AMD Radeon HD6950/6970(Cayman) Reviews

[dartaz]

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

Not many people play at 2560x1600 which makes the comparison at that resolution useless for most people. But let us calculate the average difference between HD6970 and GTX 570 from anandtech review.

GAME/HD6970/GTX 570/Advantage over GTX570

Crysis/36.6fps/32.6fps/12.2%
Battleforge/48.2fps/54.5fps/-13%
Metro/25.5/23/10%
HAWX/92/104/-13%
Civilization /34.5/45/-24%
BC2/47.8/45/6%
STALKER/39/34.3/13%
DIRT2 /56.8/64.2/-13%
ME2/56.1/55.9/0%
Wolf/79.4/66/20%

Total =(12.2-13+10-13-24+6+13-13+20)/10= -0.18%

If my calculations are accurate then that means that HD6970 is 0.18% slower than GTX 570 at 2560x1600

[zalbard]

11.1 is supposed to be made from the scratch right?

Not from the scratch, but there should be some big changes in 11.1 and 11.2.

[Sn0wm@n]

hardware cannucks did a good one once again

good job skymtl




HD 6950 2GB


totally agree with the conclusion btw!!

now on to remove those ccc oc limit

[Jodiuh]

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

That's great, yeah, but just not what I'd expected.

[flyck]

Not many people play at 2560x1600 which makes the comparison at that resolution useless for most people. But let us calculate the average difference between HD6970 and GTX 570 from anandtech review.

GAME/HD6970/GTX 570/Advantage over GTX570

Crysis/36.6fps/32.6fps/12.2%
Battleforge/48.2fps/54.5fps/-13%
Metro/25.5/23/10%
HAWX/92/104/-13%
Civilization /34.5/45/-24%
BC2/47.8/45/6%
STALKER/39/34.3/13%
DIRT2 /56.8/64.2/-13%
ME2/56.1/55.9/0%
Wolf/79.4/66/20%

Total =(12.2-13+10-13-24+6+13-13+20)/10= -0.18%

If my calculations are accurate then that means that HD6970 is 0.18% slower than GTX 570 at 2560x1600

Thanks for proving my point!

didn't had time to go through them.
The higher the settings, the the closer the gap and how closer 6970 gets to the 580.

Now add in some test higher quality settings (higher AA since only used 4x) and those results will come closer.
test which can be higher because there is still framerate room and settingroom: battleforge (-13%), Civilization(-24%) brings those to 8xAA and the gap is getting closer. (just to name 2)

So add those higher quality settings and magically the 6970 becomes faster overall.

25x might not be used very often, it is an indication for future. for 16x, most used, those cards can reach playable framerates and you won't notice the difference with all settings enabled.(which was also not tested).


edit:

6970 need to reduce price in my country to fight with 570

Maybe. As it stands currently you might be right. But i don't think either of them is a clear winner. I believe the 6950 is the best choice for any card at the moment. and GTX560 will need to be really close to 570 if it wants to compete with that product.

[PatRaceTin]

6970 need to reduce price in my country to fight with 570

[AKM]

AMD dropped the ball on this one, Radeon 6970 can't even beat Geforce 570 gtx on pure performance not even in performance/watt the Nvidia card actually uses less power!

[kadozer]

Crossfire > SLI for 69xx. That and the 6950 2 major wins imo.

[Sn0wm@n]

That's great, yeah, but just not what I'd expected.

30% on the same node is epic .... did you expect 100% improvement ????


with good yields and good pricing???

[Pantsu]

30% on the same node is epic .... did you expect 100% improvement ????


with good yields and good pricing???

Where does this 30% come from? According to the TPU review 6970 is 13,6% faster than 5870 at 1920x1200. The die size increased ~16% at the same time. Epic indeed.

As for the Power consumption, HWC is afaik the only site that puts 6970 power consumption over gtx 570. IMO it's the test method making 570 look better. There are several game power consumption tests done by other sites that give a better picture.

CrossFire is the only bright spot, the scaling is simply superb now, clearly better than SLI. Those wanting to try eyefinity (like me) should be in for a treat with a couple of these.

[Olivon]

AMD dropped the ball on this one, Radeon 6970 can't even beat Geforce 570 gtx on pure performance not even in performance/watt the Nvidia card actually uses less power!

[IMG]

Yeah, it's really the sad part ...

Finally, Fermi is not so bad ? No ?

[Tuvok-LuR-]

wtf?
I was going for the 6950 cause in most of the reviews it seemed to have same power consumption as 6870, now I see tests where it withdraws 50w more

[duploxxx]

AMD dropped the ball on this one, Radeon 6970 can't even beat Geforce 570 gtx on pure performance not even in performance/watt the Nvidia card actually uses less power!

perhaps from a more consistent source : Anandtech for example, you will see that actually the 570 consumes more, so NO AMD didn't drop the ball this time, sure it's not a major enhancement from 58xx series, I also expected i bit more from the increased SP count, but neither is the 5xx series from NV they just fixed what they screwed up before. AMD has set the design for new design in future, expect more from next generations. NV always had the fastest single gpu card for many generations (comparing generations against each other, not just a small glimps in time due to design issues), don't understand all the fuss that AMD would take single card perf crown... that stupid hyping.

http://www.anandtech.com/show/4061/a...eon-hd-6950/24

[flyck]

AMD dropped the ball on this one, Radeon 6970 can't even beat Geforce 570 gtx on pure performance not even in performance/watt the Nvidia card actually uses less power!

strange, since the 6950 consumes only <20W more then 5850 according to all the review sites and according to that review it is >50W. Seems like they used the +20% setting for power measurements used for overclocking.

