Check out this thread

[Avalon]

Someone please post a link to the patch/tweak you specifically used that gave you gains. I want to test something.

 

[MercenaryForHire]

Originally posted by: Avalon
Someone please post a link to the patch/tweak you specifically used that gave you gains. I want to test something.

Same one Hardcore linked to.

- M4H

 

[MercenaryForHire]

Now, on to my "quick" testing.

System: Athlon XP 2600+, nF2 Ultra 400, 512 DCDDR @333MHz, R9500Pro @ stock, Cat4.9s

All stated scores are averages of three cached runs, and are Average / Minimum as observed with FRAPS. Unless otherwise noted, tests were run in "High Quality" mode, with AF disabled. When AF and AA were used, they were forced via RadLinker.

640x480
Before: 47.3, 23
After: 48.0, 28
Significant minimum FPS increase

800x600
B: 35.8, 20
A: 38.3, 24
Gains in both avg and min

800x600, 8xAF
B: 33.8, 15
A: 38.7, 21
Note that avg FPS is FASTER with AF in this resolution!

1024x768
B: 29.9, 13
A: 30.3, 14
Negligible increases?

1024x768, 8xAF
B: 23.1, 9
A: 28.2, 14
10x7 with AF is now "playable."

1024x768, 2xAA, 8xAF
B: 18.8, 7
A: 21.4, 9
Impressive gains, however using AA results in the "Double Vision" effect blanking the screen totally.

- M4H

 

[Avalon]

I just wanted to make sure I wasn't using some old one. Thanks. I'm going to try some quirky settings and see if I can manage to find a framerate difference this time. Thanks M4H. I'll let you know.

 

[MercenaryForHire]

Originally posted by: Avalon
I just wanted to make sure I wasn't using some old one. Thanks. I'm going to try some quirky settings and see if I can manage to find a framerate difference this time. Thanks M4H. I'll let you know.

When you get back, check my scores posted above. It seems to really tweak AA and AF. I get "free" 8x at 8x6, and a much smaller hit at 10x7. The Cat4.9s seem to have broken the AA though.

- M4H

 

[jiffylube1024]

Originally posted by: Genx87

You might want to refine your sense of irony. When nV substituted shaders in 3DM03, they did so without telling anyone until they were caught, they used their whole company to discredit 3DM03, and they never once apologized for cheating in a benchmark.

Not quite. I am sure if we go back a year or so we will see the same people who praise this were bashing Nvidia for doing the same thing. So I ask what is the difference? Both are hacks, both do not give the same exact pic quality.

^Telltale sign of a fanboy, hyperbole to the point of lying.

------------------------------------------
[/quote]

Why are you pandering to the lowest common denominator in this discussion? nV never "gave out" driver tweaks, it basically slipped them into its drivers without telling anyone. nV certainly was cheating, though the devil comment is hyperbolic fanboy talk.

Hyperbolic fanboy talk also results in people claiming "ID is working against ATI." Why do you bother listening to people ignorant or biased enough to make such statements? Do you enjoy bringing this forum down to the level of soccer hooligans? Who or what exactly is the "it" that is proclaiming that "ID is working against ATI," BTW? Surely you can't mean ATI itself, as it hasn't commented on this--but then, who?

Did you even bother to read the original posters post?[/quote]

Do you ever bother to acknowledge when someone makes a strong point that what you are doing is helping to fuel the flames?

This frothy mouthed trollship of the first order has to stop, and insecure nVidiots need to give this schtick up!

------------------------------------------

Now that that is out of the way, on to the real discussion (ie, not flame-fest!).



Alright, surely you guys have seen the benchmarks! Nvidia crushes ATI in Doom3! The 6800 series makes mince meat out of everything, often by a twofold margin in lower resolutions, and by 35%++ gaps at high resolutions. The card of the day is the 6800GT, and nothing is going to change that. Heck, I'm dying to get a 6800GT, but the $400 price tag is just too much for me to be able to justify to myself!

With that said, use some objectivity and reason here. Yes, Doom3 was custom made for Nvidia hardware - there are some good articles on this on the web about how ATI's "fast Z clear" is not a good match for D3, for example, as well as other technologies that Nvidia runs better than ATI or ATI worse than Nvidia for Doom3. Nevertheless, is this not an (at least slightly) uncharacteristic throttling of ATI by Nvidia, particularly when comparing the 16-pipe X800XT with the two 16-pipe 6800GT and 6800 Ultra cards? They kick the X800XT's arse, and make the Pro look like last gen tech, especially at lower resolutions (1280 and below).

If you guys took the time to even liberally parse through the Beyond3D article, instead of jumping into this flame-fest feet first, you'd see the very essence of this optimization is based on a very simple idea - how Nvidia is much faster at texture lookups (what is done by default, obviously), while ATI can sometimes just do a calculation mathematically faster.

