Run some DOS benchmarks to get an idea of what speed these games think the system is. Speedsys, Doom, etc.

Ran now a few on the Pentium Dual Core E5800 / ATI Radeon 9800 Pro system:

1. 3DBench 1.0c: 800.0 FPS (I guess this is the upper limit)
2. PC Player Benchmark 640 x 480: 66 FPS
3. Speedsys CPU score: 4413.99 (miles above an Athlon 1400)
4. Doom high details: 2134 gametics in 412 realtics (so ~181 FPS)
5. Quake 640 x 480: 60.4 FPS

Try Descent 1 and Wing Commander 1 for the ultimate speed sensitivity test.

You can use the demo version of Descent if you don't own the full game.

Yeah, tested Wing Commander 1 now and it's way too fast (which I expected - though it's not a game I normally play, that's why I did not test it in the first place). So for such ultra speed sensitive games, you need two things:
1. disabling L1 cache with setmul (which I just tested, and I get 18 FPS in 3DBench 1.0C - so not quite the equivalent of a 386, but pretty close). Not sure if it would be enough for Wing Commander 1, though.
2. a different sound card, that does not use software emulation (haven't tested yet, but I'm guessing an Aureal Vortex 2 should get the job done).

Still, I'm impressed with how compatible these PCs are! And it's still a mystery WHY they work so well in DOS.

bloodem wrote on 2020-07-04, 14:59:

Now, for whatever reason, these two work perfectly in Win98 DOS

Are you talking about pure DOS or a DOS box under Windows 98?

bloodem wrote on 2020-07-04, 14:59:

- Prince - works great on both systems - Prince 2 - works great on both systems - Lion King […]
Show full quote

- Prince - works great on both systems
- Prince 2 - works great on both systems
- Lion King - works great on both systems
- Wolfenstein - works great on both systems
- Duke Nukem 3D - works great on both systems
- Doom / Doom 2 - works great on both systems

These games, at least, as far as I know, are not speed-sensitive. Lion King does have a shorter time at the title screen if the CPU is fast (making it harder to enter the cheat code / go into the key configuration menu), but during the game itself it always plays at the same speed. As far as I remember.

dr_st wrote on 2020-07-04, 18:30:

Are you talking about pure DOS or a DOS box under Windows 98?

pure DOS (MS-DOS 7.1)

dr_st wrote on 2020-07-04, 18:30:

These games, at least, as far as I know, are not speed-sensitive. Lion King does have a shorter time at the title screen if the CPU is fast (making it harder to enter the cheat code / go into the key configuration menu), but during the game itself it always plays at the same speed. As far as I remember.

That is partially correct, however, they are speed sensitive in other ways:
- games like Jazz Jackrabbit, Monkey Island, will refuse to run without the TPPATCH on faster CPUs (runtime error 200 - a known Turbo Pascal error). From my experience, Jazz also seems to have some mild sound issues on faster systems (but not on these two PCs).
- Prince of Persia 2 also refuses to start on faster systems, but with a different error.
- Lion King has sound issues on Pentium 3 and faster PCs (it starts with sound, but it goes quiet after the menu loads - at least on the sound cards that I tried)
- Lemmings, Titus the Fox, Loom, Golden Axe also have no FM sound on faster systems (most of them will revert to PC speaker).

Anyway, on these two PCs, even though they are much faster than a Pentium 3, things just work, without having to resort to tricks like throttling, patches, etc. This is what amazes me. 😀 The most likely explanation is that the Audigy 2 ZS's emulation is very well thought out... although it's still weird that even the Turbo Pascal error is somehow avoided.

Fun fact, I now switched to the Intel Core 2 Duo E4700 / GeForce FX 5900 PC, and ran benchmarks on it too. Seems that the GeForce FX is less DOS compatible than the ATI Radeon 9800 Pro. Very interesting, I would have never guessed!

1. 3DBench 1.0c: 421 FPS (compared to 800 on the ATI Radeon 9800 Pro)
2. PC Player Benchmark 640 x 480: 72 FPS (higher framerate than the Radeon, however there's major image corruption)
3. Speedsys CPU score: 3827
4. Doom high details: 2134 gametics in 843 realtics (so ~88 FPS --> more than twice as slow than the Radeon)
5. Quake 640 x 480: 52.1 FPS (tested also with Fastvid: 174 FPS)

Also tested quite a lot of games, and for the most part they work fine. However, for some reason, Doom behaves very strangely on this system: erratic mouse movement, random key presses (that were not actually pressed). If I didn't know better, I'd say that it's some kind of CPU speed issue, however it runs fine on the Pentium Dual Core E5800 (which is faster). It might also be related to the motherboard (different revision number).
I'm actually tempted to switch video cards and see if the behavior is different - however, I have a rather... "unique" (translated: bad) wire management that I'm not too happy to mess with 😀

PCPBench starts to fall to bits (when testing the LFB modes) once it gets over 70fps, with partially rendered scenes being displayed and if you watch really carefully you can see it printing "divide by zero" errors at the top of the screen before redrawing over it.

I see. Well, I ran it again, and I'm not seeing any "divide by zero" errors... just flickering and artifacts. But, yeah, it seems to just be an issue with this particular benchmark. The other benchmarks / games that I tested are fine (except Doom).

Do the Turbo Pascal games run every time?

Yes! It's bizarre! 😀

maybe it's so fast it wrapped around the error? i.e. instead of trying to divide by zero it ended up underflowing and dividing by something really big

It's quite possible, yes. 😀