Doom 3 shader mod for ++ performance

[^eMpTy^]

Since everyone seems to be only concerned about ATI's OEM business now I will attempt to bring the thread back on topic.

I have an AthlonXP2500+ Barton, 512 DDR400 Corsair XMSPro mem, and an ATI 9800 Pro running the 4.8 CATs. Before the REVISED hack I was getting 28.9FPS. After this hack I am not getting 34.8fps in timedemo1. I run at 1024x768 High Quality. And I also did the image_cachemegs thing and changed it to 96.

I see absolutely to degradation in image quality at all.

I think the big question is can this be fixed in a driver release to give it some sort of official status and get some quality control done...

 

[burningrave101]

He's a software engineer for ATi who is not on thier driver team and not working on Doom3...you may find this impressive...but I do not.

He did this in his spare time...the HACK is completely unofficial...it is supported neither by id nor by ATi...and it requires you to open and modify compiled files in your Doom3 installation...that my friend is a hack if I have ever seen one...you can church it up all you want to, but a hack is a hack...call it what it is.

And unless it gets officially supported by ID then it shouldn't be used in any review comparison between the 6800's and X800's for DOOM 3.

So the fact that Humus came up with ANOTHER cute little trick is pretty meaningless unless someone supports it and gets it included in a patch.

The majority of DOOM 3 users will never know about it.

 

[SuperRob]

He's a software engineer for ATi who is not on thier driver team and not working on Doom3...you may find this impressive...but I do not.

He did this in his spare time...the HACK is completely unofficial...it is supported neither by id nor by ATi...and it requires you to open and modify compiled files in your Doom3 installation...that my friend is a hack if I have ever seen one...you can church it up all you want to, but a hack is a hack...call it what it is.

So? A lot of game designers and software developers started by toying around with software in their spare time. Does that make anything that they did less worthy? I think John Carmack himself might take offense to that, or the guys at Valve Software, or many other game devleopment houses.

The files you modify for this hack are not COMPILED. They are simply a ZIP file named differently. Changing my AUTOEXEC.CFG file into a text file to modify it, then changing it back is NOT decompiling a file.

And yes, this is still a hack, and it's still not an insult like you're trying to make it out to be. Just because you're using the word to mean something that it doesn't mean, doesn't instantly change it's meaning. It's not "churching it up" when I use a term CORRECTLY.

To quote The Princess Bride ...

"You keep using that word. I do not think it means what you think it means."

 

[pxc]

my results...

laptop NV28M @ 225/470, 61.45, 840x525 MQ:
with hack 28.8fps
without hack 28.8fps

can't do much about that.

desktop 6800GT @ 400/1100, 61.77, application controlled AF (with hack/without hack):
1600x1200UQ 53.5fps/63.0fps hack is much slower
1600x1200HQ 55.6fps/65.6fps hack is much slower
1024x768HQ 82.0fps/84.6fps hack is slower
800x600MQ 86.1fps/85.0fps hack is faster
640x480LQ 86.7fps/85.0fps hack is faster

The hack is slightly faster in 800x600 and 640x480, but is slower in 1024x768 and 1600x1200 for me. I verified it twice by switching between the 2 versions and I got the same results both times. Very strange because it's a pretty significant hit (~10fps). <shrugs> I'll just have to live with 63fps 1600x1200UQ.

hack 2.0 results (MAD_SAT hack/no hack):
1600x1200UQ 54.8fps/63.0fps hack is much slower
1600x1200HQ 57.1fps/65.6fps hack is much slower
1024x768HQ 82.5fps/84.6fps hack is slower
800x600MQ 86.0fps/85.0fps hack is faster
640x480LQ 86.4fps/85.0fps hack is faster

 

[BoogerBomb]

hack 2.0 results (MAD_SAT hack/no hack):
1600x1200UQ 54.8fps/63.0fps hack is much slower
1600x1200HQ 57.1fps/65.6fps hack is much slower
1024x768HQ 82.5fps/84.6fps hack is slower
800x600MQ 86.0fps/85.0fps hack is faster
640x480LQ 86.4fps/85.0fps hack is faster

So this does nothing for Nvidia cards?

