so you used different tools to capture each API there? i guess if you use hypersnap for both, it should even the playing field. perhaps there is also an option to turn off that post-processing?

analog_programmer wrote on 2023-08-25, 16:01:

Joseph_Joestar wrote on 2023-08-25, 07:11:

Meatball wrote on 2023-08-25, 04:20:

Here is a comparison of Turok Direct3D and Glide:

Thanks for the detailed tests!

I've removed Turok from the preliminary list, and linked to your post as reference.

It is a subjective judgement. Yes, the "fog" with Glide looks like "vertical wall" on one of the pictures, compared to D3D's "fog", but in all other aspects (for me) Glide pictures seems to be better and brighter.

Maybe, besides "looks best" with Glide, you should leave an rating "equal" with D3D.

I preface by stating I'm not saying I'm right and you're wrong. I agree on the subjectivity, of course.

You can manually tone down gamma for Glide and turn up for Direct3D. I had to compare like-for-like at default settings, though, and I ignored this subjectivity in my final analysis. Yes, the gamma is turned up to 11 on Glide and down to -11 on Direct3D and both can be dialed back (in my opinion), but these are the default settings. However, what tilts the game in favor of Direct3D (to me) is the fog missing on Glide (check the fog photo with the spinning Tek Arrows), the softer fog (check the cave) at ground level and the sharper textures. Some good Anisotropic filtering would/could help to alleviate texture image quality, but it isn't an option for the original Turok with Glide.

Meatball wrote on 2023-08-25, 21:46:

analog_programmer wrote on 2023-08-25, 16:01:

Joseph_Joestar wrote on 2023-08-25, 07:11:

Thanks for the detailed tests!

I've removed Turok from the preliminary list, and linked to your post as reference.

It is a subjective judgement. Yes, the "fog" with Glide looks like "vertical wall" on one of the pictures, compared to D3D's "fog", but in all other aspects (for me) Glide pictures seems to be better and brighter.

Maybe, besides "looks best" with Glide, you should leave an rating "equal" with D3D.

I preface by stating I'm not saying I'm right and you're wrong. I agree on the subjectivity, of course.

You can manually tone down gamma for Glide and turn up for Direct3D. I had to compare like-for-like at default settings, though, and I ignored this subjectivity in my final analysis. Yes, the gamma is turned up to 11 on Glide and down to -11 on Direct3D and both can be dialed back (in my opinion), but these are the default settings. However, what tilts the game in favor of Direct3D (to me) is the fog missing on Glide (check the fog photo with the spinning Tek Arrows), the softer fog (check the cave) at ground level and the sharper textures. Some good Anisotropic filtering would/could help to alleviate texture image quality, but it isn't an option for the original Turok with Glide.

But of course! We also can turn down gama correction on Glide to -11 and use most optimized for speed settings to make the image even uglier 😁 As I already mentioned, it's subjective, but I think the topic is mostly for best possible settings for image quality with Glide vs with D3D.

And also I do not think that if picture with PowerVR is better than with Glide and with Glide is better than with D3D in some games, they must be discarded from the list - they're still better looking with Glide. But again... it's subjective.

dr_st wrote on 2023-08-25, 20:31:

TBH, other than the different lighting/gamma, and one place at the start of The Sanctuary of Stone and Fire, where the D3D version has some artifact not present in the 3dfx image, I cannot say there are meaningful differences.

Thanks for the detailed tests!

Looks like Rayman 2 can be put into the newly established "equal" category as well.

auron wrote on 2023-08-25, 21:13:

testing both D3D and glide on the V3 is a good call in my opinion, save for games that prohibit running D3D on it of course, like diablo 2 IIRC. the UT99 redeemer example shows that a much newer nvidia card can introduce other variables into the mix.

During my testing, I've noticed that the readme files of many games which support Glide specifically warn you against using Direct3D or OpenGL on 3DFX cards. It's possible that the developers didn't test them thoroughly under such conditions. That's the reason why I perform Direct3D and OpenGL testing on an Nvidia card.

And while I usually use my GeForce FX with 45.23 drivers, I will sometimes test on a TNT2 M64 with period-correct drivers from late 1999 or early 2000. I don't always do this, but if I see significant differences in rendering like missing graphical elements and such, I will check the game on that hardware just to take newer drivers out of the equation. So far, Unreal Tournament has been the only title where the differences could be somewhat mitigated by using an older card/driver.

auron wrote on 2023-08-25, 21:24:

so you used different tools to capture each API there? i guess if you use hypersnap for both, it should even the playing field. perhaps there is also an option to turn off that post-processing?

