shadermark 2.0 tomorrow!

xtreme2k said:
Natoma said:
Hyp-X,

Is that a 256MB 9800 Pro you've got? If it is, that would explain why your scores in the HDR benches corrolates with mine rather than the other 128MB 9800 Pro scores.

Make sure you are running your direct3d's setting texture preference and mipmap level is at 'High Quality'. It is because I find myself with your HDR score if I lower my quality as well.

I have all my texture settings in the display panel maxed.
 
Natoma said:
xtreme2k said:
Natoma said:
Hyp-X,

Is that a 256MB 9800 Pro you've got? If it is, that would explain why your scores in the HDR benches corrolates with mine rather than the other 128MB 9800 Pro scores.

Make sure you are running your direct3d's setting texture preference and mipmap level is at 'High Quality'. It is because I find myself with your HDR score if I lower my quality as well.

I have all my texture settings in the display panel maxed.

Yep that thats sorted its the memory size issue. :D
 
vrecan said:
ANTI-DETECT-MODE fights against shader and texture specific optimisations <--- is the part that is the most interesting to me. Cant wait to see what will come of this.

Anti-Detect-Mode works on both vertex and pixel shaders. How it works:

first version:
-insert NOP instruction
PRO's: worked on r3xx vpu's
CON's: errors on NV3x vpu's

second version:
- insert/double instructions which don't do any harm to the shader result
PRO's: worked on all vpu's
CON's: more instructions, slower shaders and/or going beyond shader instruction limit

third version(used in ShaderMark v2.0):
- shuffle some instructions on a random base, which have no needed order for execution
- shuffle R register's, like r1 to r3 on a random base
PRO's: worked on all vpu's
CON's: may be easy to detect by driver coders

So the ADM will be modified in future versions/updates of ShaderMark v2.0 in order to have some impact. It's a race....

Regards,
Thomas
 
xtreme2k said:
Natoma said:
xtreme2k said:
Natoma said:
Hyp-X,

Is that a 256MB 9800 Pro you've got? If it is, that would explain why your scores in the HDR benches corrolates with mine rather than the other 128MB 9800 Pro scores.

Make sure you are running your direct3d's setting texture preference and mipmap level is at 'High Quality'. It is because I find myself with your HDR score if I lower my quality as well.

I have all my texture settings in the display panel maxed.

Yep that thats sorted its the memory size issue. :D

So again I ask. Is this a portent of things to come wrt HDR perf in HL2? A 25% gain in performance going from a 128MB R350 to a 256MB R350 is nothing to sneeze at in a game where you're averaging 60fps with no AA/AF with a P42.8C. :LOL:

Especially when I want to run 4xAA and 16xAF. ;)
 
Just checked.

Shader 13 to 21 will gain significant frame rate if you move the d3d texture pref and mipmap slider down even 1 notch.

All of the test will then look very weird visually...
 
Tim said:
lost said:
I am curious, could someone try a 5900 with older drivers to compare them to the 51.75 and 52.10 to see what sort of legit PS speed improvments have gone on? TIA!

I second that, I would really like to see how the Det50 compares to the Det40s.

my test: leadtek 5900 (450/850)- xp1700+@stock - 45.33

totale.png
 
xtreme2k said:
tb said:
Natoma said:
Hyp-X,

Is that a 256MB 9800 Pro you've got? If it is, that would explain why your scores in the HDR benches corrolates with mine rather than the other 128MB 9800 Pro scores.

The Shader 20 and 21 are very sensitive to the onboard memory size, because they use large 16 bit fp render targets. Maybe future driver could improve this situation, but I'am not sure.

radeon 9800 pro 128 MB
~41 fps
~44 fps

radeon 9800 pro 256 MB
~56 fps
~61 fps

The problem is if i lower my mipmap and and texture preference in my display control panel is that I get exactly the same score as they do on a 256MB card.

So I just wana make sure this is a MEMORY issue or just a display setting issue.

just checked my settings, fresh installed catalyst drivers always set the quality one notch bellow maximum :(

Thomas
 
xtreme2k said:
Just checked.

Shader 13 to 21 will gain significant frame rate if you move the d3d texture pref and mipmap slider down even 1 notch.

All of the test will then look very weird visually...

Do I smell a future nvidia "optimization" in the works? ;)
 
ballero, did you run the test with partial precision? It doesn't look like it, because your numbers match tEd's "Partial Precision ON" results, except for tests 7 and 12.

