Does it pass all the DirectX tests from within dxdiag?

Yes it does. I should probably have mentioned that. The cube went quite a bit faster in hardware accelerated mode vs software.

I've added this card to my 486 overkill graphics list. Did you try lesser versions to fix the display properties control panel errors?

I just checked the inf files, the NVS280 PCI was not supported in an earlier Nvidia driver version than 56.xx.

I've now put the TNT2m64 PCI in there as well as a AMD 5x86-133mhz cpu. I got powerslide working ok but it's unplayable.
Only MTM2 is playable, but realistically too slow to bother with. I ran 3DMARK99 through and got 14 for the video card and
4 for the cpu. It looked as though the scores would be a bit higher than that.
In Dos the SVGA games like Duke3d are now back to working as well as the S3 Trio64v+ did. I think it's mostly games that
ran ok on a Voodoo 1 that have any chance of being playable in Direct3d. I might try Turok Dinosaur Hunter at some point.

I am still very early on in my 3D accelerated games list which work well in an extreme socket3 system, however Outlaws and Forsaken play well with D3D on a Matrox G200. I had put this testing to rest for quite some time but am trying to pick it up again.

The issue with using 3DMark99 is that CPUs without MMX or CPUs which are non-Pentiums will have their graphics score markedly lower by a factor of 20 for graphics and by a factor of 80 for CPU marks.

For example, a Cyrix 5x86-133 is generally a little faster than a Cyrix MediaGX-133, however the MediaGX-133 scores 304 in graphics, while the 5x86 scores 15. The MediaGX scores 419 in CPU marks, while the 5x86 scores 5. This considering, 3DMark99 is not a good comparative benchmark to use for non-Pentium or non-MMX CPUs. Even a Cyrix 6x86 at 133 MHz scores only 17 graphics and 6 CPU marks.