Thanks to you for making this wrapper and all its fixes!
I gladly will test more games,just when is the question.

Manx TT Super Bike fps seems better from screenshot, than some other game. Once i shared XGL200 version that works faster, later on maybe its possible look how this glidetodx6d3d wrapper version gets more fps at its limited 3dfx game support, if its possible to do and take some hints from it.

NFS2SE is totally playable at Rage XL AGP with it and no OC, all max settings just some small ingame screen issues, that you fixed already in your alpha versions. F2 better car look looks better with its shadow at back of the car and most parts look good. Later XGL200 builds gets slower and no visible visual improvements for me. I can share drivers that i use and xgl200 i use.

dgvoodoo setup does not open and i may need dx9 first or some extra files for Win95. Some games start with it, like pandemonium 1 with its glide.dll. Better to install Win98se and try it again. Your Glide2CIF wrapper gets better by every new release and if the game testing helps, i am happy to help more and more, Thank you!

Speed issue can be 3DCIF vs D3D thing? Some games are faster with D3D and slower with 3DCIF.
Turok 1 is one such game, it has its own dll file and it did not get any updated version of it @ CIF.
3DCIF can be and usually is fast, at Tomb Raider example and with 800x600 resolution @30fps.
Not all Game have option to lower the resolution of the Game and 640x480 or 512x384 is good.
I have used also 400x300 and its not that low, but sweet spot is 512 or 640 for me at ATi Rage3.

I use 5.40/2611min or barebone drivers at the moment and i have updated view files tothemax.
I am sure its tiny bit better than driver 2548,that you have used for your glide2cif game testing.

I will try your new glide2cif wrapper version out, thank you!

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

1. Manx TT Super Bike has 20 to 30 fps with Rage XL AGP and my Celeron PC (latest glide2cif alpha at the moment - for me its v4)

Alpha software tends to focus on other things at the cost of FPS. Even so I see Need for Speed 2 is running about twice as fast now as it was a couple of days ago, even though I've done no optimizing for it specifically. Just tells you there's performance potential for later. For now it's best to use a fast CPU to make sure the cost of translation is masked if you want a more accurate idea of the efficiency of the render path.

There's not inherently anything special about CIF vs others, it's an API to make development easier. It may or may not have been designed primarily to make optimal use of the hardware over, say, developer convenience. And if you think about it Glide and Direct3D as names suggest performance is the goal but ATI3DCIF just stands for ATI's 3D C API.

The driver I'm using for testing is actually pretty bad, doesn't work in Direct3D at all. But does the business in CIF.

San Francisco Rush: Extreme Racing 1997 (glide2cif: Unsupported Glide texture format 0.)

could be useful to learn something more about ati3dcif from this wrapper
https://github.com/ata4/glrage

Updates:
- add support for fog (except table fog for now)
- add support for the FX_GLIDE_SWAPINTERVAL environment variable
- add support for GR_TEXFMT_RGB_332
- use blitting instead of page flipping

FPS in GLQuake increases from 23 to 27 with a forced swap interval of 0. ATI's own OpenGL driver gets 32 or so, but none of the three versions of it I've tried has rendered the game correctly. The wrapper's starting to look like a better way to run the game.

The GR_TEXFMT_RGB_332 implementation is untested since I don't have a game that uses this format.

marxveix wrote on 2025-12-30, 17:07:

San Francisco Rush: Extreme Racing 1997 (glide2cif: Unsupported Glide texture format 0.)

Format 0 is RGB332 which was now added. But I don't think this game would run on the original Voodoo.

RaVeN-05 wrote on 2025-12-30, 18:14:

could be useful to learn something more about ati3dcif from this wrapper
https://github.com/ata4/glrage

Maybe, but they're wrapping away from CIF rather than to it so I'm sceptical. These days AI is good enough at coding that general programming is pretty trivial, but undocumented or obscure information about the hardware or API could be useful.

vvbee wrote on 2025-12-30, 23:46:

Updates: - add support for fog (except table fog for now) - add support for the FX_GLIDE_SWAPINTERVAL environment variable - add […]
Show full quote

Updates:
- add support for fog (except table fog for now)
- add support for the FX_GLIDE_SWAPINTERVAL environment variable
- add support for GR_TEXFMT_RGB_332
- use blitting instead of page flipping

The GR_TEXFMT_RGB_332 implementation is untested since I don't have a game that uses this format.

marxveix wrote on 2025-12-30, 17:07:

San Francisco Rush: Extreme Racing 1997 (glide2cif: Unsupported Glide texture format 0.)

Format 0 is RGB332 which was now added. But I don't think this game would run on the original Voodoo.

Now it shows game screen for half second and then format 1.

vvbee wrote on 2025-12-30, 23:46:

Updates:
- use blitting instead of page flipping

FPS in GLQuake increases from 23 to 27 with a forced swap interval of 0. ATI's own OpenGL driver gets 32 or so, but none of the three versions of it I've tried has rendered the game correctly. The wrapper's starting to look like a better way to run the game.

