I am kind of curious how the glide API fits in with column based renderers, because clearly the world still has the 2.5D representation on the CPU side with vertical shearing effects. Does it treat each column as a pair of triangles and convert the texture co-ordinates?

jh80 wrote on 2026-10-28, 05:26:

BitWrangler wrote on 2026-10-27, 14:25:

I don't get why anyone would put a V3 in anything slow enough to not run Duke at 800x600 well on a 2D card. It only takes a P166 or so and VBE 2.0 card.

While I get the point you're making... I don't think this statement is accurate at all. Can you confirm? Even on a P200 MMX I probably only get 20-30 fps at 800x600 on Duke3d. If you want to play at anything like 60 fps you need a much faster CPU. So, it is possible that someone may have a period-accurate Voodoo 3 in a system that otherwise could not run Duke3d at 60 fps in 800x600.

I am sorry I didn't specify that it didn't meet the arbitrary post year 2000 FPS standard. 1990-1999 if it did 15+ FPS it was playing well.

Anyway, wouldn't be having silly arguments if it was clearly stated what things actually achieve when you release them instead of us having to apply the subtext "Here's a piece of random AI generated crap I am not even sure what it does" because that's what it sounds like when details are withheld.

Edit:Hmmm do code box edits show as edit? because I swear there's more details now in first post than there was.

BitWrangler wrote on 2026-10-28, 12:51:

1990-1999 if it did 15+ FPS it was playing well.

Are you guys aiming for 60 fps on vintage DOS games? On my hyper 486's, I get average 17 fps in Duke Nukem 3D with sound enabled (640x480). It feels good enough for me. My kids can't stop playing it and never said anything about it feeling slow. Neither myself, nor my children, have any experience with modern day action games, so we aren't expecting life-like movement.

My software mode Duke Nukem 3D benchmark results are here, Re: The Ultimate 486 Benchmark Comparison
Anyone know if this Duke3D benchmark will work with the Voodoo3 patch?

TgamesFR wrote on 2026-10-27, 09:33:

For others games i will concentrate maintly on Voodoo 1 exclusives games.
My goal is to make the Voodoo 3 the 100% compatibility card over every 3DFX patches.
And also support few "high-spec" games who need the power of the Voodoo 3 to run correctly (like Duke Nukem 3D and Screamer).

Very interesting, as this happens to be while I still contemplate the merits of whether to keep my Voodoo3 or Voodoo2 SLI.

If you can make EF2000 work on Voodoo3 that will pretty much settle it: I get one more precious PCI slot, and the SLI setup will sell for more anyway.

feipoa wrote on 2026-10-28, 17:16:

Are you guys aiming for 60 fps on vintage DOS games? On my hyper 486's, I get average 17 fps in Duke Nukem 3D with sound enabled (640x480). It feels good enough for me. My kids can't stop playing it and never said anything about it feeling slow. Neither myself, nor my children, have any experience with modern day action games, so we aren't expecting life-like movement.

My software mode Duke Nukem 3D benchmark results are here, Re: The Ultimate 486 Benchmark Comparison
Anyone know if this Duke3D benchmark will work with the Voodoo3 patch?

As someone who played Duke3d quite a lot on a 486 when it came out, 17 fps is rather unbearable. Did you have your kids compare with 320x200?

Of course people tolerated low fps back in the day. Go back to 1989 and play Mechwarrior and you'll be satisfied with 5 fps.

Once games like Quake and Descent came out, and with online play, there was certainly more discussion about sufficient frame rate. You can read all of this discussion from 1996 still on Usenet. For Duke3d, I see a lot of Usenet comments suggesting 25-3o fps minimum is ideal.

Once 3D hardware started taking off in 1997 and gaming benchmarks started becoming more common alongside sites like Anandtech, then the push for 60 fps really entered the industry consciousness.

But I think you'd be hard-pressed to find a serious Duke3d player from 1996 who preferred 17 fps 640x480 vs. 60 fps 320x200. The game is simply much harder to play at low fps.

jh80, how are you determining your Duke3d framerates? When referring to 60 fps at 640x480, are you able to run the benchmark in the thread I liked previously? It uses a consistent timedemo and takes the average. If you are using the instantaneous frame rates from "dnrate" command, then the numbers are different. For example, when I load LA Meltdown with default settings at 640x480, I tend to see 15 fps as the low and 40 fps as the high (e.g. falling off the building). The normal walking about frame rate is around 21 fps.

