I've got something interesting for you.

Currently I've collected about dozen of different Socket 3 CPUs.
Also, I recently purchased an IBM PC 330-6571, PC 350-6581 motherboard from dumpster, exactly like pictured here: https://theretroweb.com/motherboards/s/ibm-pc … 71,-pc-350-6581
It has a CPU covered with a giant heatsink with no useable markings. Even the IBM FRU 96G3013 is unusual for this system, at least I was unable to find the matching FRU in well-known places. I've even have seen the same issue somewhere here at Vogons, which ended up in nothing so, instead of blindly putting this CPU into socket I tried to perform some research and analyze my other processors with my UT61E digital multimeter and you know what? I've found something really cool!

Here's the reference Socket3 pinout:
http://www.pchardwarelinks.com/486pin.htm

And here's the huge pin assignment table made by a really nice person:
http://www.pchardwarelinks.com/486pin2.htm

Since different processors have different functions on the same pins, sometimes those pins behave differently in analog domain and these differences may be measured as resistances, capacitances and PN-junction voltage drops (e.g. forward voltage), so it may be useful in identifying 486 CPU with unreachable top markings.

Here's what I've found out so far:
https://mega.nz/file/MbVSkZCA#tr0yKFxJVvNb7yr … DdRSjgIaGuwjUmc

Attached copy:

This method helped my to identify the covered unit as AMD 3 volt Dx4 with Write-back cache enable with very high degree of confidence. and it's confirmed to work at 100MHz and 120 MHz. I think it's 100 MHz but I need more statistics to say for sure. Meanwhile I will try to test it with different testing software.

My main method is measuring diode voltage drops between Vss and different pins. The explanation is simple: Digital chips are still made from analog parts. Usually digital chips have protection diode clamps and properties of those diodes (or transistors) may vary in different series. Also, different functioning modes (input, output, I/O, supply, etc) use different schematic and that also expresses in different behavior. Also, if chips are designed for different or multiple voltages, different power or have different number of gates, that will result in different analog parameters as well. The following table is my initial set of examined pins. The first always goes to the positive probe, and second goes to negative. All measurements are performed at the room temperature about 25 degrees C, but of course, much better and stable results could be done with measurement at 60 degrees using a heater with thermal probe and feedback. Unfortunately I don't have such kind of a heater but anyway all units have non-zero tolarance, and I expect variety even between same models made in different countries and in different years.

Anyway, here's the list:
Vforward A7-B10
Vforward A7-B12
Vforward A7-B13
Vforward A7-R17
Vforward A7-A3
Vforward A3-A7
Vforward A7-B7
Vforward A7-C4
Vforward C4-B7
Vforward C4-G15
Vforward H16-J1
Resistance B7-A7
Resistance C4-S4

You may see from the table that each CPU has it's own set of parameters but identical ones have their parameters nearly the same, which proves the concept that having a set of measured parameters you can narrow down your guess to a small set of couple of CPUs using such table.

If some collector having a large set of various Socket 3 CPUs and a decent DMM could perform this test on other models using this method and write down their results, that may be a very nice addition in statistics which probably may end up in a decent algorithm allowing to identify any non-CPUID processor only using a set of measurements, which probably could be useful at flea markets, identifying fakes, guessing limits, etc.

I think, there may be additional way of research using power supply but that requires a several watt PSU with adjustable output current and stable adjustable voltage so this cannot be done on the go.

Cheers.