In general though, Isn't a Pentium II or III what you actually want?
Have you recently worked for a while with a 486 system? Got a feel of what such a CPU can do, and more importantly, what it cannot do.

For example:

My experience using a Cyrix 5x86 100MHz
Windows 95 with any MP3 player: 128kB Stereo MP3s need downsampling or quality reduction to play reliably.
MPXplay in pure DOS: Can play 128kB Stereo MP3s without problems, 160kB too.

Tymo486DX2 wrote on 2024-09-18, 12:49:

Socket 7 for 486 DX2 4x 72-pin EDO-RAM 1x AGPx2 at 66 MHz 2x PCI at 33 MHz 2x ISA 1x VESA VLB cache memory 2 MB integrated Mult […]
Show full quote

Socket 7 for 486 DX2
4x 72-pin EDO-RAM
1x AGPx2 at 66 MHz
2x PCI at 33 MHz
2x ISA
1x VESA VLB
cache memory 2 MB
integrated Multi I/O controller with 2xIDE
FSB 33/66 MHz dynamic

Hmm there are no socket 7 486 chips and there are no adapters either or BIOS support ..I mean you want a Unicorn...I want 10 million USD . .but neither of us is going to get our Unicron.

I have a better chance than you since my Unicorn does exist.

Something went wrong in case Super Socket 7.
I mean PGA168 for 486 DX-2
link

Generally all Host Bus at 66 MHz and divider for PCI. No need to dynamic switiching.
Ultra ATA-66 could be installed.

Tymo486DX2 wrote on 2024-09-18, 13:36:

Something went wrong in case Super Socket 7. I mean PGA168 for 486 DX-2 link […]
Show full quote

Something went wrong in case Super Socket 7.
I mean PGA168 for 486 DX-2
link

Generally all Host Bus at 66 MHz and divider for PCI. No need to dynamic switiching.
Ultra ATA-66 could be installed.

VLB needs 33Mhz host bus .. 40Mhz is possible but very few boards could overclock the VLB that high and maintain stability and overclocking VLB does cause compatibility issues with cards as not all work reliably beyond 33Mhz.

Now if you ditch VLB then yeah you could run a 66Mhz host bus since PCI/AGP does support that.

New few cards with VLB at 66 Mhz would be possible at that case of new MB.
Only for fans.

Tymo486DX2 wrote on 2024-09-18, 13:53:

New few cards with VLB at 66 Mhz would be possible at that case of new MB.
Only for fans.

Not sure you understand how VLB works .. PCI was developed to replace VLB because of its limitations in being directly linked to the CPU bus.

I guess you could develop a chip that acts as an intermediary between VLB and the CPU to maintain 33Mhz, PCI/AGP both would run at what ever speed the chipset tells them to and ISA is locked to 4.77/8.33/12Mhz regardless so we can ignore that.

I mean even if this is a fantasy scenario you are still limited to not breaking the laws of thermodynamics.

Long time ago I have read lot of about VESA VLB. Of course there are technical issues around it.
As You say VLB should be 33 MHz.
If processor could be attached to the Host Bus 66 MHz even with own fixed divider in socket for 33 MHz?
Then VLB with processor could work at the same divider [33 MHz] from Host Bus 66 MHz special for AGPx2 and HDD ultra ATA-66 controller.

Old VLB cards reviews:
VLB controllers

I suggested that above with a separate controller handling the VLB bus speed.

I'm not sure what you want VLB for if you have onboard ATA66 are using AGP video and onboard LAN .. at this point there is nothing you need VLB for. You also cant run a VLB IDE controller if you have it onboard, one of them will be ignored by the BIOS.

So it may be better to drop VLB and make PCI/AGP run at multiples of CPU FSB, why stop at 66Mhz, you can get 486 CPUs that can run at 133Mhz and higher with the right board/apdaptor.

66 MHz is on target for fans only not for performance of course and it is strange requirements. I agree.
Yes, we can fully abandon the VLB port.

But...
All PC designs can be temporary, and MB is trying to be universal for everyone. It is hard to say about preferences, especially temporary users.

Desolder the CPU off a covington celeron slot 1 PCB and bodge wire an AMD DX5 on it 🤣

If somebody interested with this project?

Kickstarter project can be started.

I don't mean to be a downer, but what you want doesn't really make sense. There aren't any 486 chipsets that can fully take advantage of PCI, let alone AGP. While crazier projects have become reality lately, I don't see how trying to convince others to do it for you is going to make it happen.

If you want 486-like experience with AGP, just pair WinChip CPU with Super Socket 7 board.

Or build something like the ITX-Llama.

Tymo486DX2 wrote on 2024-09-18, 17:04:

If somebody interested with this project?

Kickstarter project can be started.

With respect, as far as I can tell, you are the only one here so far who sees an interest in a 486 motherboard with an AGP slot so far.

I am sincerely trying to understand what are the advantages, even theoretical, that you believe that an AGP slot would bring to a 486 based machine. I suspect that I am not the only one trying to understand.

Is there anything that a theoretical 486 CPU based machine with AGP could do that an existing AGP equipped Pentium II or Super socket 7 machine could not do as well or better ?

With a possible exception for personal "can it be done" type purely academic challenge, I do not see why someone would want to start such an endeavor because I just cannot imagine a use case for it. As for purely academic type "can it be done" personal challenge projects, they tend to require a large time and resource (money) investment and often yield a single prototype (if even that) and maybe some documentation that may be sufficient for someone with the right skillset to replicate the achievement.

In other words, if you don't do it yourself and nobody else sees any compelling reasons why it should be done, it will not happen.

Please feel free to correct me if I you disagree or if I have somehow misunderstood.

EDIT: Corrected typos

BitWrangler wrote on 2024-09-18, 16:42:

Desolder the CPU off a covington celeron slot 1 PCB and bodge wire an AMD DX5 on it 🤣

I would never do that to a AMD DX5 ....such hersey

THE EMPEROR WILL HEAR OF THIS !

This is an Ian Malcolm question if I've ever heard one.

"Your Vogoners were so obsessed with whether or not that they COULD, that they never stopped to think if they SHOULD!

as someone who went through the 486 era as it evolved and itched for AGP boards after all that.... what the hell??

When AGP graphics cards started to really happen by 1998, 486DX2 computers were OBSOLETE and couldn't run anything those AGP cards would be thinking of (i.e. incoming, quake2 crusher.dm2, unreal, etc.). The proposed AGP texturing wasn't even implemented by everyone (3dfx agp cards famously didn't) and wouldn't have benefit a 486 either. It just means a faster bus to get bigger buffers through quicker (triple buffers and high resolutions), it won't mean more complex FPU load off the table.

It was barely even worth choosing period AGP graphics boards over their PCI versions before about 1999-2000.

CPU scaling is a steep slope, like this, https://www.anandtech.com/show/288/18, so a quarter of the bottom score, then half that again, because it's 4th gen not 6th, you don't need Fraps or anything for that, Sesame Street's "The Count" can count that fast, von-two-sree, von-two-sree, von-two-sree, sree frames, sree frames per second ah ah ah.

It's worth choosing AGP cards that had 128-bit memory vs, PCI 64-bit memory.

With a low bus speed you would run low texture details anyway, but higher resolutions would be possible, or more 3D hardware calculations per pixel.