What demo are you using to get these results?
Not sure what the point of this thread is, are you asking a question? The card is definitely bottlenecked by the K6, although it's not a bad matchup at all. Why did you disable the motherboard cache though?

demo1.dm2 is what I'm using, and I've struggled over the last week + getting this thing to run stable over time. It's shown moments of stability then it craps the bed. And when I started with a TNT and thought maybe some of the issues were related so I picked up the TNT2. As for the point of the thread, it's more of looking for a confirmation of my findings. As stated, it's been a long time since I've used this board/cpu combo.

I don’t have a TNT2 to test, but my K6-2+ / GeForce 2 MX PCs can do 80+ or even 90+ FPS @ 1024 x 768. The CPU is still the limiting factor, but in your case something else might be going on. Make sure that you are not using any PCI cards (like USB 2.0 adapters), those tend to have issues on SS7 and can have a big impact on performance.
Also, use older drivers, newer drivers can affect performance considerably.

Indeed. Perhaps try to enable external cache and also use the Quake 2 3Dnow! patch and 3Dnow! OpenGL in the menu. Out of curiosity, what driver are you using for the TNT2?

In my case, I’m not even using the 3DNow patch (the performance improvement with the patch when using a GeForce 2 is minimal - 2 FPS or so)
Anyway, it does seem like performance is improved with GeForce cards (most likely due to Hardware T&L, which can improve performance even if the game doesn’t natively support it). But still, according to Phil’s video, you should still see 60 FPS with a TNT2 Pro.

Are you sure? I find it hard to believe you are achieving ~90FPS in demo1 on that CPU without the 3Dnow! patch. I'm fairly certain HW T&L does not affect the performance in this game at all.

Yes, definitely sure. My LuckyTech P5MVP3 board with a K6-2+ 550 overclocked @ 600 MHz with tight RAM timings (and with a specific RAM module, because not all PC133 modules are the same, apparently, according to my tests) + a GeForce 2 MX are able to achieve 92 FPS. The other boards that I have can push 80+ FPS when paired with a GeForce 2 MX (also without the 3DNow patch).

Garrett W wrote on 2020-06-04, 06:38:

Are you sure? I find it hard to believe you are achieving ~90FPS in demo1 on that CPU without the 3Dnow! patch. I'm fairly certain HW T&L does not affect the performance in this game at all.

It most certainly does (not the "L" part, but at least the "T"), see Phil's video from my previous post and also this article from Anandtech:

Provided that one of those two above stipulations are met, the next requirement is that the game must not use its own transforming or lighting engine. This immediately takes Unreal Tournament out the running because it uses its own T&L engine. Most other games use their own lighting engines and therefore can't take advantage of the 'hardware lighting' part but can take advantage of the 'hardware transforming' part of the T&L equation. Right off the bat, this will give games like Quake 2 and Quake 3 a small boost in performance.

So, the way I see it, the 3D now patch helps the CPU with the "Transform" part (as in, taking the vertex 3D scene and transforming it in 2D, which is the final image you will see on your display). But if you have a GPU that (in certain scenarios / games) can take care of this, the 3DNow patch apparently has very little impact.

Interesting, good to know!