I dont mean to spread fud..

[Heathen]

And how do we know that even the 4.7 drivers will have all of the 4.5 betas optimizations ?

How do we know they won't?

 

[Bjorn]

And how do we know that even the 4.7 drivers will have all of the 4.5 betas optimizations ?

How do we know they won't?

Yes, the point is, we don't know.

 

[Hellbinder]

Except for the simple fact that the X800 Pro is essentially the same basic architecture as the previous R3xx chips, so it seems rather hard to believe that significant performance increases could be had at this stage in the game (at almost two years since the R300 was released). But, I suppose we'll find out sooner or later.

Dude that is absolute total FUD.

I mean come on people...

The X800 has more paralellism in their Pipeline than even the Nv40.. It has many times more instruction processing than any version of the R300. Sure it may be *based* on the R300 but come on. Its like you are intentionally ignoring the Architcture and processing in each Quad, ignoring the improved Vertex shaders, ignoring the increased pipelines etc etc...

 

[bloodbob]

I personally think we should wait for the drivers to come out that include these optimisations and hopefully they aren't IQ dropping or application/shader/texture/mesh specific optimisations. If the speed increases don't come out in few drivers then we should get up them for their pre-sales marketing.

 

[Mendel]

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.

-cut-

As for the X800 series, it's getting a lot of its upcoming performance benefits from our work in optimizing the memory controller.

Here is one r97p owner that you just got a little bit happier! Now I can hopefully stand to stay with it until i can raise enough money for x800...
Btw, would you care to be more specific in what do you consider "longer" shader? If you could mention any current or known future game or benchmark that would have them? Finally about r97p, is Ati aware of Colourless' multipass Ruby demo wrapper for r300 and how do you think these driver updates will benefit it? Would that be good example of the "longer" shaders?

And yes, making lots of effort to optimize the memory controller of x800 seems to make sense since a lot of games are so cpu limited that you only start to see difference in speed between competitors - and even your own previous generator cards - and your current lineup when hitting the AA to the max.

Ati's driver department has come a long way since the introduction of Catalyst brand. Keep it up! 8)

 

[Dave Baumann]

The 4.5 drivers were feature and performance frozen a while ago..late feb/early mar or so. We had done a ton of work specifically for X800 - that did not make it to 4.5. The 4.5 beta contains some of these optimizations.

Development-wise we are usually several months ahead of the release. For eg:- some of the optimizations we already coded and tested will not make it even to the next catalyst - since the next catalyst was frozen a couple of weeks ago. Under special circumstances(customer requests, launches etc) we do cherry pick some changes(usually new features or performance optimizations) from the current development tree and move them into the release tree and call it a "branch". Note that show-stopper bugs can still go into a "release" tree during this process.

 

[Drak]

Does anybody find ati's version numbering a bit weird? Traditionally, a beta would be made and released before the Release Candidate. But now, we find that the 4.5 betas are actually newer than the 4.5 RCs that were sent to microsoft for WHQL.

 

[Xmas]

The X800 has more paralellism in their Pipeline than even the Nv40..

How did you measure this "parallelism"?

 

[sireric]

Here is one r97p owner that you just got a little bit happier! Now I can hopefully stand to stay with it until i can raise enough money for x800...
Btw, would you care to be more specific in what do you consider "longer" shader? If you could mention any current or known future game or benchmark that would have them? Finally about r97p, is Ati aware of Colourless' multipass Ruby demo wrapper for r300 and how do you think these driver updates will benefit it? Would that be good example of the "longer" shaders?

Well, for <10 instruction shaders, unless they are badly written, the compilers we have can probably do a good job. For longer shaders, say >20, there's a good chance that the new compilers will do more to advance things, since there is more to do. Of course, some shaders will benefit more than others (a serial dependancy stream cannot be optimized very much, for example).

And yes, making lots of effort to optimize the memory controller of x800 seems to make sense since a lot of games are so cpu limited that you only start to see difference in speed between competitors - and even your own previous generator cards - and your current lineup when hitting the AA to the max.

Yes, there's still a lot of work there, on both the X800P & X800XT. Probably won't get even at 90% until later in the summer.

 

[webmedic]

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.

Thanks sireric. This means my 9600xt will see some improvement also. Will these things end up in 4.6 or 4.7 then?

