The Official Graphics Card and PC gaming Thread

Meh, powerdvd is ok. It has a similar smoothing effect that TMT 3 has. Perhaps I haven't fine tuned enough, but the user interface seems better to me on TMT3. But perhaps I'm simply use to it Less nags too. Though those can probably be turned off.

 

DIGITAL VIBRANCE

Well, I think you're both right - it IS saturation - and maybe gamma has something to do with it.

But it might as well be a brilliant new technology, because there is an easy way to do it, and there is a hard way to do it. I'll test it out on catalyst, but I love the Riva options, off, low, medium, high, and max. Instant change to the beautiful screensaver wallpaper, to the game when I alt-tab back in. I see where it is, in the nvidia control panel - but I prefer the riva instead.

Here is a suggestion for you guys: Do a google search for "color radiance." At one time about two years ago, an official spokesman for ATI said "We know people like the nvidia digital vibrance, so we're going to give that to them, and ours will be called color radiance." Then what? No further discussion.

I am telling you guys - it is just nice to have a thing, called digital vibrance, or color radiance, and you move the damn lever, and the colors start to pop. In all fairness I will test it out in Catalyst - some future post - but if Ati said they were going to provide color radiance, then why didn't they?

Maybe there is actually more to it than just saturation. It has to be enabled in hardware, I understand, from all the forum posts about it that I have read. Maybe Ati decided they couldn't deliver, for one reason or another. Or maybe you guys are right, and it's just saturation, and I'll find that Catalyst saturation changes provide the exact same result. I hope so. Some future post.



Guess what new game I decided to boost from medium to high (but not max.) Dragon Rising - the game you wrote about so convincingly, Jeff, a year ago.

I am 4 campaigns in, and I love the game (and sometimes hate it too, lol.) Jeeeeez you were right when you said the squad controls took some getting used to - very steep learning curve on that. Grenade controls are AWFUL. You can't just throw a grenade. First you have to select grenades, then scroll through the optional grenade types. If you don't do that, you might end up tossing smoke instead of a frag grenade, as I did.

And while we're on the subject, what is with the two different type of frag grenades - I guess one is a little bigger - so f**ing what!!! What moronic idiot over there at Operation Flashpoint decided to be so picayune with the details that he burdened us with a choice of two different frag types!!! HE SHOULD BE SHOT!!!(The hate is coming out, lol.)

But there is a lot to love about the game, and thanks again for the memorable review, Jeff.

Wide open maps! With the help of enhanced saturation (digital vibrance on high) the golden brown fields light up in contrast to the deeper tones where the shadows are. Without saturation turned up, the game is very flat. A red tractor, in a patch of golden brown, next to a barn, is a stunningly beautiful momentary distraction to the fact that you are surrounded by enemies, and heading out to take down some mg nests and spotters on the ridge across the way.

I am playing on normal - there is experienced and veteran - but according to the many hints, the game play is no less intense, but you get more help, and everybody respawns at checkpoints - your entire squad is back and they are healthy. I have learned about formations - V when I kind of want one of them out in front a bit - Wedge when I want to take the lead - tight when I want them right around me when an enemy can pop up any second in close-in combat and take me out with a head shot, just as we're almost done killing everybody before the castle assault.

One of the helps that I have enjoyed are the yellow markers suggesting the way to go. But they have led me right into the path of squad death many times, so my flanking movements sometimes go way around. But the yellow is helpful for teaching about tracking maneuvers going right, then left, up narrow canyons, straight up the ridge for a bit, then jog to the left, to leave behind an enemy patrol in the dark.

The game simply suffers from too little play-testing. They could have worked out the bugs and created a smoother experience, with better training, and reduced the learning curve. They would have sold a lot more copies with much better word of mouth and better reviews.

But - overall - I love it! Thanks again, Jeff, for your review.