For my part, I had HyperSnap's post processing turned off ever since leileilol pointed out that it's inaccurate. I also did try using HyperSnap to capture both Direct3D and Glide a few times for comparison purposes, but the gamma differences between the two APIs remained, even when the in-game gamma sliders were completely equalized. I suspect Glide might handle gamma a bit differently than Direct3D.

Joseph_Joestar wrote on 2023-08-26, 01:33:

Looks like Rayman 2 can be put into the newly established "equal" category as well.

I suppose so. With that said - other than the looks, my experience with Rayman 2 has always included some instability and visual glitches when running in DirectX mode. Like - sometimes the text doesn't show up, or flashes... Glide has always been perfect (as long as you have a machine that supports it, of course).

This is consistent with your findings:

Joseph_Joestar wrote on 2023-08-26, 01:41:

I've noticed that the readme files of many games which support Glide specifically warn you against using Direct3D or OpenGL on 3DFX cards.

But I recall having the above problems with Rayman 2 in DirectX mode on every other card as well. I think even GOG ships it with a Glide wrapper (not certain).

Another game I have on that old system where 3dfx looks cool is Need for Speed II Special Edition. But I suppose it doesn't qualify for this thread, since the alternative is SW rendering only?
Still, here is a side-by-side comparison if you wish (not done by me):
https://www.youtube.com/watch?v=fUjmJGofeMI

dr_st wrote on 2023-08-26, 07:40:

But I recall having the above problems with Rayman 2 in DirectX mode on every other card as well. I think even GOG ships it with a Glide wrapper (not certain).

I haven't played Rayman 2 a whole lot, just that introductory bit until you reach the hub area with the level select where I took the screenshots. But I don't recall any particular issues with Direct3D on my GeForce FX card.

dr_st wrote on 2023-08-26, 07:40:

Another game I have on that old system where 3dfx looks cool is Need for Speed II Special Edition. But I suppose it doesn't qualify for this thread, since the alternative is SW rendering only?

Yeah, NFS2:SE doesn't fit here due to the lack of other renderers which use 3D acceleration. But I do have it in this thread: List of Glide only games

Games tested:

• Requiem: Avenging Angel - retail CD release + latest official patch 1.3

Graphics cards tested:

• 3DFX Voodoo3 2000, using 3DFX reference drivers version 1.07.00 (Glide renderer)
• Nvidia TNT2 M64 using Nvidia reference drivers version 3.68 (Direct3D renderer)

Decided to give this game a quick test as well. I've seen some reports on PC Gaming Wiki which state that the Direct3D version of the game has gameplay issues such as doors not opening. A quick screenshot comparison reveals some pretty bad banding in the skybox under Direct3D, but I don't see any other major discrepancies. I'd appreciate it if someone who's more familiar with this game could chime in and provide additional details.

I'm leaning toward including it just based on the skybox banding and the gameplay bugs, but I'd rather wait for someone else to confirm this as well. For the record, I did test this with my GeForce FX card as well, and banding was present there too.

Joseph_Joestar wrote on 2023-08-26, 11:18:

Games tested: Requiem: Avenging Angel - retail CD release + latest official patch 1.3
...
A quick screenshot comparison reveals some pretty bad banding in the skybox under Direct3D, but I don't see any other major discrepancies.

Also the doors. And in general it looks better in Glide, IMO.

To address earlier requests - I captured one more scene from Rayman 2, with HyperSnap (click for full-size):

Left to right - you have Direct3D, Glide without postprocessing, and Glide with Postprocessing. FWIW, the game when run with the 3dfx driver looks more like the "postprocessed" version, in terms of gamma, so I suppose it means HyperSnap is sort-of accurate (note that it has several post-processing presets, I chose the one that matches my card - Voodoo 3000).

In this scene you can also see the visible artifacting in Direct3D more in the grassy texture on the cliff directly to the right of the metal structure. That's the worst difference I've seen in my comparison so far, although I did not play through the entire game. It might also be a problem specific to the Voodoo card or the DirectX 6 driver on my machine. I'll see if I can reproduce it on a different system. For now I would not change the classification of Rayman 2 as "equal".

dr_st wrote on 2023-08-26, 11:35:

Also the doors. And in general it looks better in Glide, IMO.

To my eyes as well, but I wasn't sure if it was just my subjective perception, so I wanted to wait until someone else chimed in. 😁

BTW, if anyone else wants to test Requiem, the patching process is a bit convoluted. First, you have to apply the base 1.2 patch. Then, you need to download a specific version of the 1.3 patch (Software, Direct3D or Glide) depending on which renderer you're using, and apply that on top.

