Reply 120 of 176, by kool kitty89
Some chipsets (and boards actually connecting all the address lines to RAM sockets) support 4MB SIMMs, some just 1MB, though I think all SIMM and SIPP socketed boards at least support the 1MB. (none is limited to just 256kB module)
So there's definitely some 8-slot SIMM/SIPP boards that max out at 8MB, probably some 386 boards even. (I have an unusual 8-SIMM slot 286 SUNTAC based board that is limited to 8x 1Mx9 modules max and also didn't seem to like having 6x modules in there: just 2, 4, or 8 )
And I haven't run into ISA I/O conflicts causing memory errors like that, but I suppose SCSI DMA operations might do it.
I have run into certain combinations of cards and boards failing to boot (or sometimes failing to POST), but the behavior is different. It's mostly related to compact flash adapters + certain video cards + certain motherboards + aux 5V power supplied to the IDE to CF adapter. (in cases where boards can supply enough power through the ISA slot to run the CF card, the problem is also avoided by just not using external power ... I'm also pretty sure I'm using the correct jumper settings on these cards for that, but I could be screwing up there somewhere, too)
Also, that 12 MHz 287 comment from Evasive was towards one of my posts.
I was also wrong, it was actually an Intel 80C287-10 installed in that 12 MHz 286 board, and I've since seen a couple other 10 MHz 287s installed in 12 MHz boards on ebay (and don't appear to be set-up to use an asynchronous clock, so just running 12 MHz).
So it's even less a wonder it wouldn't work at 20 MHz, but I should qualify that I was thinking of the typically very-overclockable Intel 287XL-12 and assuming it was possible the plain CMOS 287 might also do well in that regard. And on top of that, I was also keeping in mind that Intel didn't bother grading any parts higher than 12 MHz for 286 and 287 to avoid cutting in on 386 and 386SX profit margins. (that's at least true for the later CMOS 286s while the original 12.5 MHz NMOS 286 was around before the 386 was on the market I think and at least well before the 386SX was marketed)
And aside from that, there's not all that much fun that can be had messing around with FPUs on souped-up 286 or XT class systems given how little software supports it. I think most folks on here just have them in there to get more complete benchmark results.
With 386s you can at least have the novelty of trying to run way out of era (mostly Pentium-oriented) benchmarks, demos, and a few games, but you can't do that with the 16-bit x86 processors, even in late model 286 boards with 386SX sockets.
But if you want to play around with some 16-bit CAD software or certain accounting or spreadsheet software that uses or requires an FPU, there's that. (there might actually be some flight simulators ... or maybe just one, that are both 16-bit real mode and can make use of an FPU, but I'm not sure about that)
It's sort of like those really fast Weitek FP coprocessors from the 386 and 486 era. Really good performers, but limited in actual software.
Except, unlike the 287, the Weitek co-pros were not well suited to accounting/statistics/scientific math programs actually needing or using double or extended precision FP functions, but really fast at 32-bit single-precision floating point math for CAD and 3D rendering applications, probably for DSP-type sound processing too. Thus actually would've been more useful for accelerating 3D games, but I don't think was popular or affordable enough to ever attract that market segment ... and I don't think any arcade games or commercial 3D simulator machines exploited those either. (I think it was used on some 3D graphics workstations, but maybe more so for RISC based platforms, like SUN's SPARK workstations)