And thanks DXR. Your comment about the overclock helping squeeze a couple more frames out of a game proved true on Dragon Rising. With 4xAA, full 2560x1600, on the oil refinery "blow up the generator" mission, all of a sudden (probably because of displaying the refinery below) my frames dropped and I started lagging. I turned AA down to 2 and I was still at only 25. I turned on the stable overclock - core from 594 to 621 - (I had tried the more aggressive 648 but the game crashed within 5 minutes) - and frames moved back to 28-30 - fairly smooth again. I left it on and played another 6 hours. I forgot to monitor temps. Anyway, I'm glad I took a few hours and played with the overclock settings last week.

Remember, I was going to get TMT 3, but they said they only worked on nvidia platforms when I ran the little test. So it sounds like you're saying it is no better at what it does than TMT 3. Okay, that's good. What about the problems you were having with frame rates on DVD upscale, etc. Did it fix those issues? LOL

(Someday, someday, I'm going to post - you're gonna see what forced me to fork out hard-earned cash for Power DVD 10. Kevin can say Meh, because I guess he already has a brilliant non-free product.)

Hahahahaha

 

You do realise rivatuner works for both nvidia and ATI graphics cards Rich?
The results of your graphics card overclock are purely coincidental, a tiny 4% overclock can never result in a 10%+ performance boost.

 

For the time being, the problem playing dvds seems to have disappeared. Even with the hardware acceleration enabled, which seemed to be the problem before. I've been watching Sliders episodes. Excellent show!

 

Haha always glad to see my writing appreciated


I also agree that the game has a lot to love. Complicated controls and glitches aside, the game is extremely deep and allows a lot of freedom in how you complete the missions.

 

GTX 570 is released to much fanfare from the fanboys.

Performance, as anticipated, 195% to an HD4870 (125% to an HD5850).
Power consumption essentially identical to a GTX470, 195W low-load [i.e. biased tests], 215W typical, 270W heavy load [CPU-unbound title]

The new coolers for the GTX570 address the noise, and for the most part the heat, of high-performance Fermi cards up until now, but the power consumption is still astronomical as ever, with two HD6850s using less power than one GTX570, despite being an astonishing 50% faster, on average, and almost the same price ($370 vs $350 / £290 vs £300)

As nice as the new single cards are, the best price/performance ratio now seems to be to buy two cards. Scaling sure has come along in recent times.
It's worth mentioning it has taken nvidia over 6 years to reach 85% scaling typical. In less than five years, AMD have reached 95% typical, often reaching 100%. Impressive stuff.

As a reminder:

HD5750: Performance/watt 94%, Performance/Dollar 69%
GTS250: Performance/watt 55%, Performance/Dollar 70%
HD5770: Performance/watt 88%, Performance/Dollar 70%
GTS450: Performance/watt 93%, Performance/Dollar 87%
GTX460: Performance/watt 82%, Performance/Dollar 83%
HD5830: Performance/watt 79%, Performance/Dollar 79%
GTX461: Performance/watt 80%, Performance/Dollar 75%
HD6850: Performance/Watt 118%, Performance/Dollar 81%
HD5850: Performance/watt 103%, Performance/Dollar 78%
GTX470: Performance/watt 70%-, Performance/Dollar 64%
HD6870: Performance/watt 115%, Performance/Dollar 72%
HD5870: Performance/watt 96%, Performance/Dollar 62%
GTX480: Performance/watt 66%, Performance/Dollar 44%
GTX570: Performance/watt 85%-, Performance/Dollar 55%
GTX580: Performance/watt 80%, Performance/Dollar 44%

Dual-card configs
2x HD5850: Performance/watt 90%, Performance/Dollar 68%
2x HD6850: Performance/watt 115%, Performance/Dollar 79%

2x HD5870: Performance/watt 84%, Performance/Dollar 54%
2x HD6870: Performance/watt 112%, Performance/Dollar 70%

2x GTX460: Performance/watt 76%, Performance/Dollar 77%
2x GTX461: Performance/watt 74%, Performance/Dollar 69%

2x GTX470: Performance/watt 63%, Performance/Dollar 58%
2x GTX570: Performance/watt 76%, Performance/Dollar 50%
2x GTX480: Performance/watt 59%, Performance/Dollar 39%
2x GTX580: Performance/watt 72%, Performance/Dollar 39%

