Any ports where pci express video cards won't generate an improvement from agp?

I mean, I imagine they speed up software rendering on most any port due to the widened bus alone.  Also, does using a (relatively modern) regular pci video card, like a geforce 5200, make a cap on the framerate in the same way that an 8-bit isa video card put a cap on the framerate of vanilla?

The 35fps cap is because the game was optimized for that speed, they could've chosen to make it 70 (like Wolfenstein), but the target hardware was only able to both render and run game logic about as fast as 35.

As for your actual question, I don't know enough to answer it. If I had to guess, I'd say it probably won't make a world of difference, especially if you're using a source port with a simpler GL renderer.

2 hours ago, invictius said:

Also, does using a (relatively modern) regular pci video card, like a geforce 5200, make a cap on the framerate in the same way that an 8-bit isa video card put a cap on the framerate of vanilla?

Doom's isn't capped at 35Hz, it is 35Hz. In the same way Rocky is the Eye of the Tiger.

The question is way too unspecific.

Some ports have uncapped frame rate others do not.

For uncapped frame rate, the entire performance characteristics of the system are important, but the graphics card is still only one piece - CPU makes far more of a difference.

For ports that are capped to 35 fps, the selected screen resolution will be a deciding factor as well, because the more pixels are being rendered, the more power you need - but again the CPU itself is the most important thing.

Which leaves the ports that stick to low screen resolutions, up to 640x480. AFAIK those have been maxed out by the hardware in the last millennium already, so having a better graphics card should make no difference.

Something else to consider. It is useless to render faster than your monitor's refresh rate, cause those frames will be invisible. The best scenario is to lock the game at the speed of the refresh. If you can render slightly faster than the monitor refresh rate, and use vertical refresh sync, you should be able to get output at the best possible speed, free of tearing artifacts. That's as good as it gets.

Any port that's CPU bound, really, which means practically all of them. Even hardware accelerated ports nowadays are pretty constrained by factors external to the GPU, so even if you had an infinitely fast video card/0-time rendering, you would still be constrained by gameplay calculations. AGP vs PCI Express would not give a significant advantage here, except in very specifically crafted examples, e.g. specific map geometries, extra effects and drivers combinations.

The case of vanilla Doom benchmarks in pure DOS/VGA mode is really a pathological case, a product of the ISA legacy bus where the CPU grossly outperforms the video/data bus's capabilities.

If a port with a radically new architecture appeared, that offloaded a significant part of e.g. geometry processing, BSP traversal etc. to the GPU, then it would make sense to start exploring new limits.