I dont mean to spread fud..

[overclocked]

Chalnoth wrote:

If there was any more performance to pull out of the R3xx architecture, then ATI has been damned slow on the driver optimization front. So I definitely expect no significant performance improvements on the R3xx.

I further suspect that the R420 is too close to the R3xx to show driver improvements, either. We'll see whether this is the case or not, of course, but yes, if these performance improvements are not across the board in shader benchmarks/games, then I suspect it's a cheat.

Well it´s good for all the R3xx owners altough it has performed great anyway, and i agree with you that they now would have "maxed out" the performance in their driver´s.
I also raised my eye when i see the kind of leap in performance with the 3D3k tests shown from the chart, i have "no doubt" that this is legitime but after all the fuzz about 3DMark and the whole situation that where with the NV30 i guess most people including myself get´s suspecius.
The inrease in FarCry is expected as they tune their driver`s for new games and nVidia has had many driver-related problems also with FarCry.
But as i said 3dmark scores get´s me a little "paranoid"..

jvd wrote:

Well I disagree with you . Where does it say the r420 isn't much of a change from the r3x0 series . Do you think they spent the last 2 years scratching thier ass and all they came up was lets double the amount of pipelines of the r3x0 ?

Well they did double the pipelines..

 

[jpaana]

And doubled the number of alus per pipe...

 

[Ostsol]

And doubled the number of alus per pipe...

Actually, not really. From what I've read around on this forum, the R3xx series had a texture sampling unit, a full vec3 + scalar pair, and a mini vec3 + scalar pair. That mini just got upgraded to a full, on the R420.

 

[Headstone]

Wouldnt just the fact there was a process change between the 3xx and 4xx make a difference in how they optimize the drivers as well? The layouts are not the same to begin with and then you add in the changes that were done. I think that there will be a lot more speed to come out of the X800 series starting with new drivers to coincide with the XT launch to market.

 

[KimB]

Wouldnt just the fact there was a process change between the 3xx and 4xx make a difference in how they optimize the drivers as well?

Not in the least. That's a hardware design issue.

 

[PeterAce]

And doubled the number of alus per pipe...

Actually, not really. From what I've read around on this forum, the R3xx series had a texture sampling unit, a full vec3 + scalar pair, and a mini vec3 + scalar pair. That mini just got upgraded to a full, on the R420.

The extra mini got upgeaded to an extra full alu? Where did you here this?

Here is R300 shader core (Eric Demers (Sireric) slide) :

And here is R420 shader core :

And Dave's blurb from his X800 Review :

The pixel shader core in R420 remains largely the same as R300, with a texture address unit, and two ALU's each of which are co-issue capable, and dual issue capable together - this can result in up to 5 floating point operations executed per cycle in the best case scenario. The second ALU is fully featured, however the first is a smaller ALU which ATI haven't fully detailed its capabilities - it does feature PS1.4 input modifiers but ATI state that it has other instruction capabilities, however we don't know what they are.

and

though the Shader core, for the most part, remains the same as R300's, which fundamentally means that it's still a Shader 2.0 shader core, it isn't exactly the same and some modification have been made.

Was one of these modifications a second Full ALU?

 

[Dave Baumann]

The question will be answered when Cat 4.6 Become available, By then I think some real comparison should be made between 4.5, Beta 4.5 and 4.6

From what I'm hearing it looks like some increase will come in at 4.6, but it'll probably be 4.7 that it fully comes in.

 

[Tim]

From what I'm hearing it looks like some increase will come in at 4.6, but it'll probably be 4.7 that it fully comes in.

Then the 4.5 beta is special benchmark drivers based on very early betas, Ati should have a good hiding for using this kind of tactic.

I think most has perception that the 4.5 betas was whql-candidates, the reality was that the 4.5 beta had very little to do with the real 4.5 cats.

 

[jvd]

From what I'm hearing it looks like some increase will come in at 4.6, but it'll probably be 4.7 that it fully comes in.

Then the 4.5 beta is special benchmark drivers based on very early betas, Ati should have a good hiding for using this kind of tactic.

I think most has perception that the 4.5 betas was whql-candidates, the reality was that the 4.5 beta had very little to do with the real 4.5 cats.

Why does it make them cheating drivers ?

Perhaps the build that went on to be the 4.6s were finished shortly before the 4.5 betas . The 4.6s had some of the speed increase but what wil lbecome the 4.7s have all of it as they are based after the 4.5 betas . A later build so to say

 

[Dave Baumann]

Then the 4.5 beta is special benchmark drivers based on very early betas, Ati should have a good hiding for using this kind of tactic.

Well, its a driver thats getting some of the specific opimistations for R420, of which you'll see a couple of releases later. Timeframes are pushed out for the freezing WHQL drivers so I don't think it unreasonable they they put out some beta's with R420 optimisations - I still think this is better than the somewhat haphazard review / WHQL release pattern (?) from NVIDIA.

 

[Ostsol]

And doubled the number of alus per pipe...

