I did some test I took from ASUS VL/I-486SV2GX4 Rev 1.7 l2 cache problem

The different lines between the chips in same bank and different bank that should be parallel are parallel.
Voltage for all the chips is 5.03v.
Could not test down to the tag chip as that one is in the middle of the board and there is 3 buffers chip by the sram banks...

Will do a more extended test this weekend and also go to a big city to see if I can find sram there, in case i'm lucky..

Fun fact, I noticed that the rtc was not working when the ISA riser was not connected, like the seconds, don't even change when the system is on.
I did do all my sram test without that riser connected, I don't think that could affect the cache but maybe I should test with the riser installed just in case...

I was also able to gain access to the disabled option I mentioned earlier by moving a jumper "OEM bios / Acer bios" back to Acer. only difference is that it show an acer logo and give me access to the read only option, but nothing related to the L2, the cache option was already enabled but maybe it was just for the L1.

Rav wrote on 2023-02-24, 02:11:

I would assume that if I got defect from aliexpress it would crash or do funny thing right, not just not detect any?
I did try with 128K (as in the picture) and the full 256K.

Hello!

Do you have a TL-866 II? It can test SRAM chips, and highly recommend it if those chips are untested. A lot of us use those 128kx8, 32-pin chips we can only get from eBay in our boards that can handle them, and it seems about 10% of those that you get are bad.

Those cache size jumpers don't actually tell the BIOS anything about how much cache is installed. All they do is redistribute address bits between the tag RAM and the cache data lines. On all these old 386 and 486 boards, the BIOS determines the cache size by putting the chipset into a special cache test mode, writing patterns (0x55AA/0xAA55 etc.) into memory, and seeing what sticks. So if a cache chip doesn't respond at all or always returns bad data it's the same as a missing one.

I don't have a TL-866 II (Yet... I did plan to get one)
So from what you say it could just be bad chips, thanks.

I found that https://github.com/johnzl-777/SRAM-Read-Write
It append I did have an Arduino Mega 2560. So I should be able to test the sram chip soon without that tl866.
It won't tell me if there on spec with regard to latency but I will be able to know if it alive or dead.

Problem solved.
I made that arduino sram tester and modified the program so it do full random fill then read it back then AA55ing the whole sram again and read it back again.

Weeded out 4 of the 10 chips, 40% defects!

Now running only with 128K only so I downgraded the ram to 16MB for now so it's fully cached.
Have to order more of that half dead sram!

Exactly why, I had issued many warnings that what you did was from chinese sellers. Not acceptable to have them take money for fakes and defective parts. In my opinion this is a big no no as this encourages them to keeping practicing these cheating on buyers like this.

Cheers,

pentiumspeed wrote on 2023-02-25, 23:31:

Exactly why, I had issued many warnings that what you did was from chinese sellers. Not acceptable to have them take money for fakes and defective parts. In my opinion this is a big no no as this encourages them to keeping practicing these cheating on buyers like this.

Cheers,

Holding the "Honk if scammed on L2 cache" sign.

Yes, it's unbelievable no prior test isn't done before selling.
Go ask a refund...........

After eliminating the dead chips, It's still cheaper than local sram or certified sram unfortunately.
imo, they should just disclose the expected failure rate to the buyer so the buyer would be aware that he need to buy more and to test them first.
They will continue this practice until more shops offer sram at there price...

4$ for 6 chips that work fine on a 40mhz bus is fine.