Finally found it. Some history….the tool was created by a German magazine. Tom Pabst helped translate it for them. Tom is the guy who founded Toms Hardware (I didn’t know who surname was Pabst). The tool was made freely available for everyone to use.

The tool was originally located here http://ourworld.compuserve.com/homepages/toms_r/ctcm.htm then moved to http://www.personal.u-net.com/~sysdoc/hardware.htm then to http://sysdoc.pair.com/ afterwards it made its way to tomshardware.com.

Here is its final resting place before it disappeared into the World Wide Web ether.

https://web.archive.org/web/19980119085255/ht … e.com/ctcm.html

vswitchzero uses it in this video https://youtu.be/1M_qp5CE1DI?si=2CTREwA6JV4pcpWd

Oh yeah, I recall that thing from back in the era...
Boards with fake cache were common back then, so when I got to know about that utility - from the March 28, 1995 issue of "Komputery i Biuro" magazine - I hastily went to download it.
And it was well worth it - it indeed happened that my new board turned out to have only 64 KB of cache, instead of the advertised 256 KB.

But now I'm questioning infallibility of that tool...
See the pic in the original post: "Write Strategy L2: Write Back", and "Dirty Tag L2: n/a" - but it isn't possible to have Write Back without the Tag, right?

And that's my 486DX2/66:

Grzyb wrote on 2024-08-25, 00:36:

But now I'm questioning infallibility of that tool...
See the pic in the original post: "Write Strategy L2: Write Back", and "Dirty Tag L2: n/a" - but it isn't possible to have Write Back without the Tag, right?

I think it is possible. The memory is just written again even though it isn't needed. Highly inefficient. My understanding of it may be wrong though.

I have a 486DX2/66 on a DataExpert EXP3406 rev 1.1 (https://theretroweb.com/motherboards/s/dataex … ert-exp3406-1.1). As per the motherboards schematics, there are 2x cache tag sockets.

But only 1 has a cache chip in it. So a dirty tag cannot be used. vswitchzero explains it nicely in his video (see link in previous post).

My motherboards L2 cache is set to WB in the BIOS. In the CTCM cache tool you can confirm the dirty tag is not being used and L2 is in WB mode.

The motherboard has terrible performance. I don't know why companies back in the day selling these motherboards just didn't spend a little more and add an extra cache chip for the dirty tag.

I've ordered another IS61C256A-15N DIP-28 chip for my motherboard (https://www.aliexpress.com/item/1005004418864759.html). Performance should massively increase when installed.

My results:

The difference in "Memory Bandwidth" is shocking: 16 vs. 75 MB/s !

Grzyb wrote on 2024-08-25, 02:38:

The difference in "Memory Bandwidth" is shocking: 16 vs. 75 MB/s !

Yep 100%. I need that dirty tag cache chip badly.

I don't think you can blame it all on cache.
See my results with both caches disabled:

Grzyb wrote on 2024-08-25, 03:59:

I don't think you can blame it all on cache.

This is true but I reckon a good portion of it will be cache. Hopefully the new cache won't take too long to arrive.