[Nintendork]

wtf?
I was going for the 6950 cause in most of the reviews it seemed to have same power consumption as 6870, now I see tests where it withdraws 50w more

No, see the anand chart.

This is a power consumption chart not an fps one (message to some people)


You won't expect to consume the same as 6870 being a faster card (and twice the memory) but Powertune is your friend, a pretty cool feature.

[Tuvok-LuR-]

Is my Corsair vx 450 enough for an E8400 @ 4ghz on p35 and a 6950? (fine on a 4890 atm)

[Nintendork]

4890 actually consumes more so you should be OK unless you want to do a useless furmark bench with powertune savings off.

[sergiojr]

Well, GTX570 has also lost 8W from its own review.

What's going on?

[Sn0wm@n]

yeahh wtf on hardware cannucks power draw test.... Oo

[~CS~]

To me it seems AMD did the most they could with what they had,the load power consumption is quite high albeit alleviated by very low idle power,and an ability to manage load consumption with powertune.If the chip consumes so much at only 390mm2 ,perhaps AMD was limited,as to how large they can go by the arch and didn't aim too low,like I first assumed.

The cards do seem well balanced overall,nothing is really bad,just underwhelming,and a 2GB buffer does make it look more appealing for some users.

[flyck]

No, see the anand chart.



You won't expect to consume the same as 6870 being a faster card (and twice the memory) but Powertune is your friend, a pretty cool feature.

Maybe same question for duploxxx, but aren't those CF power numbers of anandtech strange?

full system power
1 6950 --> 292W
2 6950 --> 472W
difference = 180W

1 6970 --> 340W
2 6970 --> 564W
difference 220W

so these numbers are without power tuning?

also that would imply that the rest of the system would only consume 292-180 = 112W

but that would mean the 5770 would consume 130W (243-112).

I know there can be illeguralities as cpu limited or other things, but those would make the gap bigger instead of smaller. (if for example card 2 only is used 80% due to limitations somewhere else, the actual card consumption would be even higher).

[zalbard]

Is my Corsair vx 450 enough for an E8400 @ 4ghz on p35 and a 6950? (fine on a 4890 atm)

Yeah, you'll be just fine.

[SKYMTL]

Well, GTX570 has also lost 8W from its own review.

What's going on?

Basically, we had to redo ALL of the tests because the Corsair HX1000 I was using died. It was replaced with the same product. Since overall efficiency changes from one sample to another it was necessary to redo them.

Most of the cards stayed the same. However, one or two changed by a bit (usually less than 2%). You will also see this when we revisit the HD 6850 and GTX 460 in the near future as both cards had their consumption trend downwards a bit as well.

[Tomasis]

All you guys whining about how it didn't live upto the hype - AMD has remained virtually silent - the Hype is of your own making. Theorising over and over about specs, when nothing concrete has been released, trying desperately to get the information you all crave. So you built up the card into some nVidia toppling king, which clearly ISN'T the case.

Look at the data you have:

5870 = 1600SP's/80TMU's
6970 = 1536SP's/96TMU's

You're expecting miracle performance from this....why? The fact it's 15 to 20% faster than a 5870 which has an SP advantage, is pretty impressive. Take into account the whole package. It's EVOLUTIONARY - not REVOLUTIONARY.

nVidia have consistently released cards with the words 'WE ARE TEH WINZ0RS! WE ARE FASTER THAN ANYONE!' and then slap on a pricetag to match. So AMD comes along, gives you 95% the performance, for a lower price, gains marketshare, as they're 'fast enough' (this is an INITIAL release, with release drivers afterall). Seriously - what gives with some of you?

I buy cards not based on the brand, but the bang-for-buck. I considered upgrading to a GTX460 from my HD4850, and then the 6850's came along, stomped on that sandcastle, and here i am again, re-assessing what i really NEED. Do i game at 2560x1600 with ALL the eye candy? No. I have a 1680x1050 monitor that i want to last before replacing - and most games i find don't need more than 4xAA & 8xAF to look 'pretty'. If you want more performance out of these cards - overclock them, perhaps? That is, afterall, what this forum specialises in, not hypothetical fanw@nking...

well written! what is to be expected with less shader and same node? I like some new things as power saving thing

69xx is quite good.. even better at Crossfire again..

people were disappointed for 68xx but missed that they scale in CF at good way with less shaders. I think 6850/6950 CF gives amazing value/perf.

Seems that Nvidia fans care about being No.1 while AMD is humble, silent player, acting to sell price effective cards for mass people. while most nvidia fans afford 460 and spend time praising the ultimate fastest card

6990 comes and rock their world down like 5970 does

[SKYMTL]

Regarding efficiency (or lack thereof), I just want to make it clear that we do everything possible to exclude outside variables from our power consumption test.

The power supply is plugged in to a Tripp-Lite line conditioner that regulates input voltage to a constant 120V. If this isn't done, input voltage WILL impact the consumption of the system and let's be honest; no one likely lives in a house with constant live voltage.

In addition, the 3DMark test we use puts a very minimal load on the CPU. This is important since the CPU's fluctuating load patterns in most of the apps out there can really mess up results.

I am also surprised to see quite a few sites still using FurMark, etc to test power consumption. AMD admits to that program being capped by PowerTune.