 

[burningrave101]

hack 2.0 results (MAD_SAT hack/no hack):
1600x1200UQ 54.8fps/63.0fps hack is much slower
1600x1200HQ 57.1fps/65.6fps hack is much slower
1024x768HQ 82.5fps/84.6fps hack is slower
800x600MQ 86.0fps/85.0fps hack is faster
640x480LQ 86.4fps/85.0fps hack is faster

How in the world did you test those results? You got nearly the same performance from UQ as you did with HQ which just shouldn't be!

Also, DOOM 3 has an in-game cap at 60 fps so if your getting anything higher then that its pointless.

 

[SuperRob]

So this does nothing for Nvidia cards?

And it's not supposed to.

 

[burningrave101]

And it's not supposed to.

Actually, it probably helps the NV30 cores in the same manner as the R300 and R400 cores.

 

[tranCendenZ]

Actually, it probably helps the NV30 cores in the same manner as the R300 and R400 cores.

Actually Humus' tweak causes a performance hit on the NV30 and NV40.

If ATI wants this implemented, they will have to break Carmack's goal of having everything under ARB2 and have seperate paths for NV and ATI cards - or just run a "special" ati path with Nvidia using ARB2.

 

[^eMpTy^]

So? A lot of game designers and software developers started by toying around with software in their spare time. Does that make anything that they did less worthy? I think John Carmack himself might take offense to that, or the guys at Valve Software, or many other game devleopment houses.

The files you modify for this hack are not COMPILED. They are simply a ZIP file named differently. Changing my AUTOEXEC.CFG file into a text file to modify it, then changing it back is NOT decompiling a file.

And yes, this is still a hack, and it's still not an insult like you're trying to make it out to be. Just because you're using the word to mean something that it doesn't mean, doesn't instantly change it's meaning. It's not "churching it up" when I use a term CORRECTLY.

To quote The Princess Bride ...

"You keep using that word. I do not think it means what you think it means."

let's see...blah blah blah blah blah blah...you agree that it's a hack...my work here is done...

 

[BoogerBomb]

Games will get to the point where vendor will have to be allowed to make optimizations to give their customers the best possible experience on their cards. Having all card vendors use the same code may level the playing field, but screw some people over that can have better performance.

 

[BoogerBomb]

let's see...blah blah blah blah blah blah...you agree that it's a hack...my work here is done...

Whats your definition of a hack? If you imply that is means someone is cheating or doing something underhanded then no I don't agree with that.

If you mean going in and changing something that helps you get better performance then yes.

 

[^eMpTy^]

Games will get to the point where vendor will have to be allowed to make optimizations to give their customers the best possible experience on their cards. Having all card vendors use the same code may level the playing field, but screw some people over that can have better performance.

Well I would think that having specific code paths for certain cards would be a good thing...but then that puts all the work on the developer and I don't think that's quite right...OpenGL and D3D are standards...I don't understand why everyone can't just make cards to run the standard as fast as possible...

 

[pxc]

How in the world did you test those results? You got nearly the same performance from UQ as you did with HQ which just shouldn't be!

Also, DOOM 3 has an in-game cap at 60 fps so if your getting anything higher then that its pointless.

fast system = win. It's consistent across different driver versions, too.

The timedemo is an average. Some sections are slower than 60fps, some are faster than 60fps. Without any CHEATS, also as seen in [h]'s fraps graphs, the curves for the benchmarks are similar using the same path (i.e. ARB2). What that means is that the slower sections are also faster and the framerate cap doesn't matter for those parts because it's under 60fps. And of course the faster sections don't run higher than 60fps (actually not true... it's possible to see > 60fps in the game).

I turned on com_showfps 1 for a little while this morning. It's pretty impressive to see the framerate so steady, even in complex scenes.

And it's not supposed to.

yeah, because nvidia's opengl fragment lookup isn't dog slow.

