Official AMD Radeon 6000 Series Discussion Thread

[EastCoasthandle]

Personally I don't like MLAA, because it blurs almost everything which means that many textures are blurred when they shouldn't be and it also produces some artifacts on some places. For example, in this image from your thread:

http://i11.photobucket.com/albums/a154/eastcoasthandle/CSAA/MLAA.png

Look at that lamp post on the lower left corner. Instead of smoothing the edges out, MLAA has interpreted the edges as actual geometry and has enhanced them instead of making them dissapear. There's also the fact that MLAA and DLAA seem to work against anisotropic filtering, it's somethin that I've seen in most pictures. All the pictures around the net are showing the same problems so far and for me they are something that deters me from using it. I have never liked CSAA or CFAA for the exact same reason. MSAA + alpha textures is IMO still the best all around AA method by far, but I use SSAA whenever posible. Of course in cases where MSAA and SSAA is not posible, MLAA is probably better than no AA, but that's not even true in all cases: I've seen some comparisons at high resolutions where the picture without AA looked better to me, because MLAA was blurring out every detail in the scene indiscriminately and was actually changing the real shape of geometry.

MLAA is a good promise for the future, but it needs A LOT of work before I can even consider it as an alternative to MSAA.

EDIT: I have to thank you for pointing out that MLAA or similar techniques are used on consoles. Now I finally know why the hell console games look so freaking blurred and look like crap and washed out. Blurring things out at 1920x1200 is one thing (an acceptable compromise), but blurring out a picture that has actually been rendered at 720p or less, just to get it blurred again by the upscaling to 1080p is a whole different thing...

Here's your answer. MLAA is not something limited to driver implementation. Developers can incorporate it into their games and refine how MLAA works. Which in the end would make it look better. If developers don't use it we still have the option of enabling it through CCC.

 

[Tatty_Two]

I agree so much i want your apes red waistcoat.

This is a tweak of a card, nothing more. The 69xx cards offer more in the way of change, allegedly so we'll see what they can do at the end of the month.

But until both companies move to 28nm, not much is really all that new.

There ya go, I like happy members ................

 

[Benetanegia]

Here's your answer. MLAA is not something limited to driver implementation. Developers can incorporate it into their games and refine how MLAA works. Which in the end would make it look better. If developers don't use it we still have the option of enabling it through CCC.

That does open some doors, but will developers really delve into that? I'm not sure, but I'm more inclined to believe they will not. What it is explained in the link, yes, they can do that, but IMO developers are not going to do anything but apply it as a post effect, the only thing that will change is when they apply it. They'll simply apply it to the entire scene, then put the HUD on top of it, but they will not create complicate algorithms to use it only on places where it's really needed. That's what MLAA is supposed to do anyway, it's just that it fails on doing so.

My reasoning is based on the fact that developers have access to refine MSAA too but very few or nearly none have done anything with it. There are some that will do it, in the end I don't think this MLAA is very different than the EdgeAA in Crysis, the AA used in Batman or the AAA in Metro2033, seing as how the end result in the picture is pretty similar.

I have always wanted Ati/Nvidia to make SSAA posible, but I understand that MSAA made more sense. Nowadays the performance hit of MSAA is not that high and when combined with alpha AA they offer the best quality (after SSAA). I just don't really see a place for these post process AA methods in a world that releases new more powerful cards every year, and games are simply console ports that can easily run on midrange cards. Cards are powerfull enough to use MSAA in most cases, so a method that offers worse quality is not needed in my book, no matter how easy on the hardware it is. If they can get away with the errors and the blurring of textures, then sure why not, but I don't see it happening because no matter when it is used by developers, this technique is nothing but a post process localized blurring effect.

 

[EastCoasthandle]

That does open some doors, but will developers really delve into that? I'm not sure, but I'm more inclined to believe they will not. What it is explained in the link, yes, they can do that, but IMO developers are not going to do anything but apply it as a post effect, the only thing that will change is when they apply it. They'll simply apply it to the entire scene, then put the HUD on top of it, but they will not create complicate algorithms to use it only on places where it's really needed. That's what MLAA is supposed to do anyway, it's just that it fails on doing so.

