Nexxen wrote on 2020-10-13, 22:10:

I installed 256kb of cache in 9 chips. […]
Show full quote

I installed 256kb of cache in 9 chips.

If disabled message at post doesn't appear, if set to enable it does.
Chkcpu states there's no cache when enabled.

Tomorrow I'll check how to test if there's cache with another program. Hope it's not a bios limit because of fake cache.

Did you flash the BIOS with a proper one ? The Fake Cache BIOS does not work on the M912 with real cache, you need a real cache BIOS.
I use CACHECHK v7 from a DOS floppy, it works very well on 486 boards. Google: cachechk v7 and get it from www sac sk

Horun wrote on 2020-10-13, 22:40:

Nexxen wrote on 2020-10-13, 22:10:

I installed 256kb of cache in 9 chips. […]
Show full quote

I installed 256kb of cache in 9 chips.

If disabled message at post doesn't appear, if set to enable it does.
Chkcpu states there's no cache when enabled.

Tomorrow I'll check how to test if there's cache with another program. Hope it's not a bios limit because of fake cache.

Did you flash the BIOS with a proper one ? The Fake Cache BIOS does not work on the M912 with real cache, you need a real cache BIOS.
I use CACHECHK v7 from a DOS floppy, it works very well on 486 boards. Google: cachechk v7 and get it from www sac sk

I have my answer, I have fake cache bios and I'll try to flash it (as it should be possible as per my manual).
I'll give it a go and see what happens. Worst case I buy a programmer.
😀

Can I ask if you ever got the bios flashed and if so what tool did you use to do it?

I've been banging my head off a wall trying to upgrade my M912 with real cache. Installed 1MB (only cost me £20) but it just hangs at post. If I change the jumpers to only use 1 bank it works fine at 512kb. Oddly cachechk reports a total of just 512kb and no L2 cache. Total should be 8k L1 and 512k L2.

I've tried 5 different ami flash tools to update the bios with no luck. Nothing will read it. Unfortunately I don't have an eeprom programmer so I'm stuck if I can't find a tool that works.

Unite wrote on 2020-11-29, 02:50:

Can I ask if you ever got the bios flashed and if so what tool did you use to do it?

I've been banging my head off a wall trying to upgrade my M912 with real cache. Installed 1MB (only cost me £20) but it just hangs at post. If I change the jumpers to only use 1 bank it works fine at 512kb. Oddly cachechk reports a total of just 512kb and no L2 cache. Total should be 8k L1 and 512k L2.

I've tried 5 different ami flash tools to update the bios with no luck. Nothing will read it. Unfortunately I don't have an eeprom programmer so I'm stuck if I can't find a tool that works.

Are you sure that you board has an EEPROM rather than an EPROM ? EPROMs first need to be erased using UV light and programmed externally in a programmer . What is the model of your BIOS chip and does it have a little "window" on it hidden by a sticker ?

486 boards usually used EPROMs . EEPROMs (that can usually be re-flashed while installed in the motherboard) became more common with Pentium based machines .

Hmm. Can you take a good picture of the full board and post it ? Try to gently peel the sticker off the BIOS chip so we can read the ROM markings before taking the picture.
Have three - 1.7 and one - 1.4, all real cache and not one came with 1Mb, three have 256k and one has 512k (and those 512k take up the whole extended length of all 8 of the cache slots).
The TAG must be 64Kx8 (same as all the main cache if 512k) to support 1Mb, if it is only 32Kx8 then 512k is all that it will show.
http://web.archive.org/web/19970327145447/htt … com/6900-3.html
One thing to keep in mind: The TH99 and many sources list the Amptron dx6900 VARIANT as v1.75 but my boards are silkscreened v1.7 not 1.75 so there could be some difference between 1.7 and 1,75 (never found any good info)
Also if you can get 512K cache working on a 486 that is great ! The board only supports 64Mb ram and 512k is more than enough for that 😀

Cachechk and other memory tests show the 1MB cache helping over 512KB.

I have a decent number of t M912 boards, both 1.4 and 1.7. Never really even tried to acquire them in the first place. They just came in lots with other stuff or in systems I bought without knowing the internals.

The AMI BIOS is way better than the AWARD BIOS as it supports large HDDs.

Unite wrote on 2020-11-29, 02:50:

Can I ask if you ever got the bios flashed and if so what tool did you use to do it?

I've been banging my head off a wall trying to upgrade my M912 with real cache. Installed 1MB (only cost me £20) but it just hangs at post. If I change the jumpers to only use 1 bank it works fine at 512kb. Oddly cachechk reports a total of just 512kb and no L2 cache. Total should be 8k L1 and 512k L2.

I've tried 5 different ami flash tools to update the bios with no luck. Nothing will read it. Unfortunately I don't have an eeprom programmer so I'm stuck if I can't find a tool that works.

Hi!
I've tried a lot of times and different bios tools to flash the bios but none worked, always an error (unsupported chipset) or hangs. This doesn't mean the right tool doesn't exist, could have been some precise version, really IDK.
Cache simply isn't working, like if non were installed.

