Great job - a custom module would be awesome!

i have the same problem as you , one of my sticks does not work at all
it never gets detected on boot , like. it boots with fake mode on always
i did try cleaning the contacts, cheching most of them with a multimeter but no luck.

i also bought the alliance chips, and got some doldering paste that melts instantly with hot air. i will need to do the same process that you did.

Well I've ordered about 200 of the alliance SRAMs in anticipation of making these modules. I've been messing around with KiCad and it'll be a while with my learning curve before I'm able to produce anything.

Has anybody ever noticed that there exist modules with 3.3V SRAM memory chips and others with 5V chips ?
For example TC55V328AJ and IDT71V256 are 3.3V parts, AS7C256 is a 5V part. There are others like HM62H256AJ or UT61M256JC I was not able to find a datasheet for. It must be noted that some of them are 3,3V only parts, not 5V tolerant parts.
I measured the supply voltage on the cache module and it is indeed 5V.
I wonder if this is assembly was intentional, because it is a guaranty for early failure and the reason why many of these sticks are non-functional anymore ?

BastlerMike wrote:

Has anybody ever noticed that there exist modules with 3.3V SRAM memory chips and others with 5V chips ? For example TC55V328AJ […]
Show full quote

Has anybody ever noticed that there exist modules with 3.3V SRAM memory chips and others with 5V chips ?
For example TC55V328AJ and IDT71V256 are 3.3V parts, AS7C256 is a 5V part. There are others like HM62H256AJ or UT61M256JC I was not able to find a datasheet for. It must be noted that some of them are 3,3V only parts, not 5V tolerant parts.
I measured the supply voltage on the cache module and it is indeed 5V.
I wonder if this is assembly was intentional, because it is a guaranty for early failure and the reason why many of these sticks are non-functional anymore ?

Perhaps some M919 supplied 3.3V to the cache stick... Just a wild guess but with PCchips and their dirty tricks you can never be sure.

after you take out the chips from the pcb

how do you test each one to see if its good or faulty?
im asking because i have no idea how to test each one, i also got a coast module that on boot it never gets detected.

i bought replacement alliance chips but i dont know which one is the faulty one

you can modify a known working cache stick with a socket and test each chip.
I once did this to identify faulty EDRAM

But if you already ordered replacement chips I doubt that this is worth the time.
You have all chips desoldered anyway, it is easier to replace them all at once with new ones

Today I soldered new chips onto my M919 cache module. I used IDT71256 and one SB61L256A because I only had eight of the IDT chips . They are all designed to work at 5 volts. Works great now.

BastlerMike wrote:

Has anybody ever noticed that there exist modules with 3.3V SRAM memory chips and others with 5V chips ?

These screen shots from the Amptron DX-9700 manual should explain that.

Amptron DX-9700 is given in several places as = PcChips M919 v2.x-v4.x.
I don't think there ever was a 4.x M919 but that's what the references say.

This might also explain the M919-02S5 and M919-01S3 part numbers.
I only have M919-02S5 to look at.
.

.
.
I might be mistaken but I thought I saw a 1998 BIOS for M919.
If it exists and someone has a copy it would be appreciated.
Thanks.
.

PCBONEZ, good find. That information seems to suggest that the cache voltage is the same as the jumpered CPU voltage. Is this really true and has anyone measured this? Could that manual you found be wrong? I have the official printed manual that comes with this board and it doesn't mention anything about cache voltage like you found.

But if what you found is true, it could explain why the board is so flakey. Perhaps I have 5 V cache chips which aren't all that reliable at 3.3 V. I have three m919 modules but have never bothered to lookup the spec sheets for the chips.

What surprises me is that people have enough interest in this motherboard to bother making new SRAM slot modules. Why the interest in this board? Is it because its the only board with BIOS-included Cyrix 5x86 LSSER and LINBRST features? Will fake cache PC Chips boards become collector's items?

I am attaching the BIOSes I have for this board. They have these dates:

9190506.BIN -- 05/06/1996
9191016S.ROM -- 10/16/1996
9190914S.ROM -- 09/15/1998

I don't recall if the 10/16/1996 or 09/15/98 BIOSes work. I just know I tested them at some point. The board came with 05/06/1996 BIOS.

Thank you.
I ordered some of those reproduction module blanks to play with.
Among other things, I want try try using sockets instead of soldering the chips directly on them.
The other parts (including the mobo to test them in) aren't here yet so I'm doing the research and collecting files.
I will need to rework the pre-tinning. Its full of blisters/bubbles as if the PCB wasn't cleaned before it was applied.
The PCB itself looks pretty good.

I've seen that same cahce-voltage information in both Amptron and PcChips docs now.
The PcChips version is all over the place but made generic probably for rebranding purposes.
Back in the day even small mom-n-pop shops would rebrand PcChips as their own house brand.
PcChips made it easy. You could order without any branding on the boxes or in the manuals.
I visited the PcChips showroom in Hong Kong in the 90's and the rep said you could order without BIOS chips too.
(It was intended for buyers ordering 100's or 1000's of boards. I just found myself in front of it so I went in.)

Amptron manual version:

PcChips manual version:

I suspect the cache module voltage is set by the same jumpers that set CPU voltage.
The board I ordered isn't here yet so I can't look into that.

Other speculations:

I've seen a lot of model different numbers on these modules.
I suspect some of the 3rd party distributors where making their own modules in-house.

I've seen a number of those 3rd party website via wayback now.
I've seen none that offered the modules for sale outright.
I suspect if you didn't order/buy the board with the module option as original you couldn't get one.

I don't understand why PcChips is "a thing" either. They gave bottom-end a lower bottom.
In the late 90's a few shops outsourced testing their returns to me because their own techs couldn't keep up with them all.
I would sometimes see 5 or 6 out of the same (factory) batch and the stability varied a lot even within batches.
It had to be flaky quality control (or none) at the factory.
They ARE fun to play with, but that's about it. I wouldn't choose one for a final build.

--- Here's a kicker
Hsing Tech (The parent company. PcChips was their distribution branch and public 'front'.) merged with ECS