekends

[Thatguy]

Originally Posted by TotalChaos

thx for the info, gonna grab them since my 6950 will be here Monday. Damn why don't couriers work weekends

they will, but usually they charge a big fee for doing so.

[Mussels]

Originally Posted by pantherx12

I think he may of realised that mussles and may of found it amusing himself and perhaps not of realised everyone else ever has seen it.

Way to be really patronising dude

i am the patron saint of patronising.