Performance points

c. 75: $110: HD5750, 81W (350W PSU)
c. 100: $115: GTS450. 106W (350W PSU) $135: HD5770. 108W (350W PSU)
c. 130: $150: GTX460. 155W (550W PSU) $165: HD5830. 165W (550W PSU) $180: GTX461. 170W (550W PSU)
c. 150: $185: HD6850. 127W (380W PSU) $200: HD5850. 151W (550W PSU) $250: GTX470. 230W (550W PSU)
c. 180: $240: HD6870. 151W (550W PSU) $290: HD5870. 188W (550W PSU) $350: GTX570. 230W (550W PSU) $430: GTX480. 290W (650W PSU)
c. 230: $300: SLI460. 310W (650/750W PSU) $330: CF5830. 330W (650/750W PSU). $360: SLI461. 340W (650/750W PSU). $530: GTX580. 285W (650W PSU)
c. 280: $370: CF6850. 254W (550W PSU) $400: CF5850. 302W (650/750W PSU) $500: SLI470. 450W (750/850W PSU) $500: HD5970. 294W (550/650W PSU)
c. 350: $480: CF6870. 302W (650/750W PSU) $700: SLI570. 460W (750/850W PSU). $860: SLI480. 580W (850/1000W PSU).
c. 410: $1060: SLI580. 570W (850/1000W PSU)


In a more understandable form,

At the 75% performance point, you could buy an HD5750 for $110, but with the far faster GTS450 just $5 more, there really is no point. This performance point may as well be currently abandoned.
At the 100% level, you could buy a GTS450 for $115, or an HD5770 for a $20 premium. Performance and power-wise the GTS450 is ever so slightly ahead for both. This is still the one sector nvidia control. Still, the Radeon is not an absurd proposition at least. An extra 2 watts and $20 is nothing compared to the tradeoffs in the high-end sector.
At the 130% level, efficiency is pants all-round. None of the cards here will run off a single-6pin system, yet are outperformed by a card that will [HD6850]. The GTX460 1GB is also the same price as said 6850, so is completely defeated. The HD5830 is borderline worthwhile, with the reasonable price of the GTX460 768MB making it the most acceptable of three rather unacceptable choices.
At the 150% level, it's HD6850 or not at all. The other cards use far more power and need bigger PSUs, and are more expensive, for no real gains at all.
At the 180% level, it's HD6870 or not at all. The HD5870 is marginally faster and runs off the same PSU, but is $50 more, a lot for a 5% increase. The ludicrous 110% markup on the GTX570 for its 14% performance benefit isn't really worth it either, especially when you remember that GTX470/570 cards are at liberty to use the full 300W of their higher-end equivalents on odd occasions.
At the 230% level, the power of dual cards really shows through. Two GTX460s do at $300 what one GTX580 does at $530. They don't even use any more power. However, the memory performance of GTX460 768MB cards is woeful when high performance is needed, thus the 5830s are taking over, and they are priced perilously close to the pair of 6850s, which as usual, mop the floor with everything else performance-wise, and on a PSU literally half the price. Another bracket currently worth avoiding.
At the 280% level, two HD6850s can take on pretty much anything. Still less than 3/4 the price of the far slower GTX580, the only SLI combination that can keep up uses an astonishing 430-550W, versus the mere 250W of the 6850s. That stings after the $130 extra they cost.
At the 350% level, two HD6870s are again out on their own. The rival SLI pairing is $220 more, almost 50%, and as before, the power consumption required is effectively double.
At the 410% level, we're waiting on AMD's HD6900 release. Excluding Tri-Crossfire of HD5850s which would reach this point [in some titles], it's basically nvidia-only with two GTX580s. Still, that's $1060, and 600 watts of GPU. 17% more performance is nice, but is it really worth $580 and 270W+? Something to think very carefully about, and with the hopefully imminent launch of the 6900s, it's worth a wait if you're going this far.

 

Very nice summary Sam. Makes me feel better about playing musical video cards to get my HD6850s XD

 

Well yeah, what you did makes more sense with two cards than it does with one, as with one card, moving from a 5850 to a 6850 loses you about 3% performance for the sake of 25 watts. This difference is more than enough to make the 6850 a far better card for a new system but doesn't make much sense as an upgrade.
With two 5850s to 6850s however, not only are you obviously losing 50 watts and two power connectors, but you're also gaining 8% performance due to the scaling difference.

