From what I know, fake chips on some 486 boards were just dummy chips: a DIP package that had no function inside at all. They were there to trick people to think there was L2 on the board.

The other kind of fake chip is where the reseller puts a new coating on the chip with a new print. It's often a cheap unrelated part that they try to sell as a more popular chip for profit. Your chip might not work at all, in such a case. Or, it is a cache chip but of a lower capacity than was advertised.

Some BIOS could report the cache on boot, even tho you have none installed. It then just echoes whatever you set in BIOS.

Speedsys should be a quite reliable tool to verify your memory.

Hezus wrote on 2023-04-13, 21:14:

From what I know, fake chips on some 486 boards were just dummy chips: a DIP package that had no function inside at all. They were there to trick people to think there was L2 on the board.

Yep, like on my PCChips M919.

And the traces on the back, which are hilarious:

Fortunately I own a cache module for it, so mine actually has cache.

So here my speedsys membench. It says i have 1MB, cachechk say i have nothing.
That means cachechk cant read out my hardware correctly?

Jackhead wrote on 2023-04-13, 21:34:

So here my speedsys membench. It says i have 1MB, cachechk say i have nothing.
That means cachechk cant read out my hardware correctly?

IMG_4153.JPG

Hi.

Sorry, but 🤣.
You use the Asus SV2GX4 board, that board does not report wrong cache sizes.
So when the boards shows 1MB L2, you have 1 MB L2.
So cachechk is just wrong. speedsys clearly shows that it is real 1MB L2 cache, too.

Remark:
You seem to have slightly mis-configured somethine, because your values are not on the highest level.

See here:
Re: Avance Logic ALG2228 VLB

EDIT: To be more precise, you L2 cache is NOT running on 2-1-1-1 read and 2T write. Do you have 2-2-2-2 read?

Jackhead wrote on 2023-04-13, 21:34:

So here my speedsys membench. It says i have 1MB, cachechk say i have nothing.
That means cachechk cant read out my hardware correctly?

Your RAM is too fast (or your L2 is too slow, depends on the point of view), so the speed difference between cached reads and non-cached reads is too low for cachechk to pick it up. Try adding RAM wait states in the setup, and see whether cachechk then recognizes the correct amount of cache.

Thanks guys! Here my settings, maybe you see what i can improve here.
So I need go up WS, or cache w/r. I have 15ns cache . And Rams are 60ns.

Jackhead wrote on 2023-04-13, 21:51:

Thanks guys! Here my settings, maybe you see what i can improve here.
So I need go up WS, or cache w/r. I have 15ns cache . And Rams are 60ns.
IMG_4154.JPG

Cache Burst Read should be 1T!

Also change Slow Refresh to Enabled. Slowly refreshing the DRAM means a faster performance. But minor improvement.
Cache Burst Read is important.

EDIT:
Interesting would be also setting local bus values to T2 and Transparent.
You have to check, if the computer is still stable then. No guarantee 😁
But you could beat my ARK 1000VL 27000kB/s value in speedsys. Maybe.

CoffeeOne wrote on 2023-04-13, 21:56:

Cache Burst Read should be 1T! […]
Show full quote

Jackhead wrote on 2023-04-13, 21:51:

Thanks guys! Here my settings, maybe you see what i can improve here.
So I need go up WS, or cache w/r. I have 15ns cache . And Rams are 60ns.
IMG_4154.JPG

Cache Burst Read should be 1T!

Also change Slow Refresh to Enabled. Slowly refreshing the DRAM means a faster performance. But minor improvement.
Cache Burst Read is important.

Here we go:

Jackhead wrote on 2023-04-13, 22:01:

CoffeeOne wrote on 2023-04-13, 21:56:

Cache Burst Read should be 1T! […]
Show full quote

Jackhead wrote on 2023-04-13, 21:51:

Thanks guys! Here my settings, maybe you see what i can improve here.
So I need go up WS, or cache w/r. I have 15ns cache . And Rams are 60ns.
IMG_4154.JPG

Cache Burst Read should be 1T!

Also change Slow Refresh to Enabled. Slowly refreshing the DRAM means a faster performance. But minor improvement.
Cache Burst Read is important.

Here we go:
IMG_4155.JPG

Great.
Now try T2 and Transparent. Graphics speed would be higher. Keep in mind, that you have already good graphics speed.
.....

When i set Transparent system dont boot. Hold after memcheck. With T2 only i see no difference on bandwidth

Jackhead wrote on 2023-04-13, 22:11:

When i set Transparent system dont boot. Hold after memcheck. With T2 only i see no difference on bandwidth

Yay, then keep T3 and Synchronize.
Still the ET4000W32p is pretty fast.
You could post values from Phil's Benchmark Suite:
Quake, Doom (fast PCs) and others.
Your PC is now competitive 😉

For now im pretty happy that "fake" nonsense is cleared.
Btw i ordered the chips here from amazon, pretty cheap. Just tryed it and looks it was worth 😀 . Maybe this help someone else.
Thanks everyone!
https://www.amazon.de/IS61C1024-15N-IS61C1024 … 81424722&sr=8-1

CoffeeOne wrote on 2023-04-13, 21:40:

You use the Asus SV2GX4 board, that board does not report wrong cache sizes.
So when the boards shows 1MB L2, you have 1 MB L2.
So cachechk is just wrong. speedsys clearly shows that it is real 1MB L2 cache, too.

Do I see it right that read speed goes from 15MB/s uncached (386 boards do better) to ~30MB/s? How does it look if cache is disabled completely?

rasz_pl wrote on 2023-04-14, 02:49:

CoffeeOne wrote on 2023-04-13, 21:40:

You use the Asus SV2GX4 board, that board does not report wrong cache sizes.
So when the boards shows 1MB L2, you have 1 MB L2.
So cachechk is just wrong. speedsys clearly shows that it is real 1MB L2 cache, too.

Do I see it right that read speed goes from 15MB/s uncached (386 boards do better) to ~30MB/s? How does it look if cache is disabled completely?

No, you don't see it right.

CoffeeOne wrote on 2023-04-14, 04:47:

rasz_pl wrote on 2023-04-14, 02:49:

CoffeeOne wrote on 2023-04-13, 21:40:

You use the Asus SV2GX4 board, that board does not report wrong cache sizes.
So when the boards shows 1MB L2, you have 1 MB L2.
So cachechk is just wrong. speedsys clearly shows that it is real 1MB L2 cache, too.

Do I see it right that read speed goes from 15MB/s uncached (386 boards do better) to ~30MB/s? How does it look if cache is disabled completely?

No, you don't see it right.

Yes, I was looking at moving 🙁 . Its ~45 to ~67, weird that cachechk doesnt pick it up

rasz_pl wrote on 2023-04-14, 10:53:

Yes, I was looking at moving 🙁 . Its ~45 to ~67, weird that cachechk doesnt pick it up

When the OP stated that cachechk doesn't pick it up, L2 timings were bad (2-2-2-2 instead of 2-1-1-1), and it went from 45 to 53, which is a smaller difference than what you see now.