And if you read even a little deeper into the B3D article, you'd read that by 'doing the math' so to speak, and using this "shader replacement," you get similar, if not more precise results than the stock way of doing it (hint: read past page 5, when Humus' results were producing inferior results). Of course, it actually hurts performance on Nvidia cards because, as said above, Nvidia does texture lookups faster than working it out; Nvidia essentially did custom tailor their card to be a Doom3 beast out of the box, and tweaks like this generally make the Nvidia cards perform worse.

The counterpoint by John Carmack to these types of optimizations is that he wanted to standardize shaders across the board, instead of introducing tons of custom shaders and paths for everything, which in turn greatly simplifies troubleshooting across hardware configurations.

Both sides make excellent points - I agree 100% with JC's decision to standardize shaders for the game, and I agree 100% with Humus' decision to play around with the shaders and, being an ATI developer as well as a hardware enthusiast, using his ATI hardware know-how to tweak a setting in D3 to work better on the hardware he knows best.

If we can take this for what it is - a clever hack by a knowledgeable hobbyist who works for ATI, which can give ATI users a tangible, if (IMO) a bit of an underwhelming speed boost, then what is wrong with that?

Instead, the mere thought of ATI getting better performance by an unofficial hack triggers a tidal wave of name calling and mud throwing, accusations and biases. Never mind the fact that the ATI cards still gets crushed by the 6800 series!


-------------------------------

This isn't the optimizations of yore by Nvidia to artificially level the playing field with inferior IQ and to make performance +/- 10% equal; this is a shot in the arm to ATI's rendering speed to just make them less crippled, and it isn't being force-fed to anyone through either company's drivers, but through a hardcore 3d discussion forum!

Rant over - sometimes the internet can be such a drag!

 

[nitromullet]

Exact same settings and improvement as me... except I'm at 800x600 because of my 9600Pro.

All kinds of people (mainly nvidia users) are saying that we are sacrificing image quality for FPS, and it may be true, but if it's image quality that we can't even tell is gone, then there's no drawback!

I for one, as a nVidia user, think this is actually pretty cool that ATi guys can tweak a few things and enjoy Doom3 as much as us nVidia guys. What I didn't like was the finger pointing at JC and id with no real proof.

This tweak seems to yield some pretty good results for some, none for others, and artifacts for yet anouther bunch. I haven't dug into this in detail as it doesn't really apply to me. If you can improve the game's performance with no noticable IQ loss, I would do it too. :beer:

Edit: and an extra one for you jiffylube1024 :beer:

 

[Avalon]

My AF levels (second column) are what is determined by the detail setting I use, unless I specifically note that I am forcing that AF setting in my control panel. My AF, when forced in the benches, will be set to Quality. It is my general understanding that High detail uses 8x AF. If I am wrong, correction would be much appreciated, but until then, I will assume as much in my benches. I'll be runnning timedemo demo1 about 2-3 times, enough for it to be a smooth run and my frames to be consistent. Before the patch framerate and after the patch framerate scores will be included for comparison.

My rig is an AXP Mobile at ~2.4ghz @ 216x11. I'm running a gig of Geil golden dragon in dual channel at 2-3-3-6 PC3500. My video card is a Connect3D Radeon 9500 128mb non pro, softmodded to a 9700, overclocked to pro levels of 325/310mhz. My motherboard is an Epox 8RDA3i nforce2 400 Ultra. For video drivers, I'm using the latest 4.7 Omegas. All other drivers are fully up to date.

Now for the scores:

10x7 Low detail 0x/0x
Before: 41.8
After: 41.7

10x7 Low detail 0x/8x Forced via CP
Before: 32.6
After: 40.5

10x7 Med detail 0x/0x
Before: 42
After: 41.6

10x7 Med detail 0x/8x Forced via CP
Before: 31.6
After: 39.3

10x7 High detail 0x/8x
Before: 37.3
After: 38.6

10x7 High detail 0x/8x Forced via CP
Before: 30.2
After: 38

10x7 High detail 0x/16x Forced via CP
Before: 28.4
After: 37.7

I took screenshots of when I was running at 10x7 High detail at 16x Forced in my control panel. I took them staring at the same spot in my game before the tweak, after the tweak, and with 0x AF before the tweak. I honestly could not spot any difference between 0x AF and 16x AF while flipping between screenies, much less any difference between 16x before and after. PM me and I'll send you my set of screenies over AIM or such, if you'd like. I need to re-read this data for a bit until I can speculate something.

