AMD Radeon HD6950/6970(Cayman) Reviews

[VoodooProphetII]

Overclock3D HD6970 Crossfire Review

Same as other review pointed out, the HD 6970 performs great in CrossFire.

In this review the the HD 6970 CF competed well against the GTX 580 SLI in most games, specially in the average and minimum frames, and winning in the Crysis Warhead.

I was surprised the HD HD 6970 CF was faster than GTX 580 SLI in Unigine Heaven 2.1

The GTX 580 SLI cost $300 more, $740 vs $1040 (40% more expensive). Guess what I am going to buy.

Not really, I'm waiting for the HD 6990, hopefully close in performance but even less expensive. Just hope it's not going to be too much gutted.

EDIT
The calculation was done, based on today Newegg lowest prices.

Do you know if tessellation was set to extreme or moderate on the heaven 2.1 bench. From what I saw in reviews the 69XX series drop as soon as the tesselation is set to extreme. But it is nice that CFX scale that well.

[Motiv]

Why is Hawx extremely bad for AMD. every test I see it's performing incredibly bad.

[Lanek]

The game have been made in collaboration with Nvidia ( same as LP2 ), let say they have code it for don't look good on AMD cards .

[Pantsu]

Done with Nvidia hardware for Nvidia hardware. They probably just checked if it would work on AMD cards and that's about it, no optimization or anything. This applies to stuff like HAWX2 and LP2, both use DX11 features applied by Nvidia. Sucks for those who like these games and use AMD. Personally neither game interests me. They do give a nice perf boost to aggregate fps results in reviews though.

[Lanek]

Not like i care, but yes, i still ask me why some use those benchmark / games on their review, specially after have made some articles about their concern on the results with them, put it on a GTX5xx review for compare with old Nvidia gpu's, why not, but otherwise, what a waste of time.

[FlawleZ]

Why is Hawx extremely bad for AMD. every test I see it's performing incredibly bad.

It's like Farcry 2. The performance for Farcry 2 is AWFUL for ATI cards.

[Stevethegreat]

I miss the days when new generation of hardware meant 100% percent update. It seems GPUs finally became like CPUs, incremental updates with each generation that only few (enthusiasts) can appreciate; the rest will be either scummed into buying into this new stuff or won't care at all. My one year old 5970 is still the fastest card out there, it's almost sad... It seems we're entering an end of an era.

[flippin_waffles]

I miss the days when new generation of hardware meant 100% percent update. It seems GPUs finally became like CPUs, incremental updates with each generation that only few (enthusiasts) can appreciate; the rest will be either scummed into buying into this new stuff or won't care at all. My one year old 5970 is still the fastest card out there, it's almost sad... It seems we're entering an end of an era.

Yeah but the blame can't be put entirely on hardware designers. I mean there's only so much you can do when you are up against electrical and physical boundries. Can't expect miracles time after time. Especially when redesigning an architecture 1 year ago from a shrink to the same node. I think the vast majority of fail is on spoiled end users thinking hardware makers OWE them something.

[Nintendork]

Done with Nvidia hardware for Nvidia hardware. They probably just checked if it would work on AMD cards and that's about it, no optimization or anything. This applies to stuff like HAWX2 and LP2, both use DX11 features applied by Nvidia. Sucks for those who like these games and use AMD. Personally neither game interests me. They do give a nice perf boost to aggregate fps results in reviews though.

That's the sad thing, they include those games in some overall perfmonace charts, so the AMD gpu's appear to be slower in the average.

Hawx2 is an awful game (by review sites, consoles included). And lets not forget sites like guru3d, pcgameshardware, computerbase, hexus and others who comply with nvidia's "reviewers guide", and set catalyst IQ to high instead of default. In some games with 5-10% less fps without noticeable quality gains.

Even guru3d had an article prior to their "updated" version where they even zoomed the image with photoshop and stated that even at that scenario is almost impossible to notice quality loss with the default setting (which enable performance optimizations, same as nvidia default wich you can't touch -or you can?-).

[Tim]

It's like Farcry 2. The performance for Farcry 2 is AWFUL for ATI cards.

Used to be great for AMD wasn't it?

[Stevethegreat]

Yeah but the blame can't be put entirely on hardware designers. I mean there's only so much you can do when you are up against electrical and physical boundries. Can't expect miracles time after time. Especially when redesigning an architecture 1 year ago from a shrink to the same node. I think the vast majority of fail is on spoiled end users thinking hardware makers OWE them something.

More performance is not needed anyway since consoles call the shots on the video-game industry. What's sad about this is that techno-evangelists like Kurzweil and co are proved wrong. The future won't be much more advanced than the present, rapid development can only last so much and in the computer hardware industry it lasted around 40 years.

That was a good run, but now it's dying...

[Nintendork]

Techpowerup review:
HD5800
9.10
10.3
10.7
10.11
10.12

HD6900
10.11
10.12a
http://www.techpowerup.com/reviews/A...ormance/1.html


Game with clear gains:

[cold2010]

And other games

[tbone8ty]

Yeah I can tell in this review that driver updates for the 6900 will clearly improve the performance alot just like when nvidia worked out the new fermi arch (gtx480)

We will see I expect he next driver revision the 6900 series will see 10% gains

Just look at the stalker clear sky results the 5870 beats a 6970?

