Hi guys, assuming you could order a custom, modern-made 486 motherboard, what specs would you request?
getting rid of most old standards. It doesn't matter how good that little experiment would be if the ecosystem to run it with isn't there
• Socket 2 and 3 compatible of course, compatible also with Pentium Overdrive
goes with the territory.
• Replaceable real 5ns L2 cache, min 512 KB, max 2048 KB (with modern manufacturing technologies I guess 5ns would be possible), fully stable @ 66 MHz with zero wait states
just put a few MB directly on die on the chipset and avoid the issue alltogether
• 8x 72pin DIMM RAM slots (with custom ad-hoc 40ns FPM RAM modules also purchaseable, fully stable @ 66 MHz with zero wait states)
Since you need a memory controller either way, just integrate a DDR2/3/4 controller and have easy access to cheap modern memory. Chances are that would be fast enough for cache not to matter at all.
• Fully stable FSB @ 66 MHz
sure, why not. The rest of the system would run asynchronous anyway because we can design it in a sane way to avoid stability issues.
• CPU multiplier ON/OFF jumper
sure, why not. BUT not all 486 even have selectable multipliers. Especially many DX/2 are hardwired.
• Fully settable FSB: 16, 20, 25, 33, 40, 50, 60, and 66 MHz
sure why not.
• 66 MHz PCI 2.1 (backwards compatible) x4 slots
no point, there is barely any hardware that supports that anyway.
• Fully settable PCI slots via jumper: 16, 20, 25, 33, 40, 50, 60, and 66 MHz
no. If we freshly design it we are keeping it sane and within specs. fixed 33Mhz or 66Mhz according to specs. Stability is key here.
• 3 EISA slots (ISA compatible), that, besides the default 8.3 MHz setting, can be forced to run also @ 12 MHz and 16 MHz
can do, but why? PCI already has anything you will ever need with the exception of soundcards and for that you only need ISA with chipset support for proper DMA/IRQ handling. EISA does nothing worthwhile on top of that that wouldn't be handled better by PCI cards.
• 32 bit 66 MHz local-bus built-in caching Controller with 2x IDE PORTS, EIDE (PIO Modes up to 4) and UATA (up to UATA 133) compatible, with zero to 8 MB RAM using the same DIMMs as above for the main System RAM
why? simple SATA ports will be easier to deal with and assuming you use an SSD the caching would do jack shit anyway. No point in going the IDE route. If for some godforsaken reason you absolutely HAVE to use an IDE device, just get an IDE to SATA adapter and be done with it.
• 32 bit 66 MHz local-bus built-in U2W SCSI Caching Controller (80 MB/sec) with two ports, with zero to 8 MB RAM using the same DIMMs as above for the main System RAM
that one makes slightly more sense, but only as an external port. There was a lot of peripheral gear that used SCSI. but for internal use what is the point? SATA will be faster and you can actually get new hardware for that. And if you insist on having the modern SCSI equivalent implement a SAS controler instead of SATA. It will be more expensive, but SATA devices will also work on it so outside of cost it wouldn't even be much of a downside. But since it is for a 486 the question again would be why.
EDIT: • Equipped with also 2xIDE-to-SD Card and 2xSCSI-to-CF Card converters, full speed (able to use up the entire UATA for the first and U2W SCSI for the second bandwidths)
sure could do, but why? You can already have SATA so not much of a point.
EDIT: • Dual Floppy Disk Drive port, up to 2x3.5" FDD and 1x5.25" FDD
missing a trick here, built in FDD emulator would be a good idea. but generally, it makes sense
• 32 bit 66 MHz local-bus built-in Cirrus Logic GD5434 SVGA with 2 MB of I have no idea what kind of memory would be better (I guess 45 ns DRAM, right?), disableable via jumper if one wants to use different SVGA cards in the PCI slots
good luck even getting such old video chips. How about instead we get some more modern integrated GPU instead? just make sure that the bios has proper VESA modes implemented. Also has the advantage that it could have HDMI output natively. Ideally with scaling, but that might be iffy.
-dedicated USB ports for mouse and keyboard that convert to PS/2 or with a bios implementation that has proper "legacy support" for them. It is getting harder to find good PS/2 mice and keyboards.
-PCIe x16 slot for a video card. The actual lane number doesn't matter just to have the option to use a video card with modern outputs (HDMI, Displayport) to avoid analogue to digital converter boxes
Most options should be configurable over the bios including some additional ones like actually available memory in case there are issues with too much memory. Various forms of "emulation" and legacy support in general, basically software defining most aspects of the system.
Maybe get the bios to support Nvme in a way that dos sees Nvme drives as regular fixed disks, not sure about the feasibility here.
I'm sure I can come up with more but that is enough for now.
Question for the most skilled/experienced: do you think such a beauty beast would be manufacturable today with modern technologies and current infrastructures? Or would manufacturing companies have to re-invest lots of money to buy totally different kinds of equipment to build the above or similar to the above?
could it be done? yes. However designing the hardware and manufacturing the components would be prohibitively expensive if you break it down to the few maniacs that would actually want to buy such a thing. It would just be not worth it. I know I wouldn't want to buy such a device at the prices you would have to ask for. It would make ebay prices look like the bargain of a century. A more sane approach would be to go the FPGA route like what is being done for old consoles (Retro AVS, Analogue NT mini, SuperNT, MegaSG) or homecomputers like Atari and Amiga (MiSTer) and completely emulate the entire hardware that way.
Making new 486 boards is an idea that pops every few months and it is just an unrealistic pipe dream. Many of the components needed have been out of production for decades and designing and manufacturing new chipsets is not going to happen for the lack of a market. And don't even bring up the argument that people already make new soundcards and midi interfaces. Those projects are orders of magnitudes simpler both on the PCB front as well as the ability to source parts since most cards use simple standard components anyway. A motherboard is a completely different beast.