My reasoning is based on the fact that developers have access to refine MSAA too but very few or nearly none have done anything with it. There are some that will do it, in the end I don't think this MLAA is very different than the EdgeAA in Crysis, the AA used in Batman or the AAA in Metro2033, seing as how the end result in the picture is pretty similar.

I have always wanted Ati/Nvidia to make SSAA posible, but I understand that MSAA made more sense. Nowadays the performance hit of MSAA is not that high and when combined with alpha AA they offer the best quality (after SSAA). I just don't really see a place for these post process AA methods in a world that releases new more powerful cards every year, and games are simply console ports that can easily run on midrange cards. Cards are powerfull enough to use MSAA in most cases, so a method that offers worse quality is not needed in my book, no matter how easy on the hardware it is. If they can get away with the errors and the blurring of textures, then sure why not, but I don't see it happening because no matter when it is used by developers, this technique is nothing but a post process localized blurring effect.

Games like GOW3, TFU2, Kill Zone 3, etc will be using M/D LAA. Here's another article discussing MLAA. We know that Star Wars Force Unleashed II is using DLAA (per my earlier link) for the console. But for the PC it's a mystery as the AA option for that game only reads Antialiasing. I'm not going to assume that DLAA is in the PC version. There are more games that will be using either MLAA or DLAA. So it's a reality at this point and it does look like developers will be using it.

 

[AltecV1]

OMG i just checked if there is any 6000 series available in my country and there was a gigabyte HD 6870.....for 285.37 EUR what a ripoff

 

[mastrdrver]

mastrdrver, go read the review on anandtech. MUCH more detailed and technical information. 68xx isn't just a simple 'refresh', there is much more to it then that. You need to read about all the features Barts and Cayman have, Cypress is definitely old news in comparison.

Been there and was the first I read. Also read TR and PCPer.

I prefer the technical to the benchmarks.

In some respects, I look at it like the 4870 > 4890 (but this is a lot more than that with this release) and that took even longer, lot's of people said then "it's just a refresh" but hey, the end product was a damn nice card IMO, people are only dissapointed because they expect too much in the first place, rarely has there been two major architectural breakthrough's by the same company in a year to be honest and I see no reason to beleive this will change significantly, it normally takes 18 months to 2 years to see something that significant...... the "refreshes" simply give the company that bit of extra time and breathing space to get the real new stuff sorted...... just my thoughts.

I just really dont know what some expect.... damn two HD6850's in crossfire faster than a 480 or 5970 (or as fast as) for half the price (of the 5970).... is this not good enough?

I guess too I keep forgetting that this is more of a 57xx replacement (sort of). When looking at it from that perspective it is quite an impressive jump.

I for one don't really care about the added AA modes. Maybe it will change this time around but they've yet to add one that doesn't take a big enough hit in performance. Add that to (at least myself) the difference in image quality is minor for the performance hit taken. I will say though that I would be interested in the new AA that will actually work on all edges in things like BC2. I play that game a lot and the lack of AA on most of the game just makes my eyes bleed.

 

[Paintface]

http://www.newegg.com/Product/Product.aspx?Item=N82E16814102908

Any 6850 out there with better custom cooling?

 

[pantherx12]

Can someone who has a 5870 now measure the heatsink mounting holes for me?

 

[Tatty_Two]

I guess too I keep forgetting that this is more of a 57xx replacement (sort of). When looking at it from that perspective it is quite an impressive jump.

Yes, my only point in general was, when you think of the 6850 as a 5750 replacement, running 2 of them in crossfire was roughly equivilent to a 5850's performance, now the same market segment has it's replacement cards (albeit to be fair they are currently more expensive) in HD5970 territory..... that is an absolutely HUGE jump. I would expect to see prices drop for these cards in time, it's not unusal to see inflated prices at initial release with limited availability but to be honest they will probably never sit in the same price bracket as the 5750 does/did.

Even so, if they eventually come down to the 5770 price area (2 of which roughly equalled a 5870) we are still seeing a Huge bang for buck in performance terms alone.

 

[pantherx12]

I can't see them getting as cheap, I think lowest 6870 will get to is £150 or so

Whilst fab process is less shitty now, the transistor count is up from the 5770 on the same fab process.

So won't ever be as cheap.

Unless I've gone wrong somewhere in my thinking.

 

[RoutedScripter]

6850 costs 186 euros in my country , crap price.

