You can change the CPU as it's socketed - you can use a very thin flathead screwdriver or a IC extractor like https://www.aliexpress.com/item/1005006471677542.html to remove the CPU.

muon wrote on 2026-04-02, 06:27:

I haven't the manual of the mobo

It looks to be similar to https://theretroweb.com/motherboards/s/aquari … com-mb-4d33-50l, but 50B rather than 50L. Post a screen shot of what get's reported when the system posts and it may be possible to know more.

Yes. I would recommend a proper LIF removal tool, to minimize the risk of damaging your board and/or bending pins:
https://vogonswiki.com/index.php/File:Intel_C … emoval_tool.jpg

Those sockets are a pain in the arse but it is a socket

red-ray wrote on 2026-04-02, 08:01:

muon wrote on 2026-04-02, 06:27:

I haven't the manual of the mobo

It looks to be similar to https://theretroweb.com/motherboards/s/aquari … com-mb-4d33-50l, but 50B rather than 50L.

its https://theretroweb.com/motherboards/s/aquari … ystems-umc-486a

As said, this is a socket do the CPU can be removed.
Since it's a DX33 I'd guess your chances are best with a DX6-66. You may also mount a DX50 but you may also need to replace the oscillator if there are no jumpers for the bus speed (I haven't been to Theretroweb for this board to look it up)

The only silkscreen jumper info is for cache size, pretty solid that this is a 5V only board, CPU clock is the 33MHz zip tied oscillator.
A DX2-66 should be a drop-in replacement but watch out for odd versions that are not 5V. If you wanted 100MHz, it would have to be DX4 100 ODPR (overdrive processor REPLACEMENT) or you would need a 3.3V interposer adapter to run a 3.3V DX4. With no additional options, the non-Intel 486 models might work, or might not, can't remember if they NEED additional jumper options, the more exotic ones would also be 3.3V.
Pentium Overdrive? NO, think that needs the extra row of pins on the later socket.
Already adaptered CPU upgrade modules like Evergreen and others, possible but probably expensive

Thank you all so much for your replies. Yes, the motherboard is the one from the theretroweb link.

The reason I would like to removed the CPU and replaced it with another one is because the cache memory isn't working. There isn't much information about the motherboard, but the cache memory jumper settings are printed directly on the board, and I can't get it to work, even after replacing all the 32kx8 ICs, it has for the 256KB cache.

When I say the cache isn't working, I mean that when I use the DOS cachechk utility, the cache timings are the same as the main memory, and even worse with some memory settings. I don't understand this. It only happens with this motherboard. But it's worse because the CPU's internal cache (8K of the 486) is nonexistent. It has the same timings as the rest of the memory. I thought the CPU must be faulty; I can't think of any other explanation.

That's why I wanted to remove the CPU and replace it with another one to see if it makes a difference. If anyone has any other suggestions, please let me know.

disabled L1 and L2 is deffo not broken CPU issue
https://theretroweb.com/motherboards/s/aquari … ystems-umc-486a doesnt have any documentation 🙁 but picture shows those two blocks of jumpers in the corner unpopulated. You might try that, but some of those might be for selecting SX/DX and still wont work. edit: looks like this was already suggested but no info if you tried: Problems with cache in 486 mainboard

ideally we would need to find original manual or someone with confirmed working setup
looks like Commodore !??!?! was using this mobo https://www.zimmers.net/cbmpics/cpcs4.html I remember someone messing with commodore 486 hmm here is one https://imgur.com/gallery/commodore-486-33c-6pg4IJg but uses different mobo 😐
ah it was Epictronics "Commodore 486 Restoration. I found the matching MOBO!" https://www.youtube.com/watch?v=zvCIwzjXi9A and again different mobo

The L1 cache must work, otherwise the bios program would not be able to start from rom. It uses some arcane trickery to store some stuff in L1 cache before the ram is detected, configured and counted.

Therefore, it might just be disabled later on. Matbe try one of those slowdown programs that can enable and disable caches and see what happens.

Couple of alternate board pics - maybe comparisons might give some clue...

red-ray wrote on 2026-04-02, 08:01:

muon wrote on 2026-04-02, 06:27:

I haven't the manual of the mobo

It looks to be similar to https://theretroweb.com/motherboards/s/aquari … com-mb-4d33-50l, but 50B rather than 50L. Post a screen shot of what get's reported when the system posts and it may be possible to know more.

Thank you, Ray.

muon wrote on 2026-04-02, 13:53:

Thank you all so much for your replies. Yes, the motherboard is the one from the theretroweb link.

muon, have you already played with the turbo switch JP18?
You do not need a physical switch for it, you also may try a jumper instead.
Try to use a tool like SpeedSys to see the graphical cache chart.

Put a not too small screwdriver between socket and CPU and try softly to rotate it left / right from all 4 sides.
It should easily go off then. Watch out for the pins before you (re)insert any CPU into that LIF-Socket ("low insertion force").

When you go for a DX2/66 please mention that you also need a heatsink with fan.
As power dissipation grows with faster CPUs I heavily recommend (for 5-Volt-486):
DX/33: no cooling required
DX/40, DX2/50: heatsink required
DX/50, DX2/66 and faster: heatsink with fan required