Seems to me that the tweak mainly only helps out when you use AF. Either using the control panel AF is inefficient compared to using High quality to gain 8x AF, or High quality doesn't give AF at all, which I had thought it had. I'm still pretty sure it does. Perhaps, then, control panel implementation of AF in Doom 3 actually is inefficient. Also, note my "After" benches on High detail. Note the 8x game, 8x CP forced, and and 16x CP forced. It's pretty close to identical, save an FPS between the highest and lowest score, which could easily be well within the margin of error for my benches. What I'm really thinking, is that this tweak is not implementing AF correctly, and in turn is like running the game without AF at all. We're really going to need some people with keen eyesight to scrutinize over some good screenshots to determine if this is what's really going on.

 

[Terranboy]

I have a AthlonXP 2600+, and a Radeon9800Pro. Here are some results and questions.

800x600, (Medium settings), Catalyst 4.7, No tweak
32.7fps

800x600, (medium settings), Catalyst 4.7, Shader tweak (new interact file)
33.3fps

800x600, (medium settings), Catalyst 4.9 Shader tweak.
39.9fps
33.4fps (high settings)*

800x600 (medium settings) Catalyst 4.9 No tweak.
35.6fps (confusing)

After I applied the new shader file, there was a tiny increase in performance. Things looked ok in the timedemo, but the game was very dark. I then replaced the pak000 file with the original one I backed up before starting all that. It asked if I wanted to replace the Interact file ( they were different sizes so I know the changes took effect ). But my levels are still dark. I can see lights, but there is NO ambience. I have no idea why, since I replaced the shader mod with the original file. Sorry, I can't sit and read 28 pages of thread to see if this has been covered.

Also, the shader tweak and the patch seem to work it tandom. But I can't fix the darkness problem.

 

[gururu]

this is really beginning to hurt...

1024, 16X AF (quality) , 4.7

medium quality
before:29.9
after: 37.0 (+24%)

high quality
before: 28.5
after: 36.2 (+27%)

jiffylube1024 wrote:

If we can take this for what it is - a clever hack by a knowledgeable hobbyist who works for ATI, which can give ATI users a tangible, if (IMO) a bit of an underwhelming speed boost, then what is wrong with that?

This is hardly an underwhelming speed boost....IMO

 

[GeneralGrievous]

Between the cats 4.9 and this tweak (if it works), the XTPE is slowly catching the Ultra. According to hardocp it was 25-30% behind, that doesn't seem to be the case anymore.

 

[Avalon]

Originally posted by: GeneralGrievous
Between the cats 4.9 and this tweak (if it works), the XTPE is slowly catching the Ultra. According to hardocp it was 25-30% behind, that doesn't seem to be the case anymore.

You make it sound like ATI is about to pass up Nvidia in Doom 3. Not going to happen. More analysis of the tweak is needed. Right now, I don't even think it works.

 

[clicknext]

Originally posted by: gururu
jiffylube1024 wrote:

If we can take this for what it is - a clever hack by a knowledgeable hobbyist who works for ATI, which can give ATI users a tangible, if (IMO) a bit of an underwhelming speed boost, then what is wrong with that?

This is hardly an underwhelming speed boost....IMO

It's a strange speed boost. For some people there's almost no difference, but for some like me, there's the full 40% advertised, with the right settings.

 

[gururu]

Originally posted by: Avalon
You make it sound like ATI is about to pass up Nvidia in Doom 3. Not going to happen. More analysis of the tweak is needed. Right now, I don't even think it works.

I'm going to finish the game quickly with the new speeds before they find something wrong with the tweak!

 

[clicknext]

Originally posted by: gururu

Originally posted by: Avalon
You make it sound like ATI is about to pass up Nvidia in Doom 3. Not going to happen. More analysis of the tweak is needed. Right now, I don't even think it works.

I'm going to finish the game quickly with the new speeds before they find something wrong with the tweak!

Haha, yeah... then you'll be forced to conform and play the game at a slow rate!

 

[Avalon]

Originally posted by: clicknext

Originally posted by: gururu

Originally posted by: Avalon
You make it sound like ATI is about to pass up Nvidia in Doom 3. Not going to happen. More analysis of the tweak is needed. Right now, I don't even think it works.

I'm going to finish the game quickly with the new speeds before they find something wrong with the tweak!

Haha, yeah... then you'll be forced to conform and play the game at a slow rate!

Well, the tweak "works" in such that it gives you free AF. Whether it is actually doing AF or not is another story

 

[Shagga]

Originally posted by: Avalon

Originally posted by: clicknext

Originally posted by: gururu

Originally posted by: Avalon
You make it sound like ATI is about to pass up Nvidia in Doom 3. Not going to happen. More analysis of the tweak is needed. Right now, I don't even think it works.

I'm going to finish the game quickly with the new speeds before they find something wrong with the tweak!

Haha, yeah... then you'll be forced to conform and play the game at a slow rate!

Well, the tweak "works" in such that it gives you free AF. Whether it is actually doing AF or not is another story

I agree. Thsi is what the reults seem to suggest to me also.

