Timedemo

[Annihilation]

AMD Athlon 64 X2 4600+ 2G RAM
NVidia GeForce 6600GT
1280x1024x32
4*AA (export __GL_FSAA_MODE=8 )
full gfx settings

3067 frames, 79.6 seconds: 38.5 fps

Interesting difference:

Intel Core 2 Duo E6700 @ 2666Mhz + 8G RAM
NVidia GeForce 8800 GTS 640Mb
1680x1050
max settings

3067 frames 36.2 seconds 84.8 fps

Turning aniso/FSAA on or off doesn't make any difference.
Latest mg binary, 64bit linux

That is odd, my first guess would be ram.  I didn't know ram could alter fps that much though.  So the onlything to assume is AMD does make shitty cpus

[Archangel]

Look at the video cards. 6600GT versus 8800 GTS? The 8800 can smoke the 6600.

I tried out the VBO patch -- Doesn't apply for me. There was a decent amount of errors (mostly SSE conflicts, but I dealt with those). The issue I have right now is

Code: [Select]

src/renderer/tr_marks.c: In function ‘R_BoxSurfaces_r_sse’:
src/renderer/tr_marks.c:260: error: ‘v4fMZeroDotFive’ undeclared (first use in this function)
src/renderer/tr_marks.c:260: error: (Each undeclared identifier is reported only once
src/renderer/tr_marks.c:260: error: for each function it appears in.)
src/renderer/tr_marks.c: In function ‘R_MarkFragments_sse’:
src/renderer/tr_marks.c:507: error: ‘mixMask0001’ undeclared (first use in this function)
src/renderer/tr_marks.c:507: error: too few arguments to function ‘v4fMix’
src/renderer/tr_marks.c:512: error: too few arguments to function ‘v4fMix’
src/renderer/tr_marks.c:514: error: too few arguments to function ‘v4fMix’
src/renderer/tr_marks.c:565: error: ‘v4fMZeroDotOne’ undeclared (first use in this function)
src/renderer/tr_marks.c:609: error: ‘v4fMZeroDotFive’ undeclared (first use in this function)

v4fMZeroDotOne, v4fMZeroDotFive, and mixMask0001 aren't defined anywhere in the patch. Is the patch incomplete?

[kevlarman]

trem doesn't put any real load on the graphics card as long as it's not a geforce 2 or something.

[Bissig]

trem doesn't put any real load on the graphics card as long as it's not a geforce 2 or something.

Maybe not in the timedemo, which would explain why I had no change after upgrading gpu. But, ingame I now get 400 fps on some parts of some maps with same settings at which my old gpu would max out at 70-90 fps and I never go below the 76 fps I set except for on UNCREATION looking down the hall from humans base. But, at least the map Insanity gives me playable fps now ;-P

Is the timedemo code and the actual ingame code really working on the base of the same engine restrictions?

To me it seems like the timedemo is more CPU dependent whereas ingame it is more dependent on the GPU.

[Annihilation]

trem doesn't put any real load on the graphics card as long as it's not a geforce 2 or something.

Maybe not in the timedemo, which would explain why I had no change after upgrading gpu. But, ingame I now get 400 fps on some parts of some maps with same settings at which my old gpu would max out at 70-90 fps and I never go below the 76 fps I set except for on UNCREATION looking down the hall from humans base. But, at least the map Insanity gives me playable fps now ;-P

Is the timedemo code and the actual ingame code really working on the base of the same engine restrictions?

To me it seems like the timedemo is more CPU dependent whereas ingame it is more dependent on the GPU.

Imo, in game is still more dependant on CPU.

If you have an extremely nice card and a P2, expect shit fps, but if you have a nice E8500 and just a 64 mb integrated you'll still get nice fps.(not 400 but a good 60+ consistantly on atcs at least.)

[gimhael]

Look at the video cards. 6600GT versus 8800 GTS? The 8800 can smoke the 6600.

I tried out the VBO patch -- Doesn't apply for me. There was a decent amount of errors (mostly SSE conflicts, but I dealt with those). The issue I have right now is

Code: [Select]

src/renderer/tr_marks.c: In function ‘R_BoxSurfaces_r_sse’:
src/renderer/tr_marks.c:260: error: ‘v4fMZeroDotFive’ undeclared (first use in this function)
src/renderer/tr_marks.c:260: error: (Each undeclared identifier is reported only once
src/renderer/tr_marks.c:260: error: for each function it appears in.)
src/renderer/tr_marks.c: In function ‘R_MarkFragments_sse’:
src/renderer/tr_marks.c:507: error: ‘mixMask0001’ undeclared (first use in this function)
src/renderer/tr_marks.c:507: error: too few arguments to function ‘v4fMix’
src/renderer/tr_marks.c:512: error: too few arguments to function ‘v4fMix’
src/renderer/tr_marks.c:514: error: too few arguments to function ‘v4fMix’
src/renderer/tr_marks.c:565: error: ‘v4fMZeroDotOne’ undeclared (first use in this function)
src/renderer/tr_marks.c:609: error: ‘v4fMZeroDotFive’ undeclared (first use in this function)

v4fMZeroDotOne, v4fMZeroDotFive, and mixMask0001 aren't defined anywhere in the patch. Is the patch incomplete?

Sure, forgot to include qsse.h and qsse.c again. The complete patch is on the patch tracker now.

If you want to test it on tremfusion, you should use this patch.