 

[Delta6326]

http://www.newegg.com/Product/Product.aspx?Item=N82E16814102908

Any 6850 out there with better custom cooling?

It has great aftermarket than stock. when in stock i would get this one

 

[TheMailMan78]

Ok enough of this trash. Where is my 6950?!

 

[bear jesus]

Ok enough of this trash. Where is my 6950?!

It better be at least a few weeks away, i want to enjoy my coming pair of 6870's before i'm forced to get a 6970

 

[Steevo]

With just a upgrade install of the 10.10 drivers I have no option for MLAA, it doesn't even occur in the registry, or xml files. However MLA does, as does quite a few other options, some old, some new.


I ran Dirt2 last night with every setting I unlocked at max quality. I couldn't find a jaggie anywhere. I only got 20-30FPS though.

 

[TheoneandonlyMrK]

did you sort the hole measure 5870 thing panther?

 

[pantherx12]

did you sort the hole measure 5870 thing panther?

I meant 6870

 

[mdsx1950]

Ok enough of this trash. Where is my 6990?!

Corrected!

Dammit! I can't wait!

 

[inferKNOX]

I can't wait to see the 6900s, although I'm really disappointed that they need more than 2x6pin.
If I do skip the HD6000s, it'll probably be for that reason, although I might make a side-grade if it's free.

Just to add the final spank to a beaten horse, I must disagree with what seems like all of you on the naming thing. I don't think that the 6800s are a 5700 successor, nor 5800 for that matter; reason being that the 5700s are now gonna be 6700s (so they are in essence not replaced) and, well, obviously the 6800s are not 5800 replacers. I think they're meant to be in-betweeners, ie, they're trying to stuff as many cards in the near $200 price range as they can, since that is the most lucrative, so they move the 5800 successors up a notch (to 6900), leave the 5700s in their place (as 6700s), and stick the 6800s in between.
Anyway, that's just what I think...

 

[TheMailMan78]

What I think? I think this guy is AWESOME!

Source

 

[yogurt_21]

What I think? I think this guy is AWESOME!

http://cdn.wg.uproxx.com/wp-content/uploads/2010/07/opulence-300x177.jpg

Source

well if he likes savings ze monies he should get a 6870, on top of all it's other features it offers nearly twice the model number per dollar over the 5870.

6870/$240 = 28.625 model numbers per dollar
5870/$400 (launch price) = 14.675 model numbers per dollar
6850/$180 = 38.05 model numbers per dollar
5850/$300 (launch price) = 19.5 model numbers per dollar.

so as you can see, the 6870 and 6850 are actually better thanth 5850 and 5870. I mean cmon, the numbers don't lie.

 

[Goodman]

well if he likes savings ze monies he should get a 6870, on top of all it's other features it offers nearly twice the model number per dollar over the 5870.

6870/$240 = 28.625 model numbers per dollar
5870/$400 (launch price) = 14.675 model numbers per dollar
6850/$180 = 38.05 model numbers per dollar
5850/$300 (launch price) = 19.5 model numbers per dollar.

so as you can see, the 6870 and 6850 are actually better thanth 5850 and 5870. I mean cmon, the numbers don't lie.

The Price of the 6850/70 would have to be compare to the 5770/50 or the 5830 & not the 5870/50...

Anyhow i putting some money aside for the 6850 sometime next month or early December

 

[reverze]

Just ordered this...

http://www.newegg.com/Product/Product.aspx?Item=N82E16814150506

I'm ordering all kind of stuff Never enough video cards!

 

[CDdude55]

Just ordered this...

http://www.newegg.com/Product/Product.aspx?Item=N82E16814150506

I'm ordering all kind of stuff Never enough video cards!

Very nice.

I love the look of the XFX version.

 

[JATownes]

well if he likes savings ze monies he should get a 6870, on top of all it's other features it offers nearly twice the model number per dollar over the 5870.

6870/$240 = 28.625 model numbers per dollar
5870/$400 (launch price) = 14.675 model numbers per dollar
6850/$180 = 38.05 model numbers per dollar
5850/$300 (launch price) = 19.5 model numbers per dollar.

so as you can see, the 6870 and 6850 are actually better thanth 5850 and 5870. I mean cmon, the numbers don't lie.

HAHAHAHAHA. That, my friend, literally made me laugh out loud. Twice the model number per dollar... HAHAHAHAHA