 

[jiffylube1024]

Originally posted by: gururu
this is really beginning to hurt...

1024, 16X AF (quality) , 4.7

medium quality
before:29.9
after: 37.0 (+24%)

high quality
before: 28.5
after: 36.2 (+27%)

jiffylube1024 wrote:

If we can take this for what it is - a clever hack by a knowledgeable hobbyist who works for ATI, which can give ATI users a tangible, if (IMO) a bit of an underwhelming speed boost, then what is wrong with that?

This is hardly an underwhelming speed boost....IMO

My reasoning for calling it underwhelming is that most of the settings that it gives large speed boosts to were either a.) totally unplayable before or provide b.) still too-low framerates for my tastes (I'm starting to sound like BFG10K here ).

Basically, I've been accustomed to 60+ fps speeds on my 9800 Pro; now this game comes out and destroys our ATI cards, yielding 30-45fps average but having some serious dips in gameplay. All the while, D3 is chugging along at >100 fps in benchies on 6800GT+ cards at similar resolutions (ie 1024).

If it improved my smoothness at 800X600 more than just marginally, or it made 1024 totally smooth (which, for me at least, IMO it did not do) than I'd be happier. But after applying an update promising "up to 40% improvement" and getting a ~5% improvement at the settings I play at (800X600 HQ), that is what caused me to call it underwhelming.

This is all my personal opinion, of course, and if you like to game at 1024 with 16X AF, then I can see how this would impress you a lot more - to each his own!
---------------------------

I value a high, consistent framerate over eye candy (let's say I'm getting 40fps avg in Doom3 - personally, I'd gladly take 5-10 extra fps (and the resulting improvement in minimum fps) at high quality over free 16X AF). That's just me...

I think if I had a 256MB card I'd be happier - judging by HardOCP's original Doom3 tweaking guide, the 256MB Radeons seemed to have much better tolerance to huge dips in framerate than the 128MB cards. Still, I though dropping down to 800X600 would help me more in this respect, but I still get those big dips in large firefights.

 

[Jeff7181]

I noticed the shader file I have was different than described in the written part that explains how to do it in the OP on those forums... so I followed it as best I could, ignoring the stuff that was different, and here's my results...

Before timedemo1 = 47.1 FPS
After timedemo1 = 51.3 FPS

So then I try the "premade" one that he linked to to download, and got 47.6... so... for those of you not seeing much improvement... here's the shader I'm using...

interaction.vfp

*EDIT* By the way, this is on my 9800 Pro (R350 core) at 380/370 with CAT 4.7's.

And that is with AA and AF off... so don't try to tell me the tweak forces it not to do any AF

**EDIT** After playing the game... I'm about 75% sure specular lighting isn't working after performing this "tweak." Which makes sense considering he's editing out a line that has something to do with specular...

***EDIT*** Ok maybe not... but it definately does something that effects specular lighting... I just ran it without the tweak with specular on and with it off and it makes no difference in FPS. The image quality is definately noticeable without specular on, and the "tweaked" image does look more like the non-tweaked with specular on than it does with specular off, but it's definately different... I'm going to take some screens and see if it can be seen in a still picture.

 

[gururu]

Originally posted by: Avalon

Well, the tweak "works" in such that it gives you free AF. Whether it is actually doing AF or not is another story

actually, in my case, the results with AF(tweak) are higher than no AF (no tweak).

 

[Jeff7181]

Ok... I noticed I forgot to ad that POW line... to make it use math to do the specular lighting or something. So... I added that in, and well, it looks like crap... I'm uploading some screenshots right now. Looking at these, it's very clear why Carmack didn't do this from the beginning... I'm going to play with it more though and see if I can change anything else and make it look better.

Pre-tweak, specular off
Pre-tweak, specular on
Post-tweak, no "POW line" added
Post-tweak, "POW line" added

Here's how I see it... without the POW line added, the tweak looks just a little darker than the original with specular. With the POW line added, the tweak looks almost like specular is completely gone, but it also reduces some nasty looking blockyness on "shiny" surfaces... I wonder if it's using 16-bit precision... I wonder if I can make it use 24-bit...

 

[Keysplayr]

Jeff, it's really difficult to see your screenies because they are small and windowed. If you need some screenies hosted in full size, let me know.

Keys

 

[Jeff7181]

Originally posted by: keysplayr2003
Jeff, it's really difficult to see your screenies because they are small and windowed. If you need some screenies hosted in full size, let me know.

Keys

check out the ones I just posted links to... those are full size jpegs with maximum quality compression in photoshop

 

[gururu]

the post tweak no POW does look darker, but not any worse than the no tweak with specular, particularly when gamma is bumped up just a notch..
the 'with POW' looks like crap. So I don't understand why you are tweaking the paths. Why aren't you using the 'fixed patch' vfp file?

 

[gururu]

sorry