 

[Sxotty]

Yes, there's still a lot of work there, on both the X800P & X800XT. Probably won't get even at 90% until later in the summer.

At first I was thinking that is pretty crappy b/c by the time you get the performance people will have new cards, but then I realized that people who buy them now are doing it b/c they are satisfied with the current price/perfornance ratio and so cannot complain.

 

[sireric]

Yes, there's still a lot of work there, on both the X800P & X800XT. Probably won't get even at 90% until later in the summer.

At first I was thinking that is pretty crappy b/c by the time you get the performance people will have new cards, but then I realized that people who buy them now are doing it b/c they are satisfied with the current price/perfornance ratio and so cannot complain.

We certainly wish we could introduce the cards with the highest performance, for many reasons

However, the reality is that we get something out that's reasonable (which hopefully meets price/performance goals), and then keep on improving it.

 

[tEd]

Yes, there's still a lot of work there, on both the X800P & X800XT. Probably won't get even at 90% until later in the summer.

At first I was thinking that is pretty crappy b/c by the time you get the performance people will have new cards, but then I realized that people who buy them now are doing it b/c they are satisfied with the current price/perfornance ratio and so cannot complain.

We certainly wish we could introduce the cards with the highest performance, for many reasons

However, the reality is that we get something out that's reasonable (which hopefully meets price/performance goals), and then keep on improving it.

so the programmabilty of the memory controller is new compared to r3x0?

 

[jvd]

Yes, there's still a lot of work there, on both the X800P & X800XT. Probably won't get even at 90% until later in the summer.

At first I was thinking that is pretty crappy b/c by the time you get the performance people will have new cards, but then I realized that people who buy them now are doing it b/c they are satisfied with the current price/perfornance ratio and so cannot complain.

Right I have a 9700pro that i got almost 2 years ago. The x800xt offers 2-3 times the performance of my 9700pro. Its a great time for me to upgrade. Not only that but if we do see alot more performance out of the cards with aa then i could see a 2.5-3.5 jump or even a 3-4 jump depending on the game . My 9700pro has gotten much faster over the first day i owned it .

 

[Socos]

Except for the simple fact that the X800 Pro is essentially the same basic architecture as the previous R3xx chips, so it seems rather hard to believe that significant performance increases could be had at this stage in the game (at almost two years since the R300 was released). But, I suppose we'll find out sooner or later.

oh whatever. You didnt say that about the GF4 over the GF3, or anything else.
Stop spouting nonsense. There are plenty of areas for optimization, some of which have been discussed by ATI in these very forums.

I think what Chalnoth is so eloquently trying to say is that ATI has added clipping planes and other such useful enhancing "OPTImizations"

 

[dizietsma]

When I first saw the Gt4 test and it's 20% increase I did raise an eyebrow as this is such massive jump, but I cannot believe Ati would do the same sort of cheat as has been done before by the " graphics contestents" in the 3dmark arena. So I think it is valid. However, of course you'd hope for 20% improvements in other DX9 titles as well, so that will be the acid test.

Unless of course there is a bug of course that Ati have not spotted, I'd laugh if after 6 months someone posts, " hey, what happened to the turtle .. it't gone, dragothic stylie "

 

[Randell]

At first I was thinking that is pretty crappy b/c by the time you get the performance people will have new cards, but then I realized that people who buy them now are doing it b/c they are satisfied with the current price/perfornance ratio and so cannot complain.

I'm sorry what new cards will these people have? Are you suggesting X800Pro or XT owners will already have replaced their cards in a couple of months time?

 

[Snyder]

Does anybody find ati's version numbering a bit weird? Traditionally, a beta would be made and released before the Release Candidate. But now, we find that the 4.5 betas are actually newer than the 4.5 RCs that were sent to microsoft for WHQL.

Sure, "traditionally" it would. But with a 1-month release cycle, a traditional release policy is out of the question, isn't it?

 

[Tahir2]

I have a suggestion ATI: slow down your drive release schedule! 12 drivers per year is too much..

edit: grammar

 

[BRiT]

I have a suggestion ATI: slow down your drive release schedule! 12 drivers per year is too much..

edit: grammar

No doubt. There should only ever be 1 driver per year.