When Tuesday does the G70 NDA expire?

Sorry I haven't been following ATI's code names so I just have to ask:
Is the R520 or R500 the PC part? (I suppose the other is the Xenos)
And what roughly pipes/configuratoin is the PC part?
What's the expected unveiling timeframe?

Thanks!
 
I don't think r520 will be able to really top the PS performance of G70. A little sad that G70 doesn't support fp16 AA though on the other side the SS transp. AA option is/looks really good.

AF is :?: right now as quality was decreased for current reviews at least there is some indication from the hardware.fr review if i'm interpreting it right so i'm not sure how to value the AF benchmarks for now.

DB branching performance? I guess we have to wait and see how it is improved or not.
 
The_Pope said:

How We Tested:

Please be aware that the way we test our video cards is not a like-for-like comparison, and it is not meant to be. We decided to concentrate on finding the “Best Playable” settings - this means that we're finding the best possible gaming experience delivered on each different configuration. There are no time demos used in our evaluations - we're focusing on the real-world gaming experience, which is, ultimately what should determine your next graphics card's purchase.
Congratulations, that's conceptually the best review I've read so far IMO.

Although it still leaves some things to be desired, e.g. different "types of gamers". Some would want 50 fps or so as minimum, some prefer AA over HDR, some want to get rid of all jaggies, using a combination of multisampling, supersampling and transparent AA, and some are limited to a certain resolution.
I know it's a lot of work, but at least I'd wished to see which are the best settings without HDR.
 
Bravo NV lots of nice things to like for those HighEnd users and I think they have stuck a nice balance with its features vrs limitations. If I had 500+ to spend and needed a new card, this would be what I would get!
 
tEd said:
I don't think r520 will be able to really top the PS performance of G70. A little sad that G70 doesn't support fp16 AA though on the other side the SS transp. AA option is/looks really good.

AF is :?: right now as quality was decreased for current reviews at least there is some indication from the hardware.fr review if i'm interpreting it right so i'm not sure how to value the AF benchmarks for now.

DB branching performance? I guess we have to wait and see how it is improved or not.

I am trying to figure out why yo dont spect R520 to have better PS performance since it really doesn't look that great on the GTX.

The fact that you(or me) dont think r520 to have better PS performance changes nothing on that hardware. I am just wondering how can you come to such beliefs without knowing anything.
 
The one option that temps me to buy this card is TSAA. Damn it looks good and fixes the single biggest problem in games today (or atleast the two games I play: HL2/CSS and BF2). Moreover, the technique being used is cheaper than supersampling the entire screen, and can be used to eventually fix other AA problems that multisampling can't fix, like aliasing introduced by shaders, such as the floor specular shaders in HL2.

Kudos to NVidia for offering supersampling options.
 
Excuse my Engrish, but what does that mean? :)
It means that to close the perfomance gap now created with the new G70, ATI wont have to break a sweat (increasing clockspeeds and adding Shader 3.0 compatbility across the board will do the job nicely), so yeah, you could say that the new cards from ATI & NVIDIA are just flexing out their respective architectures.

geo: yeah well, i read something different on iXBT, which i tend to trust a bit more than anyone else, since the're unbiased and probably the most extensive review site around. Ever since NV30's introduction, the branching pattern hasn't changed much for NVIDIA and it is no different for NV47 (static branching remains stable and fast while dynamic flounders, as opposed to ATI who for some reason remain steady across the board).
 
compres said:
tEd said:
I don't think r520 will be able to really top the PS performance of G70. A little sad that G70 doesn't support fp16 AA though on the other side the SS transp. AA option is/looks really good.

AF is :?: right now as quality was decreased for current reviews at least there is some indication from the hardware.fr review if i'm interpreting it right so i'm not sure how to value the AF benchmarks for now.

DB branching performance? I guess we have to wait and see how it is improved or not.

I am trying to figure out why yo dont spect R520 to have better PS performance since it really doesn't look that great on the GTX.

The fact that you(or me) dont think r520 to have better PS performance changes nothing on that hardware. I am just wondering how can you come to such beliefs without knowing anything.

From the raw shader test i often see a 50-120% gain to gf6ultra. Of course i can't predict exactly what r520 will offer but r520 has first to reach that 50-120% gain too.