The HD5850/5870 price drops have been reasonably aggressive but I think this is simply EOL stock clearing going on, the brands available for the old cards are becoming thin on the ground.
Sadly these deals are rarer in the UK as most retailers here do not bother stock-clearing properly. The HD5970 has been pretty cheap in the US, not so here.
Of course, the scaling difference of the 6800s makes a bit of a mockery of using an HD5970 these days.

AMD do need to get a move on with the HD6900s however, as while the single-GPU lead means less today than it used to, there are plenty of people who will happily spend a lot of money but insist on getting a single GPU, and you can understand this logic given there are still plenty of problems with dual graphics in games.
As of now you may be paying 45% and 120% more respectively, but nvidia still have a 14% and 35% respective performance advantage that AMD do not have a rival for yet.
It all comes down to if you accept/reject dual-GPUs can compete with a single GPU. Given the steps forward with scaling, both in terms of raw capability and also reliability, with the HD6 series, at least for two GPUs, I think this is viable.
In this case, the GTX570 does have a rival, two 6850s. They cost about the same, they use a similar amount of power, but are 50% faster.
Additionally, the GTX580 does have a rival, two 6870s. They're currently about $50 cheaper, use a similar amount of power, but are 45% faster.

More interestingly, if we apply the scaling of the HD6800 series to the expected performance levels of the HD6900s, dual 6900s should hopefully be a force to be reckoned with.
Two HD6950s should be effectively inseparable from the $1060 570W pair of GTX580s in unbiased titles.
This leaves Two HD6970s out on their own at an anticipated enormous 510% score, versus the pair of 580s which see at most 430%, and even three GTX480s which rarely exceed 480%.

It's not really obvious if the advances in dual-GPU scaling for the HD6800 series will affect 3 or 4-way crossfire performance. For the moment I am going to say no, as the lack of a second crossfire connector on the HD6800s is unprecedented, and suggests issues here.
I would of course love to be proven wrong here, as two HD6990s in QuadCF would prove pretty powerful if this was the case.

 

Wow, Sam, brilliant performance analysis! I saved both of those posts with my web capture.

- - - - - - - - - - - - - -

Haha. Maybe so, maybe so. But my stable overclock is 621, rather than stock 594 - so that's 27/594, about 4.5%, you're right. However the shader clocks have jumped about 100, and the memory clock is also about 100 higher: shader originally - I can't remember, and memory originally I think around 1100. So for shader and memory I have increased closer to 10% I think. Pretty much I believe I got about 3 fps extra out of the deal, moving from 25 to 28fps, and everything turned smooth.

Let it be said, Jeff, that I would NEVER be playing and enjoying Dragon Rising, a year after you played it, had you not invested the time and imagination and applied your abundant writing skills to pen an exciting review that stuck in my mind for a year.

Unlike today, where I have gotten smarter and I have a games folder, with a Dragon Rising subfolder, I didn't save your review with the web capture tool. So I had even forgotten the name of the game, lol. I remembered Big Maps, and a horde of enemy chinese coming in down the hill about a mile away (I am actually about 10 missions in but I haven't gotten to that one yet) and as you withdrew, you remembered that you could call in a fire mission (I forget half the time also) and so you were able to slow those guys down and get your buddies out in one piece.

When you say, "allows a lot of freedom in how you complete the missions", truer words were never spoken. They planted a Stryker right next to me at the start of one mission, and I watched it pepper the distant enemies thinking to myself "Why doesn't the driver pull closer?" I went on with my squad on foot, and had one helluva hard time - finally throwing smoke to the left - to work through a thin corridor with mg nests on two left hillsides, mg nests on two right hillsides, and Anti Tank squads directly ahead. It was only later when I re-played some earlier missions, and had already stolen a jeep from a fellow squad, not wanting to run a half mile to the enemy position again, that I looked over at the Stryker, and jumped in. Duhhh!