I'm about to order a chip flasher to program a new one and be done with it. Not that I'm 100% sure it'll support L2 cache for real.

cyclone3d wrote on 2020-11-29, 06:44:

The AMI BIOS is way better than the AWARD BIOS as it supports large HDDs.

Noted. I'll try both ASAP as I get the programmer.

Here's a photo of my board and a close up of the bios with the sticker removed. I'm not sure if this is an eeprom or eprom. If I can't program it myself I'll have to get someone to burn a chip for me. Playing around with the computer again today and I'm no further along. Still works with 512k but nothing more. For the TAG ram I just used another 128k x 8 chip, I had to order 10 of them to get the 8 I needed so just used a spare one as tag. I've tried swapping out the 4 chips in bank 1 with the other spare on the chance of it being a bad chip but no luck. That's why there are sockets at bank 1 and none at bank 0.

https://i.imgur.com/SJynT24

https://imgur.com/7ZhkzcV

I have no markings on my BIOS chip. No window.
Just 57085-1 printed under it.

If I try to flash it with a lower voltage, 5 instead of 12?
With higher I'm gonna fry it, if from 5 to 12.

This chip is definitely relabelled. PCChips loved to do that, even down to the ROMs

You can swap it with a regular 27C512 chip and it will work.

Unite wrote on 2020-11-29, 14:26:

Here's a photo of my board and a close up of the bios with the sticker removed. I'm not sure if this is an eeprom or eprom. If I can't program it myself I'll have to get someone to burn a chip for me. Playing around with the computer again today and I'm no further along. Still works with 512k but nothing more. For the TAG ram I just used another 128k x 8 chip, I had to order 10 of them to get the 8 I needed so just used a spare one as tag. I've tried swapping out the 4 chips in bank 1 with the other spare on the chance of it being a bad chip but no luck. That's why there are sockets at bank 1 and none at bank 0.

https://i.imgur.com/SJynT24

https://imgur.com/7ZhkzcV

Did you forget to set JP3, programming voltage for bios chip? 1-2 is 5V, 2-3 is 12V

JP3 is close to the keyb connector

Edit: tried all combinations and it changed nothing, 5, 12 volts. Got chipset not supported and flash protected. One award reported "wrong size".
I can only program a chip and replace the current one.

I finally tried, with the help of the TL866, two other Bios:
- 9121201
- 12-02-1195X

Neither had the cache working and would hang trying to boot off the HD or floppy.
Reseated and moved cache around and boots perfectly, although it doesn't show in CACHECHK, only the 16KB of the cpu.

The oringinal unflashable Bios chip now allows cache size different from 256kb, when before it wouldn't budge from 256... Go figure. (really odd)

Is something I'm missing?
Suggestions are welcome.

What happens if one of the chips has an address line broken? Could a broken lane be the culprit?
It worked past hd booting, and now only with l2 cache disabled. There is definitely something with cache detection.
When cachecheck fails, it ometimes show #1 meg not cached (or failed caching).

Tomorrow I'll try resoldering the sockets (only legs).

Nexxen wrote on 2020-12-24, 21:00:

What happens if one of the chips has an address line broken? Could a broken lane be the culprit? It worked past hd booting, and […]
Show full quote

What happens if one of the chips has an address line broken? Could a broken lane be the culprit?
It worked past hd booting, and now only with l2 cache disabled. There is definitely something with cache detection.
When cachecheck fails, it ometimes show #1 meg not cached (or failed caching).

Tomorrow I'll try resoldering the sockets (only legs).

Good question ! I know the jumpers tell the chipset+BIOS what size the cache is, but if an address line or other oddity comes up the cache may be disabled by BIOS afaik.
Am thinking of the reverse of the fake cache scenario where the BIOS and chipset were tricked into thinking there was external cache when there was not....
Am wondering if there could be something else slightly different between the real and fake cache v1.7, cannot image the board being different so maybe something as simple as a pull up/pull down resistor missing ?
Just a guess. Can you post a good picture of your board ? Will dig out my v1.7 with real cache and post a pic of it to compare with if you do not have a good pic.

I went through the cache socket joints again and resoldered the ones that looked meh.

Now it randomly boots to "128KB cache" and stops, or "Write-back cache enabled" (will correct as I don't remember the exact line).
Testing with Cachechk it outputs:

"This machine seems to have one cache!? [read]
!! cache is 16KB -- 102.2 mb/s ..."

Nothing about an external cache.

Pics

more pics

Does 3DBench 1.0 get affected by L2 cache?

I don't think this can be solved easily. If nothing turns up I'm going to use this board to test cpus dead or alive with no expecttions.
Basically I'm out looking for another board.
Instead of X99 boards from China they should produce up to 486 motherbaords, they'd have market...... 😀

Thanks!

Jumpers are correct for 256k 9chip and see no missing resistor packs. Are you sure your wire fixes are correct ?