Code:
Test    45.33   51.75   52.1    9800PRO (3.7)
 2       105     101     133     232
 3       72      70      85      160
 4                               163
 5       61      57      73      130
 6       76      71      82      170
 7       65      118     143     148
 8                               127
 9        29      27      39      116
 10      132     130     163     238
 11      116     113     132     206
 12      58      49      69      139
 13      35      34      40      87
 14      40      38      53      87
 15      53      52      52      125
 16                      40      76
 17      5       5       6       13
 18      23      21      30      105
 19                              27
 20                              47
 21                              50
 22                              36
 23                              52

NOTE: These are full precision results. 51.75 and 52.1 are from tEd, 45.33 from ballero, 9800 Pro from xtreme2k.

It looks like the 51.75 only improved test 7, while slowing down test 12. 52.1, on the other hand, has improvements across the board. Still very far from the 9800 PRO's speeds, at least in full precision.
 
OpenGL guy said:
tb said:
just checked my settings, fresh installed catalyst drivers always set the quality one notch bellow maximum :(
That should only be true for the OpenGL control panel.

btw why is that so? why don't they just set it to max quality as default like in d3d?
 
Using 45.33 Drivers (Anti-detect On then Off)... I'm not sure how to enable _PP; I get a message stating I cannot do this with the free version. Oh well.

Code:
(1024x768) X8R8G8B8 (D24X8) vram used 56623104
HAL (pure hw vp): NVIDIA GeForce FX 5900 Ultra (Anti-Detect-Mode: off, gamma correction: DAC)



options

pixel shader version: 2_0
partial precision: off
number of render targets: 1

results:
shader  2 (                                            Per Pixel Diffuse Lighting):	 101 fps	  9.9087 mspf	     505 rendered frames
shader  3 (                            Per Pixel Directional Light Shader (Phong)):	  70 fps	 14.2913 mspf	     350 rendered frames
shader  5 (                                   Per Pixel Spot Light Shader (Phong)):	  57 fps	 17.6607 mspf	     284 rendered frames
shader  6 (                                        Per Pixel Anisotropic Lighting):	  71 fps	 14.1230 mspf	     355 rendered frames
shader  7 (                                         Per Pixel Fresnel Reflections):	  60 fps	 16.6437 mspf	     301 rendered frames
shader  9 (                                          Per Pixel Car Surface Shader):	  26 fps	 38.3021 mspf	     131 rendered frames
shader 10 (                                         Per Pixel Environment Mapping):	 130 fps	  7.6747 mspf	     652 rendered frames
shader 11 (                                    Per Pixel Environment Bump Mapping):	 113 fps	  8.8496 mspf	     565 rendered frames
shader 12 (                                                Per Pixel Bump Mapping):	  50 fps	 20.1928 mspf	     248 rendered frames
shader 13 (                                       Per Pixel Shadowed Bump Mapping):	  34 fps	 29.5858 mspf	     169 rendered frames
shader 14 (                                        Per Pixel Veined Marble Shader):	  38 fps	 26.4083 mspf	     190 rendered frames
shader 15 (                                                 Per Pixel Wood Shader):	  52 fps	 19.0485 mspf	     263 rendered frames
shader 17 (                                  Fur Shader With Anisotropic Lighting):	   5 fps	209.0658 mspf	      24 rendered frames
shader 18 (        Per Pixel Refraction and Reflection Shader with Phong Lighting):	  21 fps	 46.8020 mspf	     107 rendered frames

_____________________________________________________________

video mode / device info
(1024x768) X8R8G8B8 (D24X8) vram used 56623104
HAL (pure hw vp): NVIDIA GeForce FX 5900 Ultra (Anti-Detect-Mode: on, gamma correction: DAC)