Wow! IN THE STRYKER was a new game - no need to worry about mg nests - stay far back enough to try to kill the AT guys - and the Stryker can even take a hit or two from an anti tank weapon, head-on, without appearing to suffer too much damage - it must have a lot of front armor.

I have a question for you: Having run out of ammo before I finally realized the Enter key was the magic answer for taking a dead enemy's weapon - I pointed the pistol at distant enemies, and the radial offered the "Engage Enemy" choice for my squad. But if I point my rifle at an enemy, no such "Engage Enemy" choice appears.

My question is: Do you know anything about how that works? Maybe I don't have to run out of rifle rounds next time - just switch to pistol and activate the radial - I'll try it and see what happens. I like that Engage Enemy much better than the "suppress fire" option, which causes them to just keep firing at one spot even after the enemy has been killed.

You're gonna laugh, but that mission where I ran out of bullets, I also had a hard time getting the pilots to follow me, because a strange high voice said "Follow me" and I thought it was the pilot talking, added to the fact that he didn't seem to follow very well. So, since I thought HE had given the command, I followed him around, and the squad and I battled chinese soldiers popping up everywhere - it was hairy. Finally I gave up and issued "follow me" for about the 4th time, and was shocked to see that he seemed to finally be obeying - a little behind however - so much so that the copilot was killed off - but I was able to finish the mission delivering the main pilot in one piece.

The funny part is that later on, I heard that strange high voice again, and so in hindsight, I think it actually WAS MY OWN VOICE, not the pilot's. Hahaha.

Most of the fighting is long range, but I have learned to come in close, strafing right or left, on occasion, and take some of the enemies out that way. They have an option called Focus, and I put that on my CapsLock so I can WASD move and simultaneously activate CapsLock with my little finger. Focus provides a two-to-one zoom, the same as the scope, but I like it better for strafing since the field of view is not obstructed by the scope, as long as I can target the enemy with the little dot in the center of the two horizontal lines. The little dot is not as easy to use as the crosshairs of the scope, but for close-in combat the unobstructed view more than makes up for that.

Additionally, the Focus option helps avoid the problem of scoping in, and getting killed in the fraction of a second it takes to move the scope up to the firing point. (I can't remember exactly, but I think that when you scope in, you are initially looking slightly below where you were looking a moment ago. Maybe that is more realistic and how a gun really works. But I haven't encountered that before this. An an example, as opposed to how Dragon Rising does it, the bullpup rifle on Counter Strike zooms in, not only with no obstruction of view, but also with no adjustment to viewing position.)

Lastly - wow - I played through MW2. Fighting house by house in Arcadia, I followed the Anti Aircraft Stryker through my old neighborhood - the two-story house with the swimming pool was where I grew up! Just kidding. Arcadia is about 60 miles north, up there near Pasadena and Glendale.

After playing through the game, I hung around for some of the other action, holding out on the roof with my sniper rifle and heart beat weapon through several rounds of attacks. Then I hit the Brazilian favela, and enjoyed the twin uzis for close-in action, holding down both mouse buttons when I wanted to eliminate a gang banger with no questions asked, or alternating mouse buttons, to make sure I had a full uzi for each of two german shepherds, without needing to reload!

I guess we'll be seeing Soap and Price again. I really like Price - what part of Britain is he from, Sam or Shaff?

Rich

 

Again, firstly FPS varies too much in non-scripted gameplay to ever accurately prove results differences that small.
Secondly, smoothness is also a placebo. If 25fps is not smooth, 28fps is not smooth, they're effectively indistinguishable. It's a complete placebo, and one that will see you killing off that 8800 very fast indeed.

The origin of Price and Gaz I don't think is revealed in the story. However, their voice actors, Billy Murray and Craig Fairbrass respectively, are both from east london [Bethnal green and Stepney], 21 miles from where I live, and about 15 miles from where Shaff lives
Both of them to an extent have the distinctive cockney accent that comes from that background.

Apart from voicing characters in the Modern Warfare games, both of them have appeared in a considerable amount of other media.
Interestingly, Craig Fairbrass played one of the roles in the film of the game Far Cry.

