[H] gunning for NV again.

[Regs]

There's not a single game out yet with those types of Pixel shader programs. Far cry uses 10-15 instruction sets at the most for one of its pixel shader programs let along 12k. The efficiency of PS3 is really not needed considering the ATI 9800 can handle enough instructions in one pass. And I don't know what you're talking about, but im playing Far Cry (PS2) with PLAYABLE FRAMERATES. Lets not over glorify PS3. But sooner or later, it will be needed, I just don't see it in the foreseeable near future.

But games can quickly change.

 

[Acanthus]

Originally posted by: Regs
There's not a single game out yet with those types of Pixel shader programs. Far cry uses 10-15 instruction sets at the most for one of its pixel shader programs let along 12k. The efficiency of PS3 is really not needed considering the ATI 9800 can handle over 500 instructions in one pass. And I don't know what you're talking about, but im playing Far Cry (PS2) with PLAYABLE FRAMERATES. Lets not over glorify PS3. But sooner or later, it will be needed, I just don't see it in the foreseeable near future.

But games can quickly change.

Far cry at 1600x1200, and everything cranked, over 45min fps? I seriously doubt it, but if im wrong, call me on it.

 

[Regs]

Originally posted by: Acanthus

Originally posted by: Regs
There's not a single game out yet with those types of Pixel shader programs. Far cry uses 10-15 instruction sets at the most for one of its pixel shader programs let along 12k. The efficiency of PS3 is really not needed considering the ATI 9800 can handle over 500 instructions in one pass. And I don't know what you're talking about, but im playing Far Cry (PS2) with PLAYABLE FRAMERATES. Lets not over glorify PS3. But sooner or later, it will be needed, I just don't see it in the foreseeable near future.

But games can quickly change.

Far cry at 1600x1200, and everything cranked, over 45min fps? I seriously doubt it, but if im wrong, call me on it.

Well of course not. I play 1280x960 with everything cranked. But what does that have to do with PS3? Extra registers and PS3 isnt going to help a game with small Pixel shader programs.

And I do retract one statement, the earlier generation video cards cannot handle over 500 pixel shader instuctions at once.

 

[Acanthus]

sorry the 16x12 was uneccesary, thats what i play at normally and i wasnt thinking.

 

[Regs]

I'm just more concerned that the RV420 cant push FP32 that is required by DX9c. The IQ differences will be noticable. But just like FX owners, they'll try to down play it.

 

[Acanthus]

Originally posted by: Regs
I'm just more concerned that the RV420 cant push FP32 that is required by DX9c. The IQ differences will be noticable. But just like FX owners, they'll try to down play it.

The difference wont be huge. It has been compared before i believe when NV was running 32 on the NV3X and the R9800 was running 24, and it looked the same.

Edit: Although that could change, now that the shaders are being written with FP32 in mind in the future.

 

[Regs]

Originally posted by: Acanthus

Originally posted by: Regs
I'm just more concerned that the RV420 cant push FP32 that is required by DX9c. The IQ differences will be noticable. But just like FX owners, they'll try to down play it.

The difference wont be huge. It has been compared before i believe when NV was running 32 on the NV3X and the R9800 was running 24, and it looked the same.

Edit: Although that could change, now that the shaders are being written with FP32 in mind in the future.

Yet these are till just rumors and ATI can come out with a card that fully supports all features of DX9c.

I still don't think HardOCP was gunning for Nvidia.

 

[Saist]

Correct me if I'm wrong on this, but one of the big things when DX9 came out was that Displacement mapping was supported in the DX9 spec, and that ATi's R3xx series fully supported it, and that Nv3x did not.

Since when has this changed?

 

[remagavon]

Unless there is a performance increase when using PS3 vs 2, then I'd probably stick with ATi. nVidia's FSAA is better but still not up to par with ATi, and you have to use it in 4x mode for it to be any good. I'd have to see their anisotropy in person to determine how it looks though, but I'm satisfied with ATi's.

 

[jim1976]

Originally posted by: Acanthus

Originally posted by: Ackmed
They did that before the 6800U preview. Try actually reading, then commenting. The fact is, PS 3.0 is almost impossible to make look better than PS 2.0. PS 3.0 is not a big deal, get over it already.

Youre preaching to me about reading when youre not listening to what im saying either cheif.

SM3.0 doesnt look much better, ITS FASTER, NOT BETTER LOOKING.

PLAYABLE FRAMERATES ARE YOUR FRIEND!

I do agree with you, BUT:

1. PS2 can do for know almost everything that ps3 can do.
2. We don't know R420's benchies for ps2 performance (but they should be higher right?)

1+2= For now the same result

 

[zephyrprime]

Here is what i know about SM2.0 vs 3.0

Say you have a shader and its 500 instructions.

6800 on SM3.0 can do this in a single pass.
NV3X/R3XX/R420 (if it doesnt support SM3.0) on SM2.0 can do this in 6 passes in a best case scenario (if they can break it up to multiple 9X instruction count shaders). This will be an enormous performance hit.

I don't think that's how things work. As far as I know, on the R3xx it would take at least 250 cycles to execute a 500 instruction shader program regardless of wether it was PS2 or PS3. The NV3X would take at least 500 cycles. No idea regarding the R420. How fast a shader program executes is more dependant on hardware than on shader standards.

Heres in NVidia's own words are the differences between PS2 and PS3.

link

The only performance benefits listed are those gained by dynamic branching and "multiple render targets"

I imagine that the benefit gained from dynamic branching is illustrated by this crude example:

Without dynamic branching:

a = some calculations
b = some different calcuations

c = ((case == m)*0xff) & a + ((case == n)*0xff) & b

(forgive me if this is only understandable by programmers.)
Now with dynamic branching:

if (case == m)
c = some calculations
else if (case == n)
c = some different calcuations

See? You don't have to do redundant calculations or any of the stupid bit masking stuff when you have dynamic branching available. In this crude illustration, the use of dynamic branching doubles performance (assuming the hardware platform has no visible latencies [which isn't a realistic assumption]).

I think the biggest advantage of PS3 isn't speed but ease of use and versatility. Any programmers reading this no doubt take dynamic branching for granted. Programming without dynamic branching seriously sucks. Also, having dynamic branching allows for making shaders which were previously impractical due to performance hits caused by calculating too many different cases now practical.

 

[BFG10K]

The difference now BFG, is now its playable frames, not 15fps vs 30fps

What are you basing this on? Show me the performance of PS 3.0 benchmarks compared to idenctical PS 2.0 benchmarks and then you can make that claim.

Until then you are simply accepting nVidia's marketing without any evidence to support it.

 

[Acanthus]

I cant prove how much faster SM3.0 is, but i can prove how fast 2.0 isnt