For me page flipping or breakpageflipping option works best and i use it with my OpenGL settings with ATi OpenGL ICD as well. If you get Q1 32fps with fast PC then you dont use Multitexture with ATi. Some settings dont work well with older ATi OpenGLs and from older OpenGLs i recommend OpenGL 1073, from middle its 1091 and newer 1107 V1 maybe, but 1107 V2 is the latest that i know for Win9x, maybe its better to rename it to 1108. Benc option is only for Bencmarking, it does not behave well at games most of the times, all other settings are more safe side and i use Fast and Max settings the most, Default is similar that you get from ATi out of the box or if you delete atidbg.ini file from C:\Windows folder

OpenGL game did get slower results from my Quake1 run and Slot1 Celeron
24.6fps glide2cif V4 previous version
23.5fps glide2cif V5 latest version

Similar Multitexture was made for OpenGL also from ATi, but download links are not working.
ATi RagePro cant handle all its options at the same time, RageXL maybe bit better, if selected
right combination, it probably can make it to the playable framerate, at least with ATiRageXL.

marxveix wrote on 2025-12-31, 06:59:

Now it shows game screen for half second and then format 1.

Game requires a Voodoo Banshee so not supported anyway. Format 1 is NCC compression, but I'm not aware of any other PC Glide game that uses it. Unless there is and it supports the Voodoo I'll leave it out for now.

marxveix wrote on 2025-12-31, 07:32:

OpenGL game did get slower results from my Quake1 run and Slot1 Celeron

Assuming you mean blitting vs page flipping. On my Pentium 4 test machine I see better FPS with blitting at 640 x 480, maybe slower CPUs need to use lower resolutions to see a benefit. With blitting it's more straightforward to implement some other things so I'll leave it in for now.

vvbee wrote on 2025-12-31, 09:58:

marxveix wrote on 2025-12-31, 07:32:

OpenGL game did get slower results from my Quake1 run and Slot1 Celeron

Assuming you mean blitting vs page flipping. On my Pentium 4 test machine I see better FPS with blitting at 640 x 480, maybe slower CPUs need to use lower resolutions to see a benefit. With blitting it's more straightforward to implement some other things so I'll leave it in for now.

Used RageXL, where its limit comes later than Rage Pro. I check Rage Pro results with blitting and i can boot up my P4 machine with universal AGP.

Did test with the Rage LT Pro 8MB with OpenGL at Celeron 548Mhz.
Quake 640x480x16 timedemo demo1
23.4 older version glide2cif
22.2 newer version glide2cif

Just saying, NGlide 0.91 and 0.95 versions added some San Francisco Rush support. Next i try some other Glide game.

Boss Rally 1999 (glide2cif menu works, game has black screen)
https://en.wikipedia.org/wiki/Boss_Rally

A 5% drop in FPS is acceptable. But if I were you I'd set swap interval to 0 and rerun the benchmark, very possible the new version comes out on top then because the old one doesn't support this.

nGlide has different priorities for the moment. But the black screen issue needs fixing in general. Sometimes though it's because there's an error popup that you just can't see, so hit Alt+Tab to make sure that's not the case.

The inability to set the DirectDraw video mode comes up with some games as well, don't know why. Happens with NHL 98's demo as well.

vvbee wrote on 2025-12-31, 15:46:

A 5% drop in FPS is acceptable. But if I were you I'd set swap interval to 0 and rerun the benchmark, very possible the new version comes out on top then because the old one doesn't support this.

Thanks for the answer! Every frame is needed for lower videocards 😀

Dumb question again, where i set swap interval to 0? Thanks
What ini file and where i add it with SwapInterval=0 in the ini?
http://www.o3one.org/hwdocs/video/voodoo_graphics.pdf

edit:
Wait a second, i will add it to the AUTOEXEC.BAT
SET SST_SWAP_INTERVAL=0

edit2:
Did not change fps results, what else you have there?

I had MiniGL 1.47,1.49 gives the same 22.2fps as 1.47
https://3dfx.retropc.se/minigl.html

Added two of your latest glide2cif wrappers here, hope its ok.
glide2cif V4=page flipping
glide2cif V5=blitting

FX_GLIDE_SWAPINTERVAL, you can also do it in the command prompt and it'll be active for what you run from the prompt until you close it. The OSD will say SWAP_0 or whatever value.

vvbee wrote on 2025-12-31, 17:02:

FX_GLIDE_SWAPINTERVAL, you can also do it in the command prompt and it'll be active for what you run from the prompt until you close it. The OSD will say SWAP_0 or whatever value.

Ms-dos command prompt and set FX_GLIDE_SWAPINTERVAL=0 did the trick, big thanks, now fps is higher.
Can you integrate it to the Glide2x.dll someday at your newer releases,for me command prompt is also good.

I just make .bat file with set fx and game exe inside.