Well i expect at least 50% clock advantage of r520 but with 16pipelines the big question is how much the ALU power increase per pipeline will be compared to r420. I guess it's around 33-50%.
 
alexsok said:
geo: yeah well, i read something different on iXBT, which i tend to trust a bit more than anyone else, since the're unbiased and probably the most extensive review site around. Ever since NV30's introduction, the branching pattern hasn't changed much for NVIDIA and it is no different for NV47 (static branching remains stable and fast while dynamic flounders, as opposed to ATI who for some reason remain steady across the board).

Hey, I'm not saying you're wrong, I just want to know how iXBT knows --they must have tested it somehow. I also think that it would be useful to have a benchmark that tried out various branching scenarios and intensiveness. And, y'know, this is the kind of place where suggesting a useful benchmark sometimes results in action when others agree with you. :D
 
alexsok said:
geo: yeah well, i read something different on iXBT, which i tend to trust a bit more than anyone else, since the're unbiased and probably the most extensive review site around. Ever since NV30's introduction, the branching pattern hasn't changed much for NVIDIA and it is no different for NV47 (static branching remains stable and fast while dynamic flounders, as opposed to ATI who for some reason remain steady across the board).
I'm not sure what you're referring to here. NV30 has dynamic branching only in the vertex shader, NV40 was the first chip to have dynamic branching in the pixel shader. And ATI has yet to present a chip that can do dynamic branching at all.
 
geo & XMAS
Hehe, now that's real embarassing i kinda misread what they've written there... you're right, ati are yet to introduce a chip that is capable of dynamic branching at all (either vs or ps) while on G70 ixbt have written the following excerpt that i unwittingly misread "the paradox lies in the fact that in NVIDIA's case, dynamic branching is MORE DESIRABLE than static branching"

Sorry guys :oops:

p.s
and yes geo if you're wondering, ixbt have an in-house testing app just for these kind of scenarios (dynamic vs static branching etc) it's included in their standard rightmark package. I just recalled when reading the review a few hours ago that there was a nuance on NV30 whereby dynamic branching suffered a major penalty blow so i guess they fixed it in this iteration.
 
I am curious if the texture flashing is still apparent with AF on this card. Reading Hardware.fr, the translation makes it seem that the 7800 does even less AF than the 6800? Is this right?

http://babelfish.altavista.com/babe...ttp://www.hardware.fr/articles/574/page5.html
"It is impossible to show to you this change on simple a screenshot, the only thing which comes out from it being that in places the texture seems less filtered and thus more accentuated, which involves this concern. It is with regrets that we note the appearance of this type of artifacts on a new top-of-the-range card,"
 
The one option that temps me to buy this card is TSAA. Damn it looks good and fixes the single biggest problem in games today
Yeah, many games suffer from this, and it's quite possibly the most annoying visual problem. I wonder if that type of AA can be used on HDR as well.

What exactly is the problem/limitation with AA being used to gether with HDR? Can't you at least use AA on the 'base' picture before you merge it with HDR buffer?
 
Graphics_Krazy said:
I am curious if the texture flashing is still apparent with AF on this card. Reading Hardware.fr, the translation makes it seem that the 7800 does even less AF than the 6800? Is this right?

http://babelfish.altavista.com/babe...ttp://www.hardware.fr/articles/574/page5.html
"It is impossible to show to you this change on simple a screenshot, the only thing which comes out from it being that in places the texture seems less filtered and thus more accentuated, which involves this concern. It is with regrets that we note the appearance of this type of artifacts on a new top-of-the-range card,"

I'm not sure what i should make of it but it's not promising to say at least
 
Probably the biggest reason AA+HDR isn't supported, besides needing more ROP logic, is bandwidth. Quite possibly, on anything less than a Xenos, AA+HDR performance will suck badly, unless you are talking about that half-HDR FP10 mode.
 
I'm wondering if the SS transp AA should also work on nv40 or does the G70 some special hardware support for it?
 
tEd said:
I'm wondering if the SS transp AA should also work on nv40 or does the G70 some special hardware support for it?
Doubt it. From the simplicity of the technique and the apparent "unfreshness" of the architecture, i'd take a venture and guess that it's more of a driver-related option that will open up for current-gen users after the new drivers are released and a reg-key is enabled.
 
alexsok said:
tEd said:
I'm wondering if the SS transp AA should also work on nv40 or does the G70 some special hardware support for it?
Doubt it. From the simplicity of the technique and the apparent "unfreshness" of the architecture, i'd take a venture and guess that it's more of a driver-related option that will open up for current-gen users after the new drivers are released and a reg-key is enabled.

To be effective, it does require new hardware. It requires centroid texture sampling that is multisample-mask aware.
 
Back
Top