Back in the early-mid 90's, an unspoiled kid would have been happy to play at almost any framerate.

feipoa wrote on 2026-10-29, 01:15:

jh80, how are you determining your Duke3d framerates? When referring to 60 fps at 640x480, are you able to run the benchmark in the thread I liked previously? It uses a consistent timedemo and takes the average. If you are using the instantaneous frame rates from "dnrate" command, then the numbers are different. For example, when I load LA Meltdown with default settings at 640x480, I tend to see 15 fps as the low and 40 fps as the high (e.g. falling off the building). The normal walking about frame rate is around 21 fps.

Back in the early-mid 90's, an unspoiled kid would have been happy to play at almost any framerate.

I haven't measured frame rates in Duke3d in a long time. I was basing my comment on my memories of preferring to play Duke3D at 320x200 on my 486.

"an unspoiled kid would have been happy to play at almost any framerate" -> Yes, but in this case no one is forcing you to play at 17 fps, since you have the option of going to 320x200. I vastly prefer smoother game play over higher resolution. As I said, I think the game becomes more challenging to play at a high level at low fps.

Just so we're clear that people back in 1996 did in fact care about frame rates in Duke3d, here's a snippet of the many Usenet posts you can read about the topic:

https://usenetarchives.com/view.php?id=comp.s … ndWVscGguY2E%2B

"An optimal framerate is 30 fps or above average... I have seen some posts that claim 12-15 fps is smooth to them, but that is unplayable for me......"

"Since standard television goes at about 24-30 fps (frames per second), if you have a framerate number in that range you are doing good. If we want to talk playability, anything over 15fps is playable."

In the pre-3D accelerator era, it's clear that people's expectations of what "good frame rates" were was much, much lower. The reason being that there simply was no way to attain very high frame rates.

Skip ahead to 1997/1998 and you can read LOTS of people complaining of games "only" running at 40 fps on Voodoo cards (Heavy Gear comes to mind as having lots of complaints on Usenet). The expectations shifted really dramatically in just a year or so.

It isn't clear how you were measuring framerates in Duke3D. Maybe the timedemo I used, which yielded 17 fps at 640x480, would equate to 25 fps in whatever method you used. If we aren't using the same measuring scheme, we are only comparing subjective anecdotes.

On my Cyrix 5x86 system running at 136.5 MHz, I prefer 640x480 with slower framerates compared to 320x200 and faster rates. But there is a limit; 800x600 yielded too slow of a framerate.

There will always be some individuals who demand the fastest results, be it now, or back in 1996. At this time, a large number of individuals still had computers from the early 90's and would have been thrilled if any more modern game ran at all.

BaronSFel001 wrote on 2026-10-28, 20:01:

TgamesFR wrote on 2026-10-27, 09:33:

For others games i will concentrate maintly on Voodoo 1 exclusives games.
My goal is to make the Voodoo 3 the 100% compatibility card over every 3DFX patches.
And also support few "high-spec" games who need the power of the Voodoo 3 to run correctly (like Duke Nukem 3D and Screamer).

Very interesting, as this happens to be while I still contemplate the merits of whether to keep my Voodoo3 or Voodoo2 SLI.

If you can make EF2000 work on Voodoo3 that will pretty much settle it: I get one more precious PCI slot, and the SLI setup will sell for more anyway.

Yes EF2000 is currently beeing ported to Voodoo 3.

First i will release Battle Arena Toshiden who was also a Voodoo 1 exclusive game then EF2000.

Plans are really make every Voodoo 1/2 patches working on Voodoo 3 and also at same time improve a lot performances compared to V1/V2 lonely.
Like this the Voodoo 3 is the best compatible card.

And bonus few new games... =)
System Shock 3DFX is coming for example...

I know the side effect will be the Voodoo 3 will be even more desirable on the second hand market now xD.

Looking forward to trying this out on my AMD 5x86 rig equipped with a Voodoo3, nice work!

I know this sounds crazy, but anyway to allow a 512x384 or 320x200 or similar lower resolution modes? The game looks great and runs really well just looking to eek out a bit more performance out of my 200MHz 486.

