Reviews - AMD Radeon HD 6970 and HD 6950 2GB graphics card review

[Jimbo75]

If Hexus hadn't benchmarked the 6950 with both sets of settings people would accuse them of AMD bias.

At which point there could be some proper discussion. BEFORE the benchmarks were changed.

The review also comments at least once on how little difference the increased IQ settings make to the performance of the card.

The difference in performance isn't the issue. The issue is nVidia making up what is basically a total lie and AMD suffering for it.

And Hexus has published several articles explicitly stating that they've compared the two sets of IQ settings extensively and can't see any difference themselves.

This is why I was so annoyed when I read that they would be using HQ filtering. If you can't see it, why is it being used?

[Tarinder]

Watch out for what NVIDIA does it its next big driver release. The I.Q. debate may well sort itself out.

[kalniel]

Ok I was overly hasty, and obviously Tarinder has attempted to come to a reasonable solution by only doing it for the 6950, but that isn't really the point. What about the next review? All HQ settings or not? Does this just become another unfair advantage that nVidia has done NOTHING to deserve?

No, because a) Hexus will describe their testing process (or will be poked by us until they do, we never give them an easy ride). And b) Hexus have no reason to give nVidia an unfair advantage. The fact that they're currently using the defaults that nVidia are complaining about should be enough to rule out that thought. Going beyond that is unjustified paranoia. I often complain about benchmark choices, but it's Hexus' right to use what they like, and up to us to do the research to match the benchmark with our own projected usage. There are always other reviews.

Don't think that just because it's only 1-5fps that is being lost, that it will be that in all cases. nVidia is absolutely insidious and will use everything they can to get an unfair advantage.

I disagree. They'll use everything they can to get an advantage. That's their job. I disagree with them when they do unfair things, but they're quite capable of getting fair advantages as well.

[Jimbo75]

No, because a) Hexus will describe their testing process (or will be poked by us until they do, we never give them an easy ride). And b) Hexus have no reason to give nVidia an unfair advantage. The fact that they're currently using the defaults that nVidia are complaining about should be enough to rule out that thought. Going beyond that is unjustified paranoia.

I disagree. They'll use everything they can to get an advantage. That's their job. I disagree with them when they do unfair things, but they're quite capable of getting fair advantages as well.

So you think this is one of those "fair" advantages? When many websites have looked at the issue, found pretty much nothing of note, yet all at the same time decided to review with HQ settings anyway? (it's not just Hexus btw).

Wake up Kalniel. nVidia put the heat on with this one, big time. Just like they did with the overclocked cards in the barts reviews.

[Jimbo75]

Watch out for what NVIDIA does it its next big driver release. The I.Q. debate may well sort itself out.

That's interesting. They're doing the same I suppose? That would just be...so much like nVidia lol.

[Tarinder]

Let me state some facts for you, Jimbo75,

NVIDIA had ZERO influence on this review. It hurts me personally to think that they would have my ear and force testing regimes upon us.

The I.Q settings were put in to show that NVIDIA's case has little merit. The Radeon frame rates barely drop. Also, the bang4buck and bang4watt are calculated from the regular HD 6950's numbers, not the IQ ones. Further, the I.Q. numbers aren't mentioned anywhere in the conclusion or when arriving at a rating.

The reason for including them lies with us trying to be as comprehensive as possible, to ensure that neither AMD nor NVIDIA have any fundamental rights to complain. AMD acknowledges that it has tampered with the I.Q settings in post-Cat 10.9 drivers, so we wanted to see how this affected scores. The benchmarks show that there's little performance gain and, from what I can see, no discernible difference in image quality.

We're going with driver defaults unless there's a marked difference in the two firms' I.Q.

[Jimbo75]

Let me state some facts for you, Jimbo75,

NVIDIA had ZERO influence on this review. It hurts me personally to think that they would have my ear and force testing regimes upon us.

The I.Q settings were put in to show that NVIDIA's case has little merit. The Radeon frame rates barely drop. Also, the bang4buck and bang4watt are calculated from the regular HD 6950's numbers, not the IQ ones.

We've tried to be as comprehensive as possible, to ensure that neither AMD nor NVIDIA have any fundamental rights to complain. AMD acknowledges that it has tampered with the I.Q settings in post-Cat 10.9 drivers. The benchmarks show that there's little performance gain and, from what I can see, no discernible difference in image quality.

Tarinder, I really, really hope that is the truth.

Chris Angelini at tomshardware suddenly decided to do it by himself also - also after finding nothing in a previous review. I don't understand why reviewers are making these decisions now when they have looked and found no IQ loss.

I was VERY surprised to see this in your test setup. Perhaps you could remove this line in particular?

This, we hope, provides an apples-to-apples comparison to NVIDIA's default image quality.

The IQ of AMD's Quality setting is no worse - in fact in cases it'll be better - than nVidia's highest setting, as my links show.

As for your point about it not really showing anything - what if it had? What if it had been 5 fps in each game instead of 1 fps? You'd just have felt even more pressure to use HQ settings in the next review, even though there is no discernable difference between HQ and Quality.