Actually, not really. From what I've read around on this forum, the R3xx series had a texture sampling unit, a full vec3 + scalar pair, and a mini vec3 + scalar pair. That mini just got upgraded to a full, on the R420.

The extra mini got upgeaded to an extra full alu? Where did you here this?

Here is R300 shader core (Eric Demers (Sireric) slide) :

http://www.beyond3d.com/reviews/ati/radeon9700pro/20040302_r300alu.jpg

And here is R420 shader core :

http://www.beyond3d.com/reviews/ati/r420_x800/diagram/shadercore.jpg

And Dave's blurb from his X800 Review :

The pixel shader core in R420 remains largely the same as R300, with a texture address unit, and two ALU's each of which are co-issue capable, and dual issue capable together - this can result in up to 5 floating point operations executed per cycle in the best case scenario. The second ALU is fully featured, however the first is a smaller ALU which ATI haven't fully detailed its capabilities - it does feature PS1.4 input modifiers but ATI state that it has other instruction capabilities, however we don't know what they are.

and

though the Shader core, for the most part, remains the same as R300's, which fundamentally means that it's still a Shader 2.0 shader core, it isn't exactly the same and some modification have been made.

Was one of these modifications a second Full ALU?

Oops! I read too fast. I read "The second ALU is fully featured" and assumed that it was the second ALU on the R3xx that was the mini. In that case, I guess the number of ALUs per pipe hasn't changed. . .

 

[Dave Baumann]

Basic PS ALU configuration stays the same.

 

[Tim]

Well, its a driver thats getting some of the specific opimistations for R420, of which you'll see a couple of releases later. Timeframes are pushed out for the freezing WHQL drivers so I don't think it unreasonable they they put out some beta's with R420 optimisations - I still think this is better than the somewhat haphazard review / WHQL release pattern (?) from NVIDIA.

If something is named 4.5 beta it reasonable to assume that it is a beta of the 4.5 cats, the reality is that they are very different. This clearly misleading, Ati should have made it clear that the drivers provided had optimizations that would not be available in official driver releases for 2-3 months, not in a week or two like the version number indicated.

 

[tEd]

Well, its a driver thats getting some of the specific opimistations for R420, of which you'll see a couple of releases later. Timeframes are pushed out for the freezing WHQL drivers so I don't think it unreasonable they they put out some beta's with R420 optimisations - I still think this is better than the somewhat haphazard review / WHQL release pattern (?) from NVIDIA.

If something is named 4.5 beta it reasonable to assume that it is a beta of the 4.5 cats, the reality is that they are very different. This clearly misleading, Ati should have made it clear that the drivers provided had optimizations that would not be available in official driver releases for 2-3 months, not in a week or two like the version number indicated.

If you buy a x800 then you get those beta drivers on the cd so what you see in the reviews you get right away

 

[Tim]

If you buy a x800 then you get those beta drivers on the cd so what you see in the reviews you get right away

But then you will running with drivers that contains optimizations that Ati does not consider ready for an official release before in a release or two later, when running such drivers bugs should be expected.

 

[jvd]

If you buy a x800 then you get those beta drivers on the cd so what you see in the reviews you get right away

But then you will running with drivers that contains optimizations that Ati does not consider ready for an official release before in a release or two later, when running such drivers bugs should be expected.

Who said that ? The 4.5s that we have offical are most likely 3 weeks older than when they were released.

The 4.5s betas were most likely done the day before a review . The 4.6s were most likely finalized and sent to microsoft a little before that .


Its much better than nvdiai . hey here are our magic drivers. Test with these . Oh btw they are never coming out .

 

[Evildeus]

If you buy a x800 then you get those beta drivers on the cd so what you see in the reviews you get right away

That's not true, at least in France, you have 4.4 with the CD.

 

[Bjorn]

Its much better than nvdiai . hey here are our magic drivers. Test with these . Oh btw they are never coming out .

"Better then NVidia" doesn't automatically means that it's good practice. And how do we know that even the 4.7 drivers will have all of the 4.5 betas optimizations ?

 

[sireric]

Our shader compiler is getting better with every release. The performance increases from it will benefit all cards, including 9700's. Some upcoming updates will increase performance > 10% for longer shaders. For most of the current batch of applications, the shaders are pretty short already, so I don't know if they will get much benefit. We also have other optimizations that will benefit various cards in different ways (96's, X800's vs. 97's) -- I don't know most of them.

As for the X800 series, it's getting a lot of its upcoming performance benefits from our work in optimizing the memory controller. We've got a huge amount of flexibility in how it can be programmed and used, and we've just scratched the surface (the state space is huge, and it will take some time to come up with optimal cases). People can expect significant (>>10%) performance increases, esp for AA.

 

[AlphaWolf]

If you buy a x800 then you get those beta drivers on the cd so what you see in the reviews you get right away

That's not true, at least in France, you have 4.4 with the CD.

Huh?

Do the catalyst 4.4 drivers even contain support for X800?

I think you may be confusing a driver version number (6.14.10.6444), with the catalyst naming scheme.