 

farcry had a film? How bad was it haha.

man these roumers on the 6970s are haywire. From GTX570 perofrmance to 40% better than the 580. From 1536 cores to a dual shader core on one chip with the total nu,ber of shaders giving 1920 cores. Not to mentiona picture with a small switch next to the xf connectors. Most confusing thing ina while, pleae hurry up the 15th!

 

Not seen it, but the wiki page said 'the film was generally panned by critics'.
Should say enough. I watched the doom film. It wasn't particularly good.

Rumours on the 6970 are nuts, but nowhere near as wild as that. 40% better than the GTX580? Where'd you find that? So far, my estimate of 12% faster than the GTX580 is about as high as they have been.

The reasons for all the craziness are as follows:

1. AMD's early southern islands official specification sheet stated 1920/1600 shaders for the 6970/6950. Now that the GPUs are in production, the actual specification sheet now reads 1536/1408 shaders, they have lowered it. Why this is, we're not sure. AMD seem to say it isn't to do with yields [after all, a 25% reduction is huge]
2. Someone put out a list of performance results for the card which were fake, and well below the mark. This is where people are getting the 'same as a GTX570' numbers from.


Think about it. It has a 70% higher TDP than the 6870, it has a 70% higher texture fillrate than the HD6870, and 40% more shaders.
50% higher than the 6870 seems a perfectly reasonable assumption to me, and that places it exactly where I have estimated.
Of course, this is in a neutral environment. Even at 12% beyond the GTX580, there are several games that are so biased, it will struggle to compete with the GTX570 in. That too could affect people's views.


Oh, and latest rumour control has pricing for the cards at $375/£280 for the HD6950, and $450/£335 for the HD6970. That to me suggests AMD intend the 6970 to at least be faster than the GTX570, but maybe not as fast as the GTX580? Then again, they may just be undercutting them. After all, the GTX480 was 40% more than the HD5870 when it first came out, despite only being 5% faster.
Of course, it could also be the case that these rumoured prices are total BS. Looking forward to wednesday, it should be interesting.

 

Some rumoured "accurate this time" results for the HD6970 coming through, with the release driver [10.12 pre]
Remember the performance levels we were looking at were 250-260 for the HD6970, relative to the HD4870.
We already know the HD5870 runs at 180, the GTX570 at 194, and the GTX580 at 230.
Some bias is to be expected in the more nvidia-focused games as ever, so results for those would be lower, but 250-260 is the magic number we're after, based on first estimates, and those I have used throughout.

Unigine 2 Benchmark: 251
Unigine 2 Benchmark [two cards in crossfire]: 451
Unigine 2 Benchmark [30", 8xAA]: 270

Aliens vs Predator: 210
Batman Arkham Asylum: 223
Borderlands, 4xAA: 237
Borderlands, 8xAA [24"]: 258
Borderlands, 8xAA [30"]: 287
Crysis Warhead, Gamer [24"]: 256
Crysis Warhead, Gamer [30"]: 312
Crysis Warhead, Enthusiast [24"]: 241
Crysis Warhead, Enthusiast [30"]: 299
F1 2010 [24"]: 227
F1 2010 [30"]: 243
Left 4 Dead 2 [24"]: 221
Left 4 Dead 2 [30"]: 243
Metro 2033: 235
Riddick AODA: 237
Wolfenstein MP: 254
Enemy Territory: Quake Wars: 248


Certainly some odd results here, both of the quite disappointing kind, and the ludicrously exciting kind.
The results for Borderlands are mindblowing considering the game is one of the most nvidia-biased in history.
Overall, most of the results seem on track with the estimates, with an odd boost in performance for the 30" resolution, very unusual for Radeons in the past.
With Left 4 Dead 2 being an exception [but not really a game that requires perfect performance anyway] most of the results here are good, in the ballpark or near it.
Now of course we just have to hope this is actually true


Something else interesting of note:

HD6950 3DMark Vantage test [Perlin Noise]
Standard Mode: 140fps
PowerTune +5%: 155fps
PowerTune +10%: 162fps

PowerTune, assuming it is not enabled in the results posted further above, looks to offer some impressive extra grunt.


Who's excited about the NDA lift tomorrow?