 

[^eMpTy^]

Whats your definition of a hack? If you imply that is means someone is cheating or doing something underhanded then no I don't agree with that.

If you mean going in and changing something that helps you get better performance then yes.

There's a middle ground here. I don't think the term hack is negative necessarily, except in the sense that it's not officially supported and the results of changing stuff around isn't really known (i.e. you could have artifacts or crashing or whatever). I absolutely don't think it's cheating or in any way underhanded.

 

[SuperRob]

There's a middle ground here. I don't think the term hack is negative necessarily, except in the sense that it's not officially supported and the results of changing stuff around isn't really known (i.e. you could have artifacts or crashing or whatever). I absolutely don't think it's cheating or in any way underhanded.

What does it matter if it's supported or not? No one is telling you to use it. Come on, this is HardOCP. Do you really think that there are people here not doing things to their systems that aren't supported? Does that make their machines any less fast? And even if Carmack released a patch tomorrow that enables this exact same hack on ATI hardware officially, does that make the fact that you could do it today any less viable?

The point is you're arguing a point that simply isn't valid. You say it's a hack. So do I. We clearly mean different things by that. But your intent is that this is somehow less worthy because it didn't come from from official body. Neither do any of the hundreds of hacks we all use on our machines every day. I've got hundreds of pieces of software I use that do things that weren't intended originally, and sometimes, these become official features. All this means is that I had access to these features before people who were unaware of them.

In this case, I get a performance boost, and I'm seeing no visible difference in the game. So I'll use it. Just like the changes I made to DOOM 3's autoconfig file. Are those hacks too, even though they only benefit people with better than average hardware configurations, and weren't set by default? If this hack benefited only nVidia hardware, would it be any different?

I'm sure you're trying to walk the fine line between being obviously biased and not, but you're not going a very good job of it.

 

[tranCendenZ]

In this case, I get a performance boost, and I'm seeing no visible difference in the game. So I'll use it. Just like the changes I made to DOOM 3's autoconfig file. Are those hacks too, even though they only benefit people with better than average hardware configurations, and weren't set by default? If this hack benefited only nVidia hardware, would it be any different?

As of the latest Humus tweak, there is a visual difference, whether you notice it or not is a different story. With the tweak on, the light sources are darker and there is less shine on the aluminum. Here are pics Humus captured to illustrate the differences:

tweak on (flashlight less bright/aluminum less shiny)
http://esprit.campus.luth.se/~humus/temp/shot00017.png

tweak off (flashlight more bright/aluminum more shiny)
http://esprit.campus.luth.se/~humus/temp/shot00018.png

 

[^eMpTy^]

What does it matter if it's supported or not? No one is telling you to use it. Come on, this is HardOCP. Do you really think that there are people here not doing things to their systems that aren't supported? Does that make their machines any less fast? And even if Carmack released a patch tomorrow that enables this exact same hack on ATI hardware officially, does that make the fact that you could do it today any less viable?

The point is you're arguing a point that simply isn't valid. You say it's a hack. So do I. We clearly mean different things by that. But your intent is that this is somehow less worthy because it didn't come from from official body. Neither do any of the hundreds of hacks we all use on our machines every day. I've got hundreds of pieces of software I use that do things that weren't intended originally, and sometimes, these become official features. All this means is that I had access to these features before people who were unaware of them.

In this case, I get a performance boost, and I'm seeing no visible difference in the game. So I'll use it. Just like the changes I made to DOOM 3's autoconfig file. Are those hacks too, even though they only benefit people with better than average hardware configurations, and weren't set by default? If this hack benefited only nVidia hardware, would it be any different?

I'm sure you're trying to walk the fine line between being obviously biased and not, but you're not going a very good job of it.

At this point you're pretty much talking to yourself.

Why don't you tell me precisely what the hell point you're trying to prove. I said it was a hack, you agree...so what is there to argue?

