ShaderMark NV40 vs R420

[Ante P]

By the way any idea why performance drops when going partial precision in the "PS 2.0 - Simple" shader?
I remember seeing it mentioned here at B3D before but I can't find the thread.

 

[Tridam]

can anyone here write some relevant PS 3.0 shaders to run with NV40?

 

[Hellbinder]

What about when dx9.1 comes out? Won't nVidia wipe the floor with ATi then? :|

In a Word... NO.

 

[Bob3D]

What about when dx9.1 comes out? Won't nVidia wipe the floor with ATi then? :|

In a Word... NO.

In a future...not far away...

What about when dx9.1 comes out? Won't nVidia wipe the floor with ATi then? :|

In a Word... YES.

 

[Stryyder]

Any gueses until how long this is featured as a Inquirer article without giving source credit? anyone?

 

[Evildeus]

ROFL

What about when dx9.1 comes out? Won't nVidia wipe the floor with ATi then? :|

In a Word... NO.

In a future...not far away...

What about when dx9.1 comes out? Won't nVidia wipe the floor with ATi then? :|

In a Word... YES.

 

[Eronarn]

Any gueses until how long this is featured as a Inquirer article without giving source credit? anyone?

Less than a day.

 

[digitalwanderer]

Any gueses until how long this is featured as a Inquirer article without giving source credit? anyone?

That depends, what were we talking about again? I kind of lost track...

Was it Rumsfeld or ShaderMark? :|

 

[Bob3D]

We are talking about a superior piece of hardware losing for a inferior piece of hardware, with a grain of hope about DX9.0c and shaders performance

 

[tEd]

why are there so huge results differences between sm2.0 and sm2.1?

 

[Ante P]

why are there so huge results differences between sm2.0 and sm2.1?

wth is sm2.1?

 

[tEd]

why are there so huge results differences between sm2.0 and sm2.1?

wth is sm2.1?

Shadermark 2.1 (beta) is one of them and here we test the performance with Pixel Shader 2.0

from nordichardware review

 

[Ante P]

why are there so huge results differences between sm2.0 and sm2.1?

wth is sm2.1?

Shadermark 2.1 (beta) is one of them and here we test the performance with Pixel Shader 2.0

from nordichardware review

hehe oh, I thought you were referring to shadermodel 2.1

in any case since I saw large differences in fillratetester once I installed 61.11 perhaps the same is true for shadermark 2.1

 

[digitalwanderer]

We are talking about a superior piece of hardware losing for a inferior piece of hardware, with a grain of hope about DX9.0c and shaders performance

Uhm...I think you're using the words "superior" and "inferior" incorrectly in your post, but I think I get your gist.

 

[Eronarn]

We are talking about a superior piece of hardware losing for a inferior piece of hardware, with a grain of hope about DX9.0c and shaders performance

Uhm...I think you're using the words "superior" and "inferior" incorrectly in your post, but I think I get your gist.

Oooh, ZING!

 

[oddfellow]

We are talking about a superior piece of hardware losing for a inferior piece of hardware, with a grain of hope about DX9.0c and shaders performance

Uhm...I think you're using the words "superior" and "inferior" incorrectly in your post, but I think I get your gist.

LOL

 

[Lezmaka]

They are wrong, I've updated them in an upcoming article:

By the way the second 6800 Ultra (the one on the right) is the Extreme version.

Can explain what's different between all 3 6800 ultras? You said "the second 6800 Ultra (the one on the right) is the Extreme version" but there's three. So is the extreme version the one on the right or the middle one?

By the way any idea why performance drops when going partial precision in the "PS 2.0 - Simple" shader?
I remember seeing it mentioned here at B3D before but I can't find the thread.

It was in Dave's preview of the NV40.

http://www.beyond3d.com/previews/nvidia/nv40/index.php?p=21

Looking at the 6800 Ultra results a little more closely though, we see some curious performances because in some cases the PS1.x and partial precision tests are slower than the PS2.0 tests! With the PS1.x and partial precision tests the internal precision is likely to be calculated at FP32 precision, but a type conversion must occur at some point - each of the shaders in this test are actually relatively, with the first few only using about 4 instructions - these are probably executed in two cycles and it may be the case that the the type conversion is not free and hence there is an extra cycle penalty for the partial precision and PS1.x integer shaders, meaning that the FP32 shaders are actually faster in these short shader cases.

 

[tb]

Ouch. How is it possible for NV40 to be so completely humiliated in those synthetic tests yet have a much better showing in actual games? And why can't the NV40 run many of the tests given it's full SM2.0 compliance?

I presume that this problem comes from a lack of some render target formats support. For example, NVIDIA doesn't support D3DFMT_R16F. Some tech demos are using this RT format as ATI has support for it for a while. However NVIDIA supports D3DFMT_R32F and D3DFMT_G16R16F. If the problem is the lack of D3DFMT_R16F support it's easy to use D3DFMT_R32F or D3DFMT_G16R16F instead. If this is the problem then I presume that the next revision of Shadermark will use one of these RT formats for NVIDIA GPU.

These limitations will be fixed in the upcomming update of ShaderMark, then you can compare NV4x and R3xx/R4xx with all shaders.
The most shader code will be altered a little bit, because of fixing some stuff, so v2.1 and v2.0 results are not comparable.....

Thomas

 

[Evildeus]

When do you think we will see this new version? This month, next month, later?

 

[tb]

When do you think we will see this new version? This month, next month, later?

Don't know. As it includes the latest HLSL compiler, I've to wait untill the dx9 sdk update (summer 2004) will be out of beta

Thomas