 

im just thinking that its a shame it wasn't 32nm or we would be getting dual 5870 performance, or just about.

but if the 285 pricing is true, then that's pretty good compared to the 580, but not so great when the 6950 is225 ish. Nor do i think its better than r 6850s.

but compared to nvidia line ups its spectacular.

 

Not necessarily. We don't know exactly how well this architecture scales up. The 6970 has already had several extremely negative rumours floating around regarding its comparative performance, though it does seem that this is all down to bad drivers being used.

The only estimates of price we have coincide with the 'wrong' or 'fake' levels of performance, so I'll stress that now. While ATI don't price things too excessively, it would seem unlikely for them to massively undercut nvidia here, so if the 6970 performs as well as the chart above dictates, then it won't be cheap.

Ultimately two 6850s will of course be better than one 6970, that should go without saying, as the 6850 is considerably more than one half of the 6970. However, people who want not to have to deal with the issues regarding crossfire, or that can only fit one card in their case, will still be buying the 6970.
Also bear in mind those who will be buying two 6970s for more performance that two 6850s/6870s cannot offer.


Given that the GTX570 is £280-£300 and the GTX580 is £410 [typically £450 due to poor yields], I would estimate the HD6970 as probably being at least £350, if not £400, and the HD6950 perhaps £250-£300.
How effective the cards really are though depends on the prices AMD choose, and whether the component availability means they're as rare as the HD5 series was at launch, as that's also a distinct possibility.

 

Ok, the HD6950 and HD6970 are now priced, and in stock, ready to buy.
The HD6950 will cost £220 / $295
The HD6970 will cost £280 / $375

The HD6950 will score around 1-2% faster than the HD5870.
The HD6970 will score around 10-12% faster than the HD5870, scoring similarly to the GTX570.

The HD6900 series cards are much slower than the initial rumours, and have been priced lower because of this. The GTX580 is still the top card of this generation, and will likely remain unchallenged.
Those who already own an HD5870, there is no real upgrade to be had from the HD6900 series, surprising since the HD6800s performed so well.
The HD6870 and HD6950 are almost identical cards apart from the tesselation and video memory, hence why the HD6950 price is so similar.

Frankly, the HD6900 is a complete lemon, and for the majority of people, is best avoided.

 

told you with price

 

There's one thing I forgot to mention, resolution.
All this applies up to 24" inclusive, and when AA is not so much applied.
Get to 30", especially with AA, and all hell breaks loose.
We're talking one 6970 beating dual 5970 quad crossfire configs. It's utter madness. Chart incoming.

 

Far Cry 2
HD6950: 180 [noAA], 185 [AA ->24"] 190 [AA 30"]
HD6970: 195 [noAA ->22"], 200 [noAA 24/30"], 205 [AA ->24"], 210 [AA 30"]

AvP
HD6950: 180 [noAA ->22"], 185 [noAA 24"], 195 [noAA 30"], 215-230 [AA 17-24"], 240 [AA 30"]
HD6970: 200 [noAA ->22"], 205 [noAA 24"], 220 [noAA 30"], 245-260 [AA 17-24"], 265 [AA 30"]

Crysis Warhead
HD6950: 180-185 [noAA ->24"], 210 [noAA 30"], 190 [AA ->22"], 195 [AA 24"], 285 [AA 30"]
HD6970: 200 [noAA ->22"], 205 [noAA 24", AA 17"], 210 [AA 22"], 220 [AA 24"], 235 [noAA 30"], 325 [AA 30"] (Yes this is correct!)

DiRT 2
HD6950: 175 [noAA ->24", AA ->22"], 180 [noAA 30", AA 24/30"]
HD6970: 185 [AA 17"], 190 [noAA 17", AA 22"], 195 [noAA 22-30", AA 24/30"]

Metro 2033
HD6950: 195-205 [noAA ->24"], 220-230 [AA ->24"], 235 [AA 30"]
HD6970: 215-220 [noAA ->30"], 240 [AA 17"], 245 [AA 22"], 255 [AA 24"], 265 [AA 30"]

Just Cause 2

HD6950: 170 [noAA ->30", AA 17/22"], 175 [AA 24/30"]
HD6970: 180 [noAA 17"], 185 [noAA 22/24", AA 17/22"], 190 [noAA 30", AA 24/30"]


It is worth pointing out that the enormous gains at 30" with lots of AA in certain tests are also shared to an extent by the GTX570 and GTX580. Typically, in these scenarios the HD6970 compares to the GTX580 similarly, sometimes slightly above, sometimes slightly below.