options

pixel shader version: 2_0
partial precision: off
number of render targets: 1

results:
shader  2 (                                            Per Pixel Diffuse Lighting):	 101 fps	  9.9049 mspf	     505 rendered frames
shader  3 (                            Per Pixel Directional Light Shader (Phong)):	  69 fps	 14.5406 mspf	     344 rendered frames
shader  5 (                                   Per Pixel Spot Light Shader (Phong)):	  56 fps	 17.8144 mspf	     281 rendered frames
shader  6 (                                        Per Pixel Anisotropic Lighting):	  71 fps	 14.1408 mspf	     354 rendered frames
shader  7 (                                         Per Pixel Fresnel Reflections):	  60 fps	 16.6373 mspf	     301 rendered frames
shader  9 (                                          Per Pixel Car Surface Shader):	  26 fps	 38.7019 mspf	     130 rendered frames
shader 10 (                                         Per Pixel Environment Mapping):	 130 fps	  7.6777 mspf	     652 rendered frames
shader 11 (                                    Per Pixel Environment Bump Mapping):	 113 fps	  8.8496 mspf	     565 rendered frames
shader 12 (                                                Per Pixel Bump Mapping):	  54 fps	 18.4790 mspf	     271 rendered frames
shader 13 (                                       Per Pixel Shadowed Bump Mapping):	  34 fps	 29.5726 mspf	     170 rendered frames
shader 14 (                                        Per Pixel Veined Marble Shader):	  38 fps	 26.2291 mspf	     191 rendered frames
shader 15 (                                                 Per Pixel Wood Shader):	  52 fps	 19.0485 mspf	     263 rendered frames
shader 17 (                                  Fur Shader With Anisotropic Lighting):	   5 fps	209.0658 mspf	      24 rendered frames
shader 18 (        Per Pixel Refraction and Reflection Shader with Phong Lighting):	  21 fps	 47.4830 mspf	     106 rendered frames
 
My 9600Pro vs the 5900Ultra results as above
Code:
  2 (                                            Per Pixel Diffuse Lighting):; 133;fps;            666;rendered  101
  3 (                            Per Pixel Directional Light Shader (Phong)):;  92;fps;           461;rendered 	70
  4 (                                  Per Pixel Point Light Shader (Phong)):;  94;fps;           469;rendered 	na
  5 (                                   Per Pixel Spot Light Shader (Phong)):;  75;fps;           377;rendered 	57
  6 (                                        Per Pixel Anisotropic Lighting):;  98;fps;           489;rendered 	71
  7 (                                         Per Pixel Fresnel Reflections):;  85;fps;           426;rendered 	60
  8 (                             Per Pixel BRDF-Phong/Anisotropic Lighting):;  74;fps;           373;rendered 	na
  9 (                                          Per Pixel Car Surface Shader):;  67;fps;           337;rendered 	26
 10 (                                         Per Pixel Environment Mapping):; 135;fps;            676;rendered 	130
 11 (                                    Per Pixel Environment Bump Mapping):; 119;fps;            593;rendered 	113
 12 (                                                Per Pixel Bump Mapping):;   81;fps;           404;rendered   54										
 13 (                                       Per Pixel Shadowed Bump Mapping):;  60;fps;           300;rendered 	50
 14 (                                        Per Pixel Veined Marble Shader):;  65;fps;           326;rendered 	34
 15 (                                                 Per Pixel Wood Shader):;  79;fps;           395;rendered 	52
 16 (                                                 Per Pixel Tile Shader):;  44;fps;           218;rendered 	na
 17 (                                  Fur Shader With Anisotropic Lighting):;   8;fps;           39;rendered      5
 18 (        Per Pixel Refraction and Reflection Shader with Phong Lighting):;  59;fps;           294;rendered    21
 19 (  Dual Depth Shadow Mapping With 3x3 Bilinear Percentage Closer Filter):;  15;fps;            78;rendered 	na
 20 (                                High Dynamic Range Shader (cross blur)):;  32;fps;           160;rendered 	na
 21 (                             High Dynamic Range Shader (gaussian blur)):;  36;fps;           182;rendered 	na
 22 (                          Per Pixel Edge Detection And Hatching Shader):;  21;fps;           105;rendered 	na
 23 (                                         Per Pixel Water Colour Shader):;  23;fps;           118;rendered 	na
 
Man, can't believe my 9500 Pro (and even a 9600 Pro) are getting higher framerates in all of the tests over a GFFX 5900 Ultra. Really brings to light Nvidia's shader problems.
 
sorry,but i think there's some problem in the result tEd provided.
51.75 should have score in shader16.(while 45.23 doesnt)
the 51.75 result is more likely to be 45.23.

btw,i got about half score in shader7.it's unusual ,too.

my test result:
smtemp.gif
 
Natoma said:
Hyp-X,

Is that a 256MB 9800 Pro you've got? If it is, that would explain why your scores in the HDR benches corrolates with mine rather than the other 128MB 9800 Pro scores.

Yes, it's a 256MB card.
Default clocks, all settings in Application / High quality.
 
Back
Top