Swap the existing chips, two at once. If the memory memory moves, one of the two chips is bad. If nothing changes, both chips are good. In the end, you know which chips are certainly good and by swapping them with the two from one must be bad, you can identify the single bad chip.

derSammler wrote:

Swap the existing chips, two at once. If the memory memory moves, one of the two chips is bad. If nothing changes, both chips are good. In the end, you know which chips are certainly good and by swapping them with the two from one must be bad, you can identify the single bad chip.

For simplicity, I reduced the system cache size to 128KB using jumpers, and used the same socket, going through each chip. This error did not appear in the 128KB mode, on any of the chips.

This error will return when all chips are installed and the jumpers have been set correctly to accept 256KB of L2 cache. Does this mean that there may be a problem with the motherboard?

It can also mean the bad cache chip may be good in the first 128KB

Deksor wrote:

It can also mean the bad cache chip may be good in the first 128KB

How would a defective module work in one configuration but not another? Wouldn't there be errors? This really doesn't make much sense to me now, and I say that as if it made a huge amount of sense to me in the first place. 🤣

First, a TAG module, which does make sense, but now, a chip that works one way but not another?

I may work on this later on. For now, I'm happy with the performance I'm getting out of 128KB of RAM. It's definitely far better than no cache at all. I am certainly thankful for the help, but I've worked on this cache problem for a little while now. The last thing I want to do right now is enable the 256k and switch out every single chip on the board, due to the long bootup process this machine has. No matter what, it will sit at the POST screen for a good 20 seconds, even with minimal POST testing enabled.

This happens with/without cache, so it seems to just be a quirk of the machine.

Could be that the TAG RAM is faulty or not suitable for 256kb cache. Since you don't need all sockets for 128kb, it's also possible that one of the additional sockets for 256kb is damaged. But it's hard to say if you don't have spare cache chips and no other way to test them.

derSammler wrote:

Could be that the TAG RAM is faulty or not suitable for 256kb cache. Since you don't need all sockets for 128kb, it's also possible that one of the additional sockets for 256kb is damaged. But it's hard to say if you don't have spare cache chips and no other way to test them.

I do have 4 other 20NS cache chips, though they're not identical to the 9 ones I got. I figured this would mean that I could not use that kind of chip alongside the other chips. The 4 other chips are of a Samsung make and model, while the 9 chips I've been trying to use are ISSI chips.

Unfortunately, this is the only motherboard I have that takes 28-pin cache, so I have no way of testing it in any other boards.

I've seen boards with different branded cache chips. I've seen even more with the TAG chip being a different brand than the cache chips.

I would go ahead and try one of the Samsung chips. It will either work or it won't.

For that matter you could replace every other one with the Samsung chips and I would think it should work no problem.

cyclone3d wrote:

I've seen boards with different branded cache chips. I've seen even more with the TAG chip being a different brand than the cache chips.

I would go ahead and try one of the Samsung chips. It will either work or it won't.

For that matter you could replace every other one with the Samsung chips and I would think it should work no problem.

I changed out some of the top row of chips with the Samsung chips, and HIMEM.SYS no longer reports errors when the system is set up with 256KB of L2 cache, so I'm pretty pleased with those results. I ran CACHECHK, and while I'm not sure how to read the results, this is what it shows when it gets done running:

TL866-II has built in SRAM test which can be used to test 386/486 cache SRAMs.
And it's a necessary tool for every retro computing/retro console fan anyway.

Neat ! I didn't even knew that !
I knew it could test some ram chip, but I didn't know it could test THESE chips.