 

[s_s256]

As of the latest Humus tweak, there is a visual difference, whether you notice it or not is a different story. With the tweak on, the light sources are darker and there is less shine on the aluminum. Here are pics Humus captured to illustrate the differences:

tweak on (flashlight less bright/aluminum less shiny)
http://esprit.campus.luth.se/~humus/temp/shot00017.png

tweak off (flashlight more bright/aluminum more shiny)
http://esprit.campus.luth.se/~humus/temp/shot00018.png

This was already talked about on Beyond3D, the difference is because the flashlight had moved.

Here is a link to the modified file (updated w/ no artifacts)
http://esprit.campus.luth.se/~humus/temp/doom3PerformanceTweak.rar

 

[BoogerBomb]

As of the latest Humus tweak, there is a visual difference, whether you notice it or not is a different story. With the tweak on, the light sources are darker and there is less shine on the aluminum. Here are pics Humus captured to illustrate the differences:

tweak on (flashlight less bright/aluminum less shiny)
http://esprit.campus.luth.se/~humus/temp/shot00017.png

tweak off (flashlight more bright/aluminum more shiny)
http://esprit.campus.luth.se/~humus/temp/shot00018.png

Look at the fps increase that he got for a barely noticeable less shine if you play the game at full speed. 28 to 44 is definately in the realm of a good increase.

 

[SuperRob]

As of the latest Humus tweak, there is a visual difference, whether you notice it or not is a different story. With the tweak on, the light sources are darker and there is less shine on the aluminum. Here are pics Humus captured to illustrate the differences:

tweak on (flashlight less bright/aluminum less shiny)
http://esprit.campus.luth.se/~humus/temp/shot00017.png

tweak off (flashlight more bright/aluminum more shiny)
http://esprit.campus.luth.se/~humus/temp/shot00018.png

You're right, there is a visual difference in those screenshots. Whether that difference is due to the slight movement of the flashlight between the two shots, or becuase of the rendering tweak. I really don't know. But when I'm playing a game, I'm not comparing the difference side-by-side like that. If the results after the tweak are that it's less than 1% darker in an already dark game, and I can simply offset that by boosting the brightness a notch, I'm satisfied.

Edit: Humus just updated, saying that the light difference was indeed because the flashlight had moved between the screenshots, and that the areas unchanged by that movement had zero difference.

 

[rancor]

well all that really matters is is ID going to endorse this in the next patch?

 

[tranCendenZ]

This was already talked about on Beyond3D, the difference is because the flashlight had moved.

Nah, the entire cone of the flashlight is dimmer, not just the objects it relatively shines on. (though as a result of the light source being dimmer it looks like the shine is less intense as well).

 

[SuperRob]

well all that really matters is is ID going to endorse this in the next patch?

What does it matter if they endorse it or not if we can do it ourselves? Since when did we ever wait for someone else to do something that we could do for ourselves? Isn't that how software like Powerstrip came about? Because someone else wouldn't (or couldn't) do something for us?

 

[Mchart]

If you want to ruin your IQ for 4 more frames, be my guest..

 

[^eMpTy^]

What does it matter if they endorse it or not if we can do it ourselves?

lol...I guess it doesn't matter at all...you're satisfied...who gives a shit about the rest of the world right?

 

[rancor]

What does it matter if they endorse it or not if we can do it ourselves?

It kinda matters since most of the players of Doom 3 won't do it like us

 

[avatar_of_might]

ah how ironic we can be?
i remember this just a few weeks ago except for a different game: Far Cry
both sides busting their asses to take the top performance spot.history repeats itself over again. ah guess this will get just get as intense when hl2, bfme, stalker, fear, etc. come out.
*sigh* well im content running d3 and farcry @ 1280x1024 with my aiw 9800pro. if my 128mb 9800pro can run it at 1280x1024 im waiting to see what the 512mb cards can do then i might upgrade.

 

[BoogerBomb]

well all that really matters is is ID going to indorse this in the next patch?

Chances are no since they would have to create an ATI path and Carmack says he doesn't want to do that. If he did then he will have all Nvidia owners yelling "Cheat!!" in his left ear which will give his right ear a break from the ATI owners yelling "Nvidia sellout!" in it.

Carmack will lose no matter what he decides to do to D3.