-----------------------------------------------------------
Celeron 533@68x8 (glide2cif v5)
GLQuake 28,8fps Rage XL , all settings same, 640x480x16 timedemo demo1
GLQuake 26.0fps RagePro,all settings same, 640x480x16 timedemo demo1

Running GLQuake with -nomtex to disable multitexturing on ATI's native OpenGL driver drops the FPS from 33 to 28. Also fixes the inverted lighting on my driver version. But at 28 FPS it's only 1 FPS higher than the wrapper at 27 FPS, so most if not all the difference is in multitexturing. With the wrapper one possible easy optimization would to render small triangles like particle effects on the CPU, like ATI's documentation suggests. Don't know how much a difference it would make but I notice you can have 1000+ extra triangles from explosions, so probably accounts for something. It's a bit of a cheat but ATI's recommendation, maybe they're doing it in their driver for that extra 1 FPS.

Turns out Gex is another game that uses NCC compressed textures, at least a demo of it does.

Great that other glide games did get also boost, not just OpenGL games with v5 and this glide command, its vsync on and off? Totally working Table fog support would be magical with 8 bit paletted textures, even if its starts normally only with rage xl. I think any driver up to 4.11.2582 and table fog does not show up on any card, including xl. Also it does not show up with any later driver with Rage Pro, only xl starts to show something bit later and bit broken at Win9x drivers.

Glide Manx TT Superpike seems totally playable now with Rage XL AGP, just text has some black background.

vvbee wrote on 2026-01-01, 00:12:

Running GLQuake with -nomtex to disable multitexturing on ATI's native OpenGL driver drops the FPS from 33 to 28. Also fixes the inverted lighting on my driver version. But at 28 FPS it's only 1 FPS higher than the wrapper at 27 FPS, so most if not all the difference is in multitexturing.

ATi was not at its best with OpenGL. 8 bit paletted textures are not enabled or not working with their OpenGL ICDs and at D3D its automatically enabled. One point that it was disabled, that it cant do 2 or 3 things at the same time, multitexture and paletted texture and mipmapping. 8bit paletted textures is not that big loss vs performance gain from multitexture.

vvbee wrote on 2026-01-01, 00:12:

Turns out Gex is another game that uses NCC compressed textures, at least a demo of it does.

Can it be tricked, like it has this feature and instead ncc compression it uses no compression or VQ Compression

Swap interval is like vsync sort of but it's supposed to be controlled by the game. I doubt there would be hardware table fog but AI's suggestion earlier was to emulate it by baking it into vertices at runtime, some caveats but seems obvious enough as a workaround. Potentially better than nothing but may be exactly what ATI were doing if later drivers had some. Haven't gotten around to it though. Gex is running thanks to Mr. AI implementing the 3DF and NCC routines, but needs some tweaking.

Maybe this Fog in Direct3D.pdf helps a bit, Riva128 came out at the same time with Rage Pro or 1 month later.
Maybe this can be done some later Rage3 cards ( XL / XC ), Riva128 emulates table fog using vertex based fog.

I wouldn't worry about it, the main thing keeping this feature from being implemented is that it's on a todo list with many other things. There's a good chance I don't have to think about it, it's a wheel that's probably been invented many times so AI will do it and then I'll verify its work, but then there are questions about edge cases like large polys etc., and that doesn't interest me right now.

What would interest me though is why Redline Racer produces no 3D and why Need for Speed 2's menu is black, even though the same CIF commands sent to a software wrapper work in both games. I did a bunch of testing earlier so the immediately obvious options to do with the APIs seem to be off the table. But it could also be something obvious in the logs that I'm just not noticing.

vvbee wrote on 2026-01-01, 14:42:

What would interest me though is why Redline Racer produces no 3D and why Need for Speed 2's menu is black, even though the same CIF commands sent to a software wrapper work in both games. I did a bunch of testing earlier so the immediately obvious options to do with the APIs seem to be off the table. But it could also be something obvious in the logs that I'm just not noticing.

Take your time and go where it pulls you, Wrapper is already in working shape and speed is better. I can wait, we all can wait.
Redline Racer color change is visible if you drive with the shift blindly,menu is working,similar to NFS2SE if with mirror view.
Demo has some logos and the colors are changing ingame with the sun reflection. Menu misses some items@RedlineRacer.
NFS2SE background colors are not the normal black,so something with the colors maybe,you know better,Thank you vvbee.
I test some other 3dfx voodoo games to know what game starts and what not, already downloaded some 3dfx game demos.

Uprising 3Dfx demo (3do) - glide2cif v5 @ Glide2x.dll
SportBaseball Edition 2000 demo (Interplay) glide2cif v5 @ OpenGL32.dll (MiniGL 1.49 and OpenGL Beta2)

The term
grtextssource refers to the source code for the 3Dfx Interactive Texture Utility Software, often known as TexUS.
TexUS was a command-line utility and library designed for developers to prepare image files for use with the 3Dfx Glide API. Its primary functions included:

File format conversion for compatibility with 3dfx hardware.
MIPmap creation, which is a technique for optimizing texture quality and performance at different viewing distances.
Support for 3Dfx's proprietary Narrow Channel Compression.

The source code for this and other 3dfx development tools has been made available in various community and developer source code releases over the years, primarily for enthusiasts and retro-computing communities (like the VOGONS forums) who work on preserving and emulating the hardware and software.