AMD Radeon HD6950/6970(Cayman) Reviews

[TheBreezyBB]

Nope, only in states they have a physical presence in that also has a sales tax.

Well, NY does.

[Pantsu]

While Nvidia's 500 series might be just a bug fix for Fermi, Cayman is simply a letdown in terms of raw performance increase. The new architecture obviously isn't all that effiecient yet. It does fix the problems with tessellation and AA-performance, though it still doesn't beat fermi in either area. There might be some promise for the next gen and with future driver updates, but that's just speculation.

In the end this release means Nvidia doesn't have to lower prices at all, but on the other hand they do have to compete with 6970 vs. GTX 570. It's just sad to see 5870 being so close to these cards, there's been hardly any improvement since last year. In my books the whole fermi fiasco and these new slightly updated cards are about on the same level as the GeForce 9000-series update:

[hurrdurr]

Graphics power increase relies heavily upon shrinking nodes. Considering these are still at 40nm, I wouldn't say that this year's increases have been very disappointing.

[dartaz]

It is mentionned countless of times, but go read a review that uses the 10.11 drivers which are available.... Anandtech used the 10.10. Hardocp for example, the 6950 NEVER looses to a 570GTX.. the 6970 trade some blows with the GTX580 but is overall slightly slower. Still blows the 570GTX by a significant margin (e.g.10%).

My friend, Did you notice that Hardocp was using different settings when they compared GTX570/580 to 6950/6970 ? Did you ?

Example, Look here they are using 8x AA for GTX 570, but only 4x AA for HD6950 in Civilization




Thats not a fair comparison. If they were using the same settings, then GTX 570 would win easily

[Kristers Bensin]

The 6970 card runs slower than the GTX 580 and runs hotter than the GTX 580? WTF?

ATI really messed up here.

Not according to this site.

[Motiv]

Not according to this site.

Looks like it may be running 'hotter' but using that as a benchmark is struggling to justify your point (almost a strawman argument).

It looks like it does run 'hotter' but it is quieter and has less of a power draw. 'Hotter' is a bit of a sham argument really, as other factors are in effect, and the temp difference seems to be within a few degrees C.

[Pantsu]

Graphics power increase relies heavily upon shrinking nodes. Considering these are still at 40nm, I wouldn't say that this year's increases have been very disappointing.

Ultimately the fault for the current situation lies with TSMC for epicly failing 40nm, and skipping 32nm. Also Nvidia failed to see that coming and the result was Fermi half a year late and neutered. If Nvidia had managed to release a full Fermi at the same time as 5800-series, the current situation would be much much better for the consumer. The performance would be better and the prices would be lower.

So better pray TSMC knows what it's doing with 28 nm.

[Nintendork]

My friend, Did you notice that Hardocp was using different settings when they compared GTX570/580 to 6950/6970 ? Did you ?

Example, Look here they are using 8x AA for GTX 570, but only 4x AA for HD6950 in Civilization




Thats not a fair comparison. If they were using the same settings, then GTX 570 would win easily

Look at "apples to apples"



Tried to look at the other games of that review?

[takamishanoku]

Not bad cards but looks like Nvidia really snuck one up on ATI and us with the new 500 series. Had they not been released (which i think we can safely say wasn't expected) ATI would have been sitting pretty. As it stands performance wise the 500 series is better and runs cooler, less power. WTF how did the roles reverse in the space of 1 month???

[Olivon]

Someone please explain?!!

WTF AMAZON 6970=483$
Amazon 6950=372$

AMD sent new prices at last time so several shops don't have the good price.

HD6X00 series is a disappointment IMHO.

I expected many more from AMD

[zalbard]

The performance is all over the place.
It performs much better than 5870 in some titles; in some titles it's barely faster. Why? Most likely drivers.
Some of the numbers are quite promising.
But yeah, I agree, quite a raw product so far.

[Lokinhow]

good cards, performon well, nice features (dual bios, power tune, mlaa), not very hot or loudy and have a good price.
but, I think almost everybody expected more, it really looks like AMD didn`t expected GTX580 and GTX570.
People got too much hype over those cards, and now it looks disappointing. But in anyway they are bad cards, AMD did a nice work.

Now lets hope newer drivers brings at least 5~7% improvement and fix some games where those cards performs strangely slow.

Oh, and HD6950 in crossfire looks like a very good deal.

[hurrdurr]

It would be very stupid to buy a card ASSUMING that it's going to get much better with drivers. For one thing, remember X2900XT. Second, if you're relying on drivers because you think the performance varying wildly means "lack of driver compatibility," you have to take into account that there has been a lot of architectural changes in Cayman which would favor some games and dislike other games. Much like Fermi - its performance varied a lot from game to game compared to other cards.