 

[rancor]

Chances are no since they would have to create an ATI path and Carmack says he doesn't want to do that. If he did then he will have all Nvidia owners yelling "Cheat!!" in his left ear which will give his right ear a break from the ATI owners yelling "Nvidia sellout!" in it.

Carmack will lose no matter what he decides to do to D3.

Very true

 

[burningrave101]

well all that really matters is is ID going to endorse this in the next patch?

I can pretty much guarantee you that that will be a big fat NO.

I dont think JC even likes ATI after what they did concerning the DOOM 3 Alpha lol.

 

[SuperRob]

FYI, Reverend at Beyond3D says that he he just got "the longest email I have ever received from JC" about Humus's tweaks.

http://www.beyond3d.com/forum/viewt...tart=280&sid=9e276e85abda29b0b9c32c5244ffe85b

Looks like we're going to have some "official" clarification here soon, for those of you hung up on that sort of thing.

 

[BoogerBomb]

I can pretty much guarantee you that that will be a big fat NO.

I dont think JC even likes ATI after what they did concerning the DOOM 3 Alpha lol.

I doubt Carmack would screw over half his customers over something an ATI employee did.

 

[burningrave101]

I doubt Carmack would screw over half his customers over something an ATI employee did.

I would lol.

And i doubt over half his customers are ATI users. nVidia has a slightly higher market share then ATI, although they are about tied right now i believe. DOOM 3 is also an nVidia branded and endorsed game so i can safely say more nVidia users are running it then ATI users. Especially after the DOOM 3 benchmarks.

PLUS, nVidia and ATI are just a small portion of the GPU market. They both have around like 23% or so market share. Intel owns the majority of the graphics market and VIA has a small chunk as well.

Will DOOM 3 even run on an Intel integrated graphics setup? lol

He's not exactly screwing anyone over by not adding it in ither.

Its just like the SM 3.0/2.0b path in Far Cry. They give extra performance but they aren't really needed.

And it looks like this HACK in D3 for the R300/R400/NV30 cores lowers the IQ so that will be a major reason not to implement it.

 

[^eMpTy^]

I doubt Carmack would screw over half his customers over something an ATI employee did.

I have to agree...I'm sure JC cares more about the people playing the game than the graphics companies antics...I'm super curious to see what JC says...it's his game, if he says it's cool, then that's the end of that...

 

[^eMpTy^]

I would lol.

He's not exactly screwing anyone over by not adding it in ither.

Its just like the SM 3.0/2.0b path in Far Cry. They give extra performance but they aren't really needed.

And it looks like this HACK in D3 for the R300/R400/NV30 cores lowers the IQ so that will be a major reason not to implement it.

well the longer this goes on the less the iq issues seem to be showing up...so maybe if JC decides to put it in a patch all the iq issues can be solved and we'll actually have something worth talking about...

The only reason I think there might be something to this is that, like Humus said, he doesn't really understand why ATi cards would have this issue...and that being as it is, maybe they didn't see the problem because it doesn't make sense for it to be there...keeps getting curiouser and curiouser...

 

[BoogerBomb]

Due to the oh so very slight IQ degradation it gives in screenshots he probably doesn't like it. But can he really blame us for wanting to get some more performance out of the game when the degreadation is virutally unoticable during gameplay?

I can assure you that this will not be the last tweak/hack that will come out for D3.

 

[s_s256]

After looking at the two screen shots for 2 minutes, I did see some IQ problems with the flashlight in the tweaked version. The flashlight was slightly dimmer, I would never notice it in game, but I guess Humas has a little more tweaking to do.

 

[’m‚³‚ñ]

This is what happens when two evil geniuses come together to work on a problem, but have completely conflicted view points. *Goes to unsubscribe from this thread before an email server dies from too many emails.*

 

[BoogerBomb]

This is what happens when two evil geniuses come together to work on a problem, but have completely conflicted view points. *Goes to unsubscribe from this thread before an email server dies from too many emails.*

If they could handle the D3 benchmark thread then this a walk in the park.