Does the Xenos benefit from general ATI optimizations?

Squilliam

Beyond3d isn't defined yet
Veteran
Supporter
For the desktop ATI gpus we can definately see performance improvements over time with different driver revisions etc. I was curious to see whether or not these improvements carried over to the Xbox360 due to the different but similar unified shader architecture.

So if a game was being coded now would the developers have any additional available GPU performance now than 3 years ago due to general ATI development?
 
Gosh, this question comes up every once in a while, doesn't it!

No. The drivers are about allowing PC developers to use the hardware more efficiently. On a console there are no 'drivers' to get in the way. There will instead be improvement with both MS's API's and with developers learning new techniques that better use the hardware, but PC driver improvements is totally independent from the console space.
 
There is no driver yes, but the hlsl compiler still continue to be optimized. MS API still continue to give more and more access with new low level API (not really new features, but ways to optimize CPU or command buffer size).

And don't forget improvement to PIX and some other tools that give devs all the key to understand, find bottleneck (and why it's one) in their GPU or CPU frame.
 
One little question, in regard to xenos ALU.

Does they work in the same way as current ATI?
I mean they try to take advantage from both TLP and ILP.

In this case perfs must be sensitive to shader code, and I guess there is a lot of room for optimisation.
 
i don't know about the current ATI hardware, but x360 shaders performance is more a question of GPR count than ALU. Since this count drive the amount of thread that can be run in //
 
One little question, in regard to xenos ALU.

Does they work in the same way as current ATI?
I mean they try to take advantage from both TLP and ILP.

In this case perfs must be sensitive to shader code, and I guess there is a lot of room for optimisation.

No, Xenos' ALU aren't as flexible as R6x0+'s, they aren't "scalars", they're vec4+1 (co-issue) instead of VLIW 5-way.
 
One little question, in regard to xenos ALU.

Does they work in the same way as current ATI?
I mean they try to take advantage from both TLP and ILP.

In this case perfs must be sensitive to shader code, and I guess there is a lot of room for optimisation.
As was already stated R600 and beyond are more flexible with ILP, but Xenos still takes advantage of TLP and ILP.
 
Back
Top