In this scene you can also see the visible artifacting in Direct3D more in the grassy texture on the cliff directly to the right of the metal structure.

Ouch, that texture distortion looks pretty bad. If you can reproduce it on a non-Voodoo card, I think that would qualify Rayman 2 for "looks best in Glide".

Games tested:

• Requiem: Avenging Angel - retail CD release + latest official patch 1.3

Graphics cards tested:

• 3DFX Voodoo3 2000, using 3DFX reference drivers version 1.07.00 (Glide renderer)
• Nvidia TNT2 M64 using Nvidia reference drivers version 3.68 (Direct3D renderer)

Did some more testing with Requiem and I found a bug which only appears in Glide mode. Take a look at the lion statues in these screenshots. The shadowing is correct under Direct3D while it's broken under Glide. So this game appears to be buggy in different ways on different renderers. I suppose it can go into the "equal" category since it's equally bad in both cases. 😁

Star Wars Rogue Squadron 3D testing:

Windows ME Voodoo5 w/1.2 patch applied to retail CD installation. 1024x768@75Hz on a CRT. Both Glide and Direct3D render in 16-bit. I disabled DirectDraw/Direct3D completely to ensure the game was forced into Glide during testing - this game has a nasty habit of ignoring your Glide selection setting within the hardware configuration utility (and even when the user profile says Glide is in use, once 3D begins, Direct3D is still activated). I checked the 1st two stages.

There is a slight edge to Direct3D as color banding/crush on the stage bitmap background is of lower quality when using Glide. However, you're unlikely to have noticed if you weren't looking for it. Otherwise, lighting, explosions, textures, fog, and so on are equal.

the glide shot looks quite green tinted there.

auron wrote on 2023-08-27, 00:00:

the glide shot looks quite green tinted there.

I hadn't noticed during gameplay. I wonder if it's the screenshot software interaction with the CRT?

Joseph_Joestar wrote on 2023-08-26, 14:52:

https://www.vogons.org/download/file.php?id=172237&mode=view Requiem_Lions_Direct3D.jpg […]
Show full quote

Requiem_Lions_Direct3D.jpg

Did some more testing with Requiem and I found a bug which only appears in Glide mode. Take a look at the lion statues in these screenshots. The shadowing is correct under Direct3D while it's broken under Glide.

This D3D image with shadows on the lion's statues also looks broken.

I remember the demo for this game, but in 1999 I think I've played it in software render or in D3D mode on a S3 deccelerator and it looked terrible to me even it was a good game for its time.

analog_programmer wrote on 2023-08-27, 04:41:

This D3D image with shadows on the lion's statues also looks broken.

While neither looks particularly great, Direct3D at least matches software rendering.

The shading in Glide is visibly different from both.

Meatball wrote on 2023-08-26, 23:20:

Star Wars Rogue Squadron 3D testing:

Cheers!

Added to the "equal" list.

The Wheel of Time Retail CD w/v333 patch applied:

Test Setup:
Windows ME
Voodoo5 5500 AGP w/1.07.400 Reference Driver
Gateway VX900 CRT

Surprise... the game looks better in Glide. Like Unreal (of course), there's something overall about this game that just 'pops' in Glide. "The Wheel of Time" in Glide mode has sharper textures, more depth, glassy water. Take a look at the pond shot; you'll see how crisp the lily pads appear in Glide versus Direct3D. I didn't notice any special effects missing between the renderers, but take a look at the manor photo, there's a little rock just to the right of the big rock, and it's practically invisible in Direct3D. I tried adjusting brightness and gamma, but the issue remains. OpenGL wouldn't let me select highest quality textures, so I didn't bother with a capture.

I want to check the game on a 5900XT for Direct3D to see if it's a Voodoo5 issue (forgive my lack of technical wording), but there's this weird, I don't know, it appears as some kind of "lighting box" where textures get progressively blurrier on Direct3D. There are multiple boxes, and they are obnoxiously noticeable especially when looking downward. Still images don't capture as well, but you will notice them while moving. Take a look at the Direct3D cave shot, and you'll see one of them at the second blood spatter on the ground. Glide also has at least one of these boxes, but it blends into the scene better.

The Wheel of Time Retail CD w/v333 patch applied:

Glide photos attached.

is there anything in the .ini that can be forced for the D3D renderer? particularily, it looks like detail textures are off by default, but at least in UT99 they can be forced on.