r420 may beat nv40 in doom3 with anti-aliasing

[pocketmoon_]

There's still some tuning work to be done on the X800, especially for AA... Expect large improvements with AA as we find more optimal settings. Stay tuned.

I think it's fair to say the same for NV40.

 

[Dave Baumann]

While OpenGL guy can speak for R420, I'm not sure he's in quite the same position to speak for NV40

 

[pocketmoon_]

While OpenGL guy can speak for R420, I'm not sure he's in quite the same position to speak for NV40

I woun't expect him to be But Nvidia are facing the same challenges with what is a much 'fresher' architecture.

I'd eat my virtual hat if we don't see ' large improvements ' in NV40 AA

 

[Dave Baumann]

It all depends on where the largest changes are - NV40's shader arcitecture is significantly different, but are its ROP's and memory bus?

The memory interface is one of the single largest areas of change between R300 and R420.

 

[Quitch]

LOL, it used to be "flexible" and now it's "fresher"

 

[Luminescent]

Does NV40 also have a programmable memory controller?

 

[Dave Baumann]

Although I've not heard anything for definite, I'd imagine that they are all programmable to some degree - although whether they are configurable by the driver or set in the BIOS is another question. I'd guess at NVIDIA being somewhat more up to speed with the latencies of dual bank memory schemes used in GDDR3 because of their greater use of GDDR2, although having multiple quads attempting to access the bus might cause some headaches.

I think one of the biggest indications that ATI still has some "discovery" to do with their memory bus is from the lacklustre X800 PRO performance relative to 9800 XT in SS:SE - logically, from a processing capabilities POV there should be no reasons why the R420 board should be in this position.

 

[pocketmoon_]

LOL, it used to be "flexible" and now it's "fresher"

Fresher as in newer, as in less time to produce optimised drivers for.

 

[Quitch]

I know how you were spinning it

 

[anaqer]

Fresher as in newer, as in less time to produce optimised drivers for.

And I thought you were just phaking da phunk. W3rd, yo!

 

[vb]

There's still some tuning work to be done on the X800, especially for AA. As someone mentioned in a review, the memory controller is very programmable, however it's not always obvious what the best settings are (i.e. what's best for one mode may not be best for another) and the combinations are enourmous. Expect large improvements with AA as we find more optimal settings. Stay tuned.

Can r420 change states on aplication start, during app when changing options (AA/AF) or only at boot times?

 

[OpenGL guy]

There's still some tuning work to be done on the X800, especially for AA. As someone mentioned in a review, the memory controller is very programmable, however it's not always obvious what the best settings are (i.e. what's best for one mode may not be best for another) and the combinations are enourmous. Expect large improvements with AA as we find more optimal settings. Stay tuned.

Can r420 change states on aplication start, during app when changing options (AA/AF) or only at boot times?

If a driver update can change the behavior, then I'd say it's not limited to boot time

 

[Pete]

Boot-time optimization might make benchmarking more of a chore than it already is, no?

 

[Dave Baumann]

Wait until they give us application profiles and allow us to set our own memory tuning!

 

[nutball]

Wait until they give us application profiles and allow us to set our own memory tuning!

I think that's almost the definition of time to go do something else isn't it? :?

 

[Bambers]

I think one of the biggest indications that ATI still has some "discovery" to do with their memory bus is from the lacklustre X800 PRO performance relative to 9800 XT in SS:SE - logically, from a processing capabilities POV there should be no reasons why the R420 board should be in this position.

Whats up with the 9800pro vs 9700pro in that first graph, is that an error? :?

 

[Eronarn]

I think one of the biggest indications that ATI still has some "discovery" to do with their memory bus is from the lacklustre X800 PRO performance relative to 9800 XT in SS:SE - logically, from a processing capabilities POV there should be no reasons why the R420 board should be in this position.

Whats up with the 9800pro vs 9700pro in that first graph, is that an error? :?

I'm willing to bet it was just a slip of the keyboard or... something. It'd be scary if my 9700 were beating out a 9800!

 

[Dave Baumann]

Mmmm, looks like AF may still be on with the 9800 PRO results. Error.

 

[AAlcHemY]

There's still some tuning work to be done on the X800, especially for AA... Expect large improvements with AA as we find more optimal settings. Stay tuned.

Will that be available for normal users, through registry tweaks? ( Adjusting some settings for the memory controller )

 

[pat777]

It all depends on where the largest changes are - NV40's shader arcitecture is significantly different, but are its ROP's and memory bus?

The memory interface is one of the single largest areas of change between R300 and R420.

R3x0 was king of shader architecture last generation while the 59x0 had the most memory bandwidth. ATI focused on bandwidth(programmable memory controller)in this generation while nVIDIA focused on shader architecture. I like how both companies focused on their weakpoints.