have you got an uptodate CGX v4 or v3, or just the one of the CD?
with the CGX stuff, i usually have the following...
HIRESCSR = yes
HIDE15BIT = yes
SAVEMEM = no
ALERTEMU = no (i find if set to yes, it usually crashes out completely on a recoverable alert anyway)
CPUP2C = no - having this set to yes is probably what is slowing it down
PLANES2FAST = yes
KEEPAMIGAVIDEO = no
SUPERLAYERS = yes
SUPERGELS = yes
USESEMAPHORES = yes (probably spelt this wrong)
for the graphics card options, if you're using video overlay, or 3D functions, set multimedia mem=yes.
iirc, on the A2000, in zorro2 mode, the CV64/3D only 3.5Mb graphics ram is available in total, and setting multimediamem to yes allocates 1Mb for texture memory and overlay. the remaining 2.5Mb is all thats available for framebuffer, so be carefull that your screenmodes aren't eating too much ram. drop your workbench screenmode to 8 or 16 bit to free up memory on the board, as it will probably want to cache both the workbench screen and whatever the game screen is.
getting GLquake68k to work, usually wants a shed load of ram allocated, at least 32-64Mb (--edit-- in a continuous block), and a massive stack. usually 500k-800k or above. try lowres 15bit screenmodes like 320x256 to check it all works. i have run glquake68k at 1600x1200 which is pretty much a slideshow, but very pretty, and left it run all day as a stability test and it was fine.
if you want to push it further, in the monitor tooltypes put in advanceclock=yes and memclock=70 (this is in mhz - ymmv!) you should now be able to push faster refresh rates, and bigger resolutions. but you'll probably want to heatsink the virge chip. i'm not sure what memory you have on your card, but if you can get it replaced with 50ns or 60ns ram, you may be able to get memclock up to 80mhz.
word on the street is the ViRGE DX chip is pin compatable with the standard ViRGE, and is able to take a higher clock, up to 100mhz, as well as an apparent 30% increase in 3D operations. but that its colour lookup tables are different for the first 4 bits in 8bit screenmodes, so it can look a bit funky.
if you have a rev2 DCE CV64/3D card, you may have faster ram, and or DX chip anyway :-)
i have found P96* to be fairly rapid in full screen operations, but i went back with CGX as i found some of the features like pip overlay and Warp3D were problems...
(*this may have been a fairly early rev of P96, so my pov may be a bit off ;-) )
i've currently got issues with CGX 4.7b something and warp3D 4.2. i can get the PPC and 68k gears demos to work, but nothing like quake, heretic2, or wipeout want to work without generating a PPC exception, or just good old crashing. so have given up with warp3D for the time being. having too much fun playing oldie games. :-D
otherwise, CGX seems to do everthing i want it to. :-)