[Tarinder]

There isn't a discernible difference in I.Q between AMD and NVIDIA, sure, but that doesn't mean AMD has kept the same settings as on Catalyst 10.9 drivers.

Here's AMD's response when I asked what the difference was between 10.9 and subsequent drivers, as pertaining to a Radeon HD 5870 and Radeon HD 6870:

"the following settings should be basically apples : apples in terms of image quality/optimization level –"

"Run 58xx on 10.9 with Catalyst AI "set to standard, versus 68xx on current driver with texture filter slider set to “High Quality”

AMD clearly admits that if you want to get Cat 10.9 I.Q you have to manually change the Catalyst CC settings.

This is what pains NVIDIA, that is, AMD has lowered its IQ. Now whether it's noticeable or not in fast game-play is actually a tangential issue.

FWIW. I think NV's default is now very, very slightly better than AMD's. That's purely subjective, of course, and based on poring over screenshots.

This is why I felt it fair to draw some attention to it in this review. I've said all I'm going to on the matter

[Terbinator]

I can't believe you let Nvidia pressure you into flunking the AMD results Taridner.

I thought Hexus reviews were the last bastion of neutrality against Nvidia's black (green) balling.

I'm never coming back to Hexus again!

[kalniel]

So you think this is one of those "fair" advantages? When many websites have looked at the issue, found pretty much nothing of note, yet all at the same time decided to review with HQ settings anyway? (it's not just Hexus btw).

Wake up Kalniel. nVidia put the heat on with this one, big time. Just like they did with the overclocked cards in the barts reviews.

You seem to be making out that it was nVidia's decision to force AMD use lower quality default settings in a way that doesn't affect most gamers. AMD did this, not nVidia.

And if the results of all this 'heat' and 'unfair advantage' is that Hexus provide a few pokey examples of the high IQ setting in their graphs then I can live with that. On the other hand, you don't even seem to want to read the review to form an accurate opinion in the matter, but I guess you mustn't let facts stand in the way of a pre-judgement eh?

[ehhhhhhh]

I'm more interested in deciding what will be my next card, rather than arguing over pity questions such as the ccc settings

CF 6870 seems a good performance boost over a single 6970 for around £50 more, though I'm not sure of the negative aspects involved in running 2 cards.

6970 can be had for 300, which is around 20-30% slower than a 6870CF setup....

on the nvidia side I'm not really sure which one is the viable alternative...

questions, questions

[Jimbo75]

Isn't the high quality 5-series is the same as Quality on the 6-series because of the improved AF algorithm in the 6-series? The AF has improved, that's for sure.

http://www.anandtech.com/show/3987/a...range-market/5

Scroll down and mouse over each and it should be pretty clear that Barts is every bit as good as the gtx 480.

AMD also said that their quality setting is "on par" with nVidia's in your recent article.

This whole IQ thing has been rumbling on for months, it started with nVidia calling the fp16 demotion a "cheat" and for me, it's a clear case of if they keep repeating something long enough people will actually start believing it.

[Terbinator]

Isn't the high quality 5-series is the same as Quality on the 6-series because of the improved AF algorithm in the 6-series? The AF has improved, that's for sure.

http://www.anandtech.com/show/3987/a...range-market/5

Scroll down and mouse over each and it should be pretty clear that Barts is every bit as good as the gtx 480.

AMD also said that their quality setting is "on par" with nVidia's in your recent article.

This whole IQ thing has been rumbling on for months, it started with nVidia calling the fp16 demotion a "cheat" and for me, it's a clear case of if they keep repeating something long enough people will actually start believing it.

I don't understand how you've made the jump from Hexus trying to verify Nvidias claims = Yes, 3 bags full sir.

The review even says they didn't notice any difference.

I'm glad they tested them as i asked in the IQ article comment section, i am an Nvidia employee also.

[Jimbo75]

I don't understand how you've made the hump from Hexus trying to verify Nvidias claims = Yes, 3 bags full sir.

The review even says they didn't notice any difference.

I'm glad they tested them as i asked in the IQ article comment section, i am an Nvidia employee also.

http://forums.hexus.net/2003565-post33.html

That was barely a month ago, can you see more or less consistency after today?

[Noxvayl]

Interesting release by AMD... Unfortunate that most places are out of stock. Looks like the HD6950 is a winner, particularly for my setup.

There is a mistake in the article I thought I should bring to your attention;

"This clearly means that its bang4buck is sweeter, though not quite as good as the GTX 570's." This is referring to the HD6950 on page 20, under Analysis, second paragraph.

Looked at the results table and found the HD6950 to have a bang4buck of 1.72 and the GTX570 had one of 1.67...

[chis]

Let's just keep in mind is that this is a brand new architecture. Even if it is based on previous efforts, do you all think the current drivers are offering best optimisation? The figures look roughly even right now, but if AMD are working hard to improve the drivers for this new and potentially untapped architecture, then the situation may very well change.

For now, I'm still deciding whether I should wait for a passive 67xx or just plump for Gigabyte's 5750...