Think about it. To make fakes like these you need a mold for the plastic packages, and not just any cheap one - and industrial mold. It's not a 10 or 100 pieces to be made. Then you need the pins, each has to be stamped from a metal sheet with, again, a custom tool. Then everything has to be put together precisely so it actually fits in the PCB. Not to mention any fake markings to make the "chip" look like a real one.

In other words, someone had to design, make and run a whole production line for these fakes, source the materials and pay for labour, and ALL that was still cheaper than buying mass-manufactured parts. Tells you a lot about how cheap production in China can be, huh. Though obviously someone pocketed the price difference and is now a millionaire, always on a lookout to make more money in similar ways.

Didn't some of those boards also have the traces under the cache chips just running around in circles? It would be nice if there was a way to get the cache on a junk M537 here working, there is an option for enabling it in the BIOS but it doesn't actually do anything (even if it says there's 512kb)...

Deunan wrote:

In other words, someone had to design, make and run a whole production line for these fakes

Maybe they already had a production line for making real chips and just ran a batch of fake ones using cheap plastic.

I've "de-faked" two 486 boards so far (Azza 4SGS-3VL and TK8498F) and they recognized the cache with their original BIOSes, so you might be lucky too. The TK8498 had some fakes that were pretty hard to tell, very well made, proper epoxy package, nice markings (but no known/existing brand), but the Azza had obviously fakes/soft plastic like yours, and was also missing two 74F244 buffers that I had to install for the cache to work.

Nprod wrote:

Maybe they already had a production line for making real chips and just ran a batch of fake ones using cheap plastic.

It's not so easy to just change the material for injection molds. There's always going to be some mixing with leftovers, not to mention any stops to make this change are major PITA. There will be a warmup period after a shutdown, some very early mold outputs are basically thrown away.
So that's my point here, if one were to re-use an existing chip line it would be trivial to just do everything like normal but not put the actual bonded silicon in the mold but rather clean metal cutouts, made from same material. Make "empty" chips in other words - I mean how expensive can that be?
But no, the pins on the fake don't actually go into the chip so that's a penny of metal saved, and the plastic is a different, cheaper kind, another penny. But for these pennies to even matter that had to be a massive operation that actually paid for the custom tooling, materials and machines in general.

BCH wrote:

Hello,

I got one of these motherboards with fake cache chips so I decided to desolder them and put some sockets with real ram (and possibly change the BIOS to be able to actually enable the cache). I know the m912 came also with real cache, so Im looking for its BIOS file to use it on mine.

Check the 486 BIOS thread, there is a dump from an M912 with working L2 ("486 Deep Green")

BIOSes on these boards actually support cache, they just lie when they display it's present, but other than this it works when real cache is installed.

oh well, I think I managed to kill the board as now it won't post.

I was very careful while desoldering the chips, and I cannot see any damaged tracks but still, the board is such a bad quality (probably the worst I've ever seen; it flexes like a cardboard) that perhaps a via got disconnected or something Im not able to see.

Are these 4 layer boards? They surely looks like 2 layers to me.

I will still try it with the cache chips once I got them, and also see if I can troubleshoot with a diagnostic card...

A photo or two would help. If the cache was connected (just had fake chips) then perhaps you damaged a via. You don't actually have to rip one out, some poor quality mobos can have the via crack near one of the layers. Repair is possible on a 2-layer PCB, you just solder a wire (or a compononent pin) through and make a solder joint on both sides. On 4-layers it might still work sometimes in the same scenario, as you fill the hole, solder can "catch" copper in the inner layer and restore the connection. But thats more down to luck at that point.

Maybe it'll work again if you install sockets in there ?

Here are some pictures of the areas I worked on (cache ram and bios chip):

I will solder some sockets to see if it helps.

Wait, before you solder anything there. That's going to make tracing any connection a major PITA though at least now you have these photos.
So it would seem that the cache chips and memory sockets are connected via the same signals. I don't like this setup, I'm having hard time figuring out how it works. But the point here is - you have to check each of these lines for continuity, from chipset to SIMM socket. The crack I mentioned could be not in the component pin hole but on one of these smaller vias. Due to temperature change or PCB flexing.

That's a lot of work but trust me, it's still less work then having to desolder the sockets you've just installed once you find that didn't help, and then doing the check anyway. I might be old fashioned but I've learned my lessons on rushing repair jobs. In order to repair stuff one must first know exactly what is wrong with it. Otherwise you are just wasting your time and effort.

Deunan wrote:

Wait, before you solder anything there. That's going to make tracing any connection a major PITA though at least now you have these photos.
So it would seem that the cache chips and memory sockets are connected via the same signals. I don't like this setup, I'm having hard time figuring out how it works. But the point here is - you have to check each of these lines for continuity, from chipset to SIMM socket. The crack I mentioned could be not in the component pin hole but on one of these smaller vias. Due to temperature change or PCB flexing.

That's a lot of work but trust me, it's still less work then having to desolder the sockets you've just installed once you find that didn't help, and then doing the check anyway. I might be old fashioned but I've learned my lessons on rushing repair jobs. In order to repair stuff one must first know exactly what is wrong with it. Otherwise you are just wasting your time and effort.

Yes, you are right. Im going to check the board for continuity first. I just have to snag some free time to do it

I've had a closer look at it and I can't see any obvious problems either. But at least it's making more sense now. There's two cache banks and one tag chip. Each bank is 4x 8bits.

Data lines: Those should be connectected across the banks, or in other words: U24 with U31, U23 with U30, etc. All 32 bits should then go the SIMM socket on one end, and the mobo chipset on the other. I'd start with the 72-pin SIMM socket pinout and test each data bit connection to both banks and then find the pin on the chip.

Address lines: Shared by all chips in a bank, then most likely going to the SIMM socket as well - but the actual mapping might be confusing. It would be best to write that down, might help find out if there is a connection broken in one of the banks.

Hello!
I need your help identifying the components at several locations to enable cache,

near the cache chips locations C26/27/29/65/37/36/64/46; between ISA slots J7 and J8, C16.

I'm going to cut out the fake chips instead of desoldering after reading what happened.
Has anything changed, did you manage to make it work?

BCH. that is one hell of a clean job

how did you do it? did you use hot iron or you have soldering station with hot air, or you have expensive desoldering tool that costs 300 dollar?

i did the job long time ago with a sis chip motherbpard and results were awesome, i dod t need a bios mod, cachechk recognized the memory.

the reason im asking is because i have soldering station with hot air but i never use it for desoldering with hot air. i have seen videos tho that people use hot air to take out components like capacitors

SETBLASTER wrote on 2020-03-25, 01:36:

BCH. that is one hell of a clean job […]
Show full quote

BCH. that is one hell of a clean job

how did you do it? did you use hot iron or you have soldering station with hot air, or you have expensive desoldering tool that costs 300 dollar?

i did the job long time ago with a sis chip motherbpard and results were awesome, i dod t need a bios mod, cachechk recognized the memory.

the reason im asking is because i have soldering station with hot air but i never use it for desoldering with hot air. i have seen videos tho that people use hot air to take out components like capacitors

I think flux and copper braid + some suction tool.
Even if you use hot air it doesn't give that result right off.

Yeah, the PCChips m912 is pretty good for desoldering things 😀

If you don't have a desoldering iron, the simplest way is to cut half of the pins with a pair of cutting pliers then to break the other half using metal fatigue. Then you can easily remove all the pins.