[D.D.]

heheh

Hi John,

No doubt you heard about GeForce FX fiasco in Half-Life 2. In your opinion, are these results representative for future DX9 games (including Doom III) or is it just a special case of HL2 code preferring ATI features, as NVIDIA suggests?

Unfortunately, it will probably be representative of most DX9 games. Doom has a custom back end that uses the lower precisions on the GF-FX, but when you run it with standard fragment programs just like ATI, it is a lot slower. The precision doesn't really matter to Doom, but that won't be a reasonable option in future games designed around DX9 level hardware as a minimum spec.

source: http://english.bonusweb.cz/interviews/carmackgfx.html

 

[dogulation000]

heheh



source: http://english.bonusweb.cz/interviews/carmackgfx.html

Interesting, even Carmack is saying that Half-Life 2 is going to show how well each line of cards is going to perform on most of DX9 games. Unless Nvidia makes a new line, of course.

 

[D.D.]

I would also like to point this out:

"Doom has a custom back end that uses the lower precisions on the GF-FX"

so....which card is using lower precision?

 

[Guest]

Doom 3 regulates all all shaders through an 8 bit backbuffer. You wont see the difference on the two cards. 24 bit precision in this game is pretty much pointless due to the the fact that the game has been designed around FX 16 integer

 

[op3n_skill]

err... isnt doom3 OGL?

 

[Guest]

err... isnt doom3 OGL?

Yes but the shader functions are similar. Like I said. Doom regulates all shaders through an 8bit backbuffer. And was designed around FX 16 integer precision and FX12 (Pixel Shader 1.4 and 1.3) 24 FP, 16 bit FP. and 32 bit FP (which are a part of DirectX 9.0) as mentioned here. Are whats slow on the Nv30 architecture.


That Being said. I still Think the Doom 3 engine is gonna look better than Half life 2

 

[D.D.]

hmmm.....so the precisions don't matter in doom III cuz it was designed on the capabilities of ps 1.3 and 1.4? Thats good news for my gf4...looks like it still has some life in it yet, unlike for HL2....

BTW, if the "special codepath" for the nv2x/3x in doom III uses lower precision, would it be fair to assume that other nv3x codepaths for dx9 games would use lower precisions to "optimize" the experience?

PS: I just played a little UT2K3 today and I've got to say that it still has pretty amazing graphics compared to HL2 and doom III...

 

[dogulation000]

Both Gabe Newell and John Carmack said that Half-Life 2 is going to represent how most later Directx 9 games would perform on both cards to come. That's unless, ATI decides to go 32 bit precision, and Nvidia fixing their cards with a new line. Now remember, I'm pretty sure that ATI and Nvidia don't sit idly while they get criticized by high profile people from developers.

 

[fivefeet8]

Look

Driver optimizations? Cheats? Who cares. Large performance increase, very little IQ degradation.

 

[Raiden01]

WELL personaly i dont notice any differences bettween those shots exspect the frame rate which is 3 x more on some screen shots which is amazing and hopefully half life 2 will perform alot better on these drivers

 

[Alias]

Look

Driver optimizations? Cheats? Who cares. Large performance increase, very little IQ degradation.

So? The screenshots are from UT2k3, a directx 8.1 game (and a very minimal one at that). It's already known that GF series (esp the FX) excels at directx 8, as a large of the die spaces was devoted to integer registers. Additionally, FX isn't the strongest in AA/AF either.

Directx 9 games are where GF FX stalls, because its floating point performance sucks, which is where nvidia resorts to IQ degredation to speed things up. Don't believe me? If the half-life scores don't convince you, look at any article based on aquamark 3.

 

[fivefeet8]

So? The screenshots are from UT2k3, a directx 8.1 game (and a very minimal one at that). It's already known that GF series (esp the FX) excels at directx 8, as a large of the die spaces was devoted to integer registers. Additionally, FX isn't the strongest in AA/AF either.

Directx 9 games are where GF FX stalls, because its floating point performance sucks, which is where nvidia resorts to IQ degredation to speed things up. Don't believe me? If the half-life scores don't convince you, look at any article based on aquamark 3.

It's a directx8 game that has just seen a performance increase of 2-3x just from a driver update. It's still one of the best GAMES out right NOW which a lot of people play. And from what I've read, those two DM maps uses quite a few dx8 pixel shaders for effects. An increase of that magnitude is surely welcomed by any game player. The FX AA/AF performance may not be the strongest, but it's not such a huge difference of performance from the Rad's AA/AF.

You're arguing a point that has already been proven(I've seen/read all the articles) and I wasn't even trying to argue with it. Where did I even mention it would increase Dx9 games? :???:

 

[Alias]

Sorry fivefeet8, I was just saying that I don't think nvidia has to cheat in dx 8 games because their cards do so well in that particular api anyway (nvidia had a big hand in developing dx 8, but not as much with dx9, so they were vary familiar with dx 8). DX 9 games is where nvidia usually uses optimizations in.

For today's games, GF FX is the fastest. Tomorrow's games are a different story. And yes, I know newer cards will be out by then from both companies.

 

[Guest]

Nvidias AA performance is fine IMO. Tho It's AA quality (multisampling) COuld use some touchups. 2x is fine. 4x needs to be rotated.

 

[saulin]

Yep Nvidias AF is amazing...


BTW fivefeet8 those shots look nice. I guess I'll install the game today and take some shots. I'm still using the StarStorm Detonators. They give me superb quality.

AF looks the best with this drivers.

 

[saulin]

Ok here are some shots of UT 2003. Actually after running some benchmarks on the game I just lost like 1-2 fps with these drivers compared to the original detonators. On 3DMark 2001 I lost like 900 3DMarks yikes.


Using StarStorm Detonators 45.23q
1024x768 32-bit
4x FSAA and 8X AF

Too bad the Jpeg compression makes the image a bit blurrier . But the AF can be noticed to be better than using the regular Detonators.

 

[D.D.]

hmm...might just give those star storms a try...

 

[saulin]

yeah the StarStorm Detonators were not such a performance hit for UT 2003 or emulators for me. But you can notice the image quality right away even on your desktop.


I used the dm-inferno map since it is the one that gives me the worst scores. But using 4X FSAA and 8X AF is still very playable for me.

Here are my scores at 1600x1200 no FSAA and no AF with the regular Detonators.

http://members.shaw.ca/navasa/16x12.html

...and here the new scores with the same settings using the StarStorm Detonators

http://members.shaw.ca/navasa/results.html

 

[fivefeet8]

Hey Saulin, post up scores from the DM-Icetomb and DM-Rustatorium maps. I'd like to see how those drivers do with those two maps.

 

[saulin]

I see those maps are in the Unreal Tournament 2003 Epic Bonus Pack. I'll download it and then benchmark those 2 maps.