[flyck]

My friend, Did you notice that Hardocp was using different settings when they compared GTX570/580 to 6950/6970 ? Did you ?

Example, Look here they are using 8x AA for GTX 570, but only 4x AA for HD6950 in Civilization

Thats not a fair comparison. If they were using the same settings, then GTX 570 would win easily

see comment Nintendork..

and i was wrong about anandtech, they either changed the text or i remembered it wrong.

Altough bothered by anantech showing all the cards and crossfire setups when the resolution is 25.... and for resolutions 16xxx they show only the single cards. So if you look closely the 25 the 6970 is between the 570 and 580 and the 6950 is continously around the 570. for 1920 the 570-580 takes the advantage and 16xx the difference is completely in nvidia favour.

HardOcp tested the best image quality settings possible while still playable and did an apples to apples which were alway high res and qualitity settings.

Anand showed that only the first graph and afterwards they lower the settings and quality and show them single card vs single card. making the clearest chart about performance also the bottom one and the one with the lowest quality settings (settings you won't use when you buy a card of >300euro)

[dartaz]

I'm not sure if hardocp is trustworthy

I mean techpowerup did use Catalyst 10.11, and results was so different. GTX 570 was clear winner against HD6950 in techpowerup review

[tajoh111]

It is mentionned countless of times, but go read a review that uses the 10.11 drivers which are available.... Anandtech used the 10.10. Hardocp for example, the 6950 NEVER looses to a 570GTX.. the 6970 trade some blows with the GTX580 but is overall slightly slower. Still blows the 570GTX by a significant margin (e.g.10%).

Correction

Anand used 10.11

AMD Catalyst 8.79.6.2RC2 = 10.11 RC2

http://www.anandtech.com/show/4061/a...eon-hd-6950/12

HARDOCP is the only website showing such results out of all the website. Hardocp has never been the most credible website.

[Olivon]

[Dami3n]

IMHO this new cards are a

[flyck]

Correction

Anand used 10.11

AMD Catalyst 8.79.6.2RC2 = 10.11 RC2

http://www.anandtech.com/show/4061/a...eon-hd-6950/12

HARDOCP is the only website showing such results out of all the website. Hardocp has never been the most credible website.

I already corrected my statement, see previous comment of mine, two posts before yours. and no, anandtech shows the same thing, again read my previous comment. The difference is an nvidia advantage for low resolutions, but if you turn those up, see first graph for anand viewers, after some searching in there graph you will see 6970> 570 and 6950~570

[Motiv]

Correction

Anand used 10.11

AMD Catalyst 8.79.6.2RC2 = 10.11 RC2

http://www.anandtech.com/show/4061/a...eon-hd-6950/12

HARDOCP is the only website showing such results out of all the website. Hardocp has never been the most credible website.

Strange, I always like HardOCP, and find they give very good reviews. They show what settings, give equal types of performance. It's a different way of doing things, but still gives a fair comparison.

[Milos-stancene]

If I could sum up the 6970 in one word it would be "disappointing". And I do not understand the marketing logic of producing what is for the moment the "top" AMD card when it can only perform around the level of Nvidia's number two card, the GTX 570. OK, so it is priced at a similar level to the GTX 570. But this only confirms the reality, that GTX 580 levels of performance, from a single GPU at least, are no longer within AMD's capabilities.

[Nintendork]

good cards, performon well, nice features (dual bios, power tune, mlaa), not very hot or loudy and have a good price.
but, I think almost everybody expected more, it really looks like AMD didn`t expected GTX580 and GTX570.
People got too much hype over those cards, and now it looks disappointing. But in anyway they are bad cards, AMD did a nice work.

Now lets hope newer drivers brings at least 5~7% improvement and fix some games where those cards performs strangely slow.

Oh, and HD6950 in crossfire looks like a very good deal.

Those games share similarities with the VLIW5/4 thing in 3dmark. That games actually don't seem to adress the total capabilities of the gpu.

For example barts vs cypress is a very constant comparison without surprises.


11.1 is supposed to be made from the scratch right?

[Jodiuh]

I just figured 3870 to 4870...big jump...4870 to 5870...big jump...5870 to 6870...wait, huh?

/reboot

5870 to 6970...wait, huh?

So performance's less than what we'd all come to expect from a Radeon. But price doesn't line up either.

[flopper]

Hardocp (Kyle the bar bouncer) has a no bull approach.
if he approves, you know its good stuff.

[flyck]

I just figured 3870 to 4870...big jump...4870 to 5870...big jump...5870 to 6870...wait, huh? reboot 5870 to 6970...wait, huh?

yeah only 30% on the same node. How terrible!