[[XC] hipno650]

Not like i care, but yes, i still ask me why some use those benchmark / games on their review, specially after have made some articles about their concern on the results with them, put it on a GTX5xx review for compare with old Nvidia gpu's, why not, but otherwise, what a waste of time.

the reason lots of sites use them in reviews is because lots of people play them. im sure they could use a bunch of random indie games that run pretty even across Nvidia and AMD but nobody plays them...

as much as it sucks that LP2 and HAWX are Nvidia optimized people still play those games... and if your playing LP2 on AMD your just not going to have the best experience. Nvidia overall seems to have allot more consistent performance even in AMD sponsored games. like in LP2 it is so bad when it comes to similar performing cards when I recommend a new video card to someone I ask them if they will be playing LP2. if they are then I will lean towards an Nvidia Card, if not then pretty much whatever is cheapest

[Bigchrome]

From the relative performance graph: GTX 480 faster than GTX580 eh?

To chime in on games being optimized for nvidia... I used to be totally against it and be really angered when new games came out that were twimtbp because I knew they'd suck on my card. I've since realised that nvidia don't have any areas in which they take a massive hit, unlike ati, so I guess, dirty tactics or not, that nvidia is just a better choice. If ati were really concerned about us getting good performance in these types of titles they'd just pay the developers to optimize for them to. It appears as if they are not though, so 2 fingers up to amd for this generation!

Having said that I did nearly get the HD6970 instead of the GTX570 that's going to be my christmas present; just as good performance if not a little better in non-nvidia titles, 2GBs of ram, and eyefinity are all great reasons. The draw of the folding power on nvidia cards was too much for me though. Should amd fix folding performance it will be entirely the better card, but waiting on them to fix things has not proven to be the best strategy in my experience.

[Manicdan]

From the relative performance graph: GTX 480 faster than GTX580 eh?

the graph set all baselines at 100%, and the change is only relevant for that cards transition through new drivers.

[Bigchrome]

the graph set all baselines at 100%, and the change is only relevant for that cards transition through new drivers.

You think I would have picked that up! Sorry, still on my first cup of coffee today! I guess if I looked at the 4870 result I could have guessed what was going on

Makes me wonder whether those ati cards will actually get the performance boost people are talking about though, and whether nvidia has changed the architechture from 4xx to 5xx significantly enough that there is still performance to be eked from them with new drivers?

[omega1alpha]

Customer service with HIS in North America is worse than bad ( they outsource support for their products in NA and the company doing so is ran by useless people who would make subpar crash test dummies ). I have nothing good to say about their company. Don't waste your time and money with those jokers. Hint: I don't like them.

If HIS and Powercolor are your two options, take the Powercolor ( at least they will respond to their customers in some capacity )

I second that! Avoid HIS unless you can get the card at a SIGNIFICANT discount. I've heard from multiple people that their customer service is horrible. So, you better hope you never have an issue with their cards. I once called them out of the blue to see if I could get a hold of someone and I could never get through on multiple attempts. Plus, the voice mail system was barely in english in the USA.

[Shadov]

the reason lots of sites use them in reviews is because lots of people play them. im sure they could use a bunch of random indie games that run pretty even across Nvidia and AMD but nobody plays them...

as much as it sucks that LP2 and HAWX are Nvidia optimized people still play those games... and if your playing LP2 on AMD your just not going to have the best experience. Nvidia overall seems to have allot more consistent performance even in AMD sponsored games. like in LP2 it is so bad when it comes to similar performing cards when I recommend a new video card to someone I ask them if they will be playing LP2. if they are then I will lean towards an Nvidia Card, if not then pretty much whatever is cheapest

Depends on the approach I guess and as far as I remember AMD said in an interview that they will never de optimize games for their competitor.

That way proprietary standards like PhysX are bad, since you can have normal physics in a game or Nvidia can add a bit of green fairy dust (also known as money) to a title in production and suddenly normal features like e.g. fog are implemented in a way that will work only on NV cards.

You can call me biased, but I prefer to have standard features available to everyone.

[Manicdan]

You think I would have picked that up! Sorry, still on my first cup of coffee today! I guess if I looked at the 4870 result I could have guessed what was going on

Makes me wonder whether those ati cards will actually get the performance boost people are talking about though, and whether nvidia has changed the architechture from 4xx to 5xx significantly enough that there is still performance to be eked from them with new drivers?

why does my memory think 9.4 wasnt the first driver released for the 4870? so i would have liked to see the first 2-3 months the card was released, then the newest one so that it covers the entire history. the 285 and 4870 show like no gains, while the 5870 and 480 have clear gains, and i wonder if thats just because we dont see the first driver releases.

[Klarko]

I have a HIS card and have never had any problems with it itself, I usually think videocards are pretty stable unless you OC it to the maximum for long periods lol

[Oztopher]

I get some graphical glitching on metro, mostly when outside. It's something to do with the shadow/lighting system, it's all over the place. When i move the camera around the shadows move too, consistently.

Apart from that it looks and plays great with very high setting. I don't bother playing in DX11, (i know there is) but i see no real visual difference and FPS is quite a bit lower compared to DX10. That's with just tessellation enabled, with advanced DoF it halves my framerate.

Object motion blur on the very high preset is an absolute must

[zerazax]

why does my memory think 9.4 wasnt the first driver released for the 4870? so i would have liked to see the first 2-3 months the card was released, then the newest one so that it covers the entire history. the 285 and 4870 show like no gains, while the 5870 and 480 have clear gains, and i wonder if thats just because we dont see the first driver releases.

9.4 came out nearly a year after the 4870 was released..

[Manicdan]

9.4 came out nearly a year after the 4870 was released..

see, i didnt black out all of 2008, just most of it....

which is why i would have liked to see them use 8.6 or whichever it was that came on the disks compared to current. see if its more than 1% between first and newest