Real Doom 3 performance boost

  • Thread starter Thread starter ranger
  • Start date Start date
R

ranger

http://www.beyond3d.com/forum/viewtopic.php?t=14874

Forget the meager tweaks posting everywhere. This is a simple working
Shader mod for ATI that addresses one of the real problems why ATI cards
are underperforming in Doom 3.

Saw this and thought, yeah right, another tweak that does very little.
Big surprise though - this really made a VERY nice improvement overall.
Very nice smooth performance on high quality settings now with 9800pro
on P4 2.4 (400bus). Definitely improves frame rates, smoothness,
visuals, etc., with no sparkles or atifacting either. No more choppiness
when the big doors open on a new area with baddies leaping at you. Some
of the heaviest places and events in the game are now much, much
smoother. The actual frame rate increase is not large for me but
replaying heavy areas you'd think the frame rates had doubled! Amazingly
smooth. I've made a spare copy of the new pak000 on CD so I don't lose
it - it's now a must have.

A couple of tips - make a backup copy of the pak000 and store outside
the Base folder first. Also, before you start, go to My Computer, then
Tools, then Folder Options, then File Types, and create a new file type
called "VFP" and set it to open with NotePad (for example). Then use
WinRAR (if you don't have it you can do a Google search for it and DL
the freeware version - it's a puny DL), to follow the simple directions
for modifying this Shader. The directions in the article are simple and
completely accurate. Follow them and you're in business.

Get this in your Base folder. It's real.
 
ranger said:
http://www.beyond3d.com/forum/viewtopic.php?t=14874

Forget the meager tweaks posting everywhere. This is a simple working
Shader mod for ATI that addresses one of the real problems why ATI cards
are underperforming in Doom 3.
Click to expand...

Tried it and it works very nicely. I gained 5fps in timedemo with this tweak
and
after giving D3 a 10 minute run I saw no anomalies. Thanks for the pointer
Ranger
and a big thanks to Humus.
 
Tried it and it works very nicely. I gained 5fps in timedemo with
this tweak and
after giving D3 a 10 minute run I saw no anomalies. Thanks for the
pointer Ranger
and a big thanks to Humus.
Click to expand...

Same here. 5fps additional in timedemo, but the increase in normal play seems to be greater - as
Humus says in his post, the timedemo makes a lot of use of the fireball and particle effects that
only occur occasionally ingame, so the real increase should be higher - it certanly feels that way
ingame.


S
 
Sham B said:
Same here. 5fps additional in timedemo, but the increase in normal play seems to be greater - as
Humus says in his post, the timedemo makes a lot of use of the fireball and particle effects that
only occur occasionally ingame, so the real increase should be higher - it certanly feels that way
ingame.
Click to expand...

hmm.. doesnt seem to make a scrap of difference on my 9800pro..
followed the instructions exactly - 40fps before and after
then downloaded his latest version and tried that - 39.60
turned off v-sync makes absolutely no difference
at all, same framerates off or on..
 
seems to be greater - as


and particle effects that


certanly feels that way



hmm.. doesnt seem to make a scrap of difference on my 9800pro..
followed the instructions exactly - 40fps before and after
then downloaded his latest version and tried that - 39.60
turned off v-sync makes absolutely no difference
at all, same framerates off or on..
Click to expand...
Try running at 16x AF.
I don't know really how this works, but I got these results,
Before:
High 1024x768 8x AF CP = 46.5 fps
High 1024x768 16x AF CP= 33.2 fps

After:
With shader tweak no AA:
High 1024x768 8x AF CP = 46.5 fps
High 1024x768 16x AF CP = 46.4 fps
 
Sleepy said:
Tried it and it works very nicely. I gained 5fps in timedemo with this tweak
and after giving D3 a 10 minute run I saw no anomalies. Thanks for the pointer
Ranger and a big thanks to Humus.
Click to expand...
After further play I can see this fix isnt perfect - in Alpha Labs 2 when
youre
escorting the Scientist in the dark - torchlight doesnt work properly.
 
After further play I can see this fix isnt perfect - in Alpha Labs 2 when
youre
escorting the Scientist in the dark - torchlight doesnt work properly.
Click to expand...

To me a performance boost = in a reduce load of processing in the gpu so it is normale to experience
less quality in lightning effect or else.
 
Judas Pissed said:
hmm.. doesnt seem to make a scrap of difference on my 9800pro..
followed the instructions exactly - 40fps before and after
then downloaded his latest version and tried that - 39.60
turned off v-sync makes absolutely no difference
at all, same framerates off or on..
Click to expand...

Me too. No big difference.
 
After further play I can see this fix isnt perfect - in Alpha Labs 2 when
To me a performance boost = in a reduce load of processing in the gpu so it is normale to experience
less quality in lightning effect or else.
Click to expand...
In that section I found the torch didnt work AT ALL. Its pitch black because
the powers out
and the torch doesnt illuminate the ground right in front of you. I
eventually got past that section and the torch
works again but the problem may reoccur so think twice about hacking around
with your pak file.
 
There would be no point if you are running at 40FPS already and explosions
which may have slower FPS aren't helped anyway.... apparently the human
brain can really only see up to around 30FPS anyway. Therefore IMO a more
useful measure in a benchmark should be the minimum FPS rather than average?

ie. If during an explosion sequence it saw 5 FPS but elsewhere it saw 80 FPS
then the 5FPS tells me that the card sucks whereas if another card saw a
minimum of 15 FPS but elsewhere it "only" got 40 FPS then the minimum is 3
times the performance of the other, even though the other card's performance
sounds about twice as fast thanks to the average.
 
If you pay attention, the caracter told you some magnetic pule kills all
light, default or updated code, same happens. Well on my machine anyway.
 
Give me a break - Im NOT talking about the lights that go out
when the power'[s cut - Im talking the friggin torch not working
properly in that section of the game after Ive dont the shader code hack.
 
Bad information, the rate of decay of the display has far more to do with
seeing 30fps or whatever than the human brain. And the framerate affects
more than just visuals, low framerates are hard to play.

Mike
 
apparently the human brain can really only see up to around 30FPS anyway.
Click to expand...

Rubbish - if you turn your monitors refresh rate down to 60hz (on a CRT)
you can see the flicker as the picture is redrawn 60 times a second proving
that your eye sees a lot more than 30 fps. The 30fps myth stems from the
fact
that films are run at around 24-30fps and appear smooth. They appear smooth
because the images are not static but include blurred images and that
'looks' okay
to your brain. I think it tricks the brain and the brain fills in the blanks
..
 
Erik said:
Wow that was a gigantic boost.
In timedemo1 I went from 31.9fps to 46.2 fps (8xAF).
A64 3500+
9800XT (omega 4.7)
1280x1024 ultra settings.
Click to expand...


How do I download that rar file? I click on it and it takes me to a
page with a bunch of garble! ANyone have a link or can send it to me?
If so please email it to (e-mail address removed) Thanks
 
The change actually replaces a table lookup with a math function, so the GPU
workload is increased but texture access is reduced. The problem is that one
of the exponents used in the function is based on Humus' intuition, so the
end result is not completely equivalent. Someone with inner knowledge of the
game (i.e. an iD employee) can rewrite the shader with the correct exponent
and get equal, if not better image quality while maintaining the speed
boost... on a Radeon.

Remember, this change works to Radeon's advantage. It may do nothing or
reduce performance on the Geforce.
 
Back
Top