I was able to test this patch on my Cyrix Media GX1-300 machine. I can confirm that the Voodoo3 patch is functional with a Voodoo Banshee, however the framerate is worse than software mode. At 800x600 and comparing identical scenes using the in-game dnrate command, I would get 16 fps with the Voodoo3 patch and 28 fps in software mode.

Next, I wanted to try Dukemark, using dukemarke /bqbench1. At 800x600, in software mode, I got an average of 27 fps (min: 15, max: 49). When running the Voodoo3 patched version, I received an identical result, but it looked to me like the timedemo was still running in software mode.

With the Banshee, playback felt more laggy, although the objects were more realistic looking.

Perhaps the patch isn't well matched for the Banshee. Next, I will run a Voodoo3 3000 on a Cyrix 5x86-133/4x system to see if the patched version also runs slower than in software mode.

I ran the Voodoo3 patch on a Cyrix 5x86-133/4x system. Similar to the other system, using the patch reduces the frame rates. Using dnrate and comparing the same scene (800x600) again, the patch yielded 7 fps, whereas software mode yielded 12 fps. This is curious because GLQuake yields exponentially better performance compared to software.

I tried with himem and with/without emm386, but the results didn't change.

Chadti99 wrote on 2026-10-30, 01:00:

I know this sounds crazy, but anyway to allow a 512x384 or 320x200 or similar lower resolution modes? The game looks great and runs really well just looking to eek out a bit more performance out of my 200MHz 486.

This Duke3D should run faster on older hardware, but it has nothing todo with 3dfx version patch. https://github.com/Darkcrafter07/Duke07

feipoa wrote on 2026-10-31, 14:12:

I ran the Voodoo3 patch on a Cyrix 5x86-133/4x system. Similar to the other system, using the patch reduces the frame rates. Using dnrate and comparing the same scene (800x600) again, the patch yielded 7 fps, whereas software mode yielded 12 fps. This is curious because GLQuake yields exponentially better performance compared to software.

I tried with himem and with/without emm386, but the results didn't change.

After seeing the ROTT patch, I'm now fairly sure the way it's working is converting columns into triangle pairs which generates a significant amount of textured triangles for the graphics card to process. I think the banshee is actually hitting a throughput limitation. The method is somewhat similar to what Doom PS1 used as a method to bypass the cpu limitations of software rendering by offloading columns to another processor.

A "true" patch would go deeper into the rendering pipeline and convert wall/floor/ceiling elements to triangle pairs before they are columnised which is likely what was done with the original blood 3dfx patch, it could run smoothly on a voodoo 1 since that would require much less triangles to render.

This is not a problem for games that already use triangle based rendering (E.G. Tomb Raider) since their triangle budgets were already tuned for hardware of the time

Using the same comparitave resolution, has anyone else compared the frame rates with software mode vs. the Voodoo3 patch?

The differences seem to be variable. I think the voodoo3 is even bottlenecked at the higher end of fps due to the (I'm speculating) conversion method. Going around the first level the framerate is higher in software mode.

Going thorugh some of the menu demos where the load is much higher, where the software renderer dips to high 20s (~28fps) the 3dfx version stay around 40 minimum and hits 60fps regularly where as the software version hovers around 35-45.

On my specific system, the increase in minimums makes the 3dfx version play better. But it does feel like there is a lot of room for optimisation. Dukemark doesn't seem to use duke3d.exe (it's probably itself a patched version of duke3d.exe) so those results aren't useful.

I tested the patch on another system at 800x600.

Cyrix MII running at 300 MHz. Voodoo3 3000 AGP. The colours aren't displaying properly for some reason.

Looking at the numbers with dnrate, the system is ~40% slower with the patch than without it.

In software mode, everything played really smooth. Dukemark showed:
Average = 48 fps
Min = 19 fps
Max = 87 fps

Not sure why the colours were messed up in DOS with the Duke3d Voodoo3 patch. The colours were fine in GLQuake under Windows 98.

Update: 01/11/2026, V1.00a

1Many bugs fixes related to last reports
2Lot of performances optimizations (for better FPS)

Performance optimisations are a mixed bag, there is improvement when out on the street but the rooftop at the start is now only 20fps

Also I can see now that the framerate is soft limited to 60, which is probably the right thing to do. But since there's no framerate limited software version it is a bit tough to compare out right performance figures.