 

HD6950 and HD6970 cards are now reviewed at all the major sites.
Newegg are selling the cards for $300 for an HD6950 and $370 for an HD6970.
This compares to $240 for an HD6870, $360 for two HD6850s, and $350 for a GTX570. The GTX580 is priced at $530 but you can't currently buy it.
In the UK, OcUK are selling the HD6950/70 from HIS at £220/280 respectively, and ebuyer are doing the same for Powercolor.

Having processed a 30" Benchmark suite, I have the final full metrics for the new cards.
These are 30" figures, so they will be skewed down for the weaker 6870, and skewed up for the 6950/6970, which perform a lot stronger at 30" than lower resolutions:
This is still 4870reference figures. To compare to 5850reference divide by 1.55.


HD6870: 158 (Normal value: 171)
HD5870: 180 (Normal value: 180) [reference point]
HD6950: 192
GTX570: 194 (Normal value: 194)
HD6970: 217
GTX580: 232 (Normal value: 230)

So, instead of the expected 215/250, the 6900 cards came in at 192/217. A disappointing result, but then they are also cheaper than we expected.
The dreams of a giant Radeon that costs a fortune but outstrips everything from a single-GPU perspective, have been shattered.
What we have here, is a more power-hungry, noisier, HD5870, with 2GB of memory as standard, that costs the same as what the HD5870 did in its latter period, for a 20% performance gain.

The good news, is that the results of the overstress tests [for instance, running stuff like Crysis at 4xAA maxed out] are true. So in crossfire, two 6970s can take on what no other cards really can [GTX580 SLI excluding] at 2560x1600.

Power consumption/Temp figures:
HD6870: 20/151W 45/80C
HD6950: 19/171W 41/85C
HD5870: 22/184W 42/85C
HD6970: 20/212W 41/89C
GTX570: 25/235W 43/85C
GTX580: 31/289W 43/86C

OK Crossfire and lower resolution scaling figures are in, here's how it stacks up for everyone except the 30" club: these are the figures that matter to most people

Remember, this chart takes the average of all mainstream titles, both biased and unbiased, including the ridiculous bias-fests that are Lost Planet 2 and HAWX 2.


GTX460: 93
SLI460: 137
HD6850: 142
GTX461: 153
HD5850: 155
HD6870: 164
HD5870: 180
HD6950: 191
GTX480: 206
GTX570: 207
HD6970: 208
GTX580: 245
HD5970: 256
CF6850: 261
SLI461: 272
CF6870: 300
CF5870: 305
CF6950: 334
SLI570: 339
CF6970: 360
SLI580: 386

Taking this apart, what do we see?
1. 768MB on the weaker GTX460 really hurts. All the tests in this group that measure 30" performance measure lower resolutions too, so there's no real high-res bias here, but dear god, does the 460 fall flat.
2. you notice the biased games have an effect on the HD6800 cards, they're weaker than they would otherwise be by a good 10-15 units here.
3. you can see the bias take hold with the GTX480/570/580 sitting a good 10-15 points higher than they should be.
4. at 83/84% versus 64%, the average scaling of the 6850/6870 is a good deal higher than that of the GTX570, which oddly fares far worse than the 460 1GB cards at 78%. At 75/73%, the crossfire scaling of the 6950/70 is weaker, as unlike the 6850/70, they do not have 95-100% scaling in a majority of titles, just one in this suite, Bad Company 2.
5. Despite the ridiculously biased games bing included in this test, two 6970s are within 7% of two 580s. 7% performance difference for $320 and 154 watts. Makes the 6970s seem pretty good value to me!


I will attempt to recompile this chart without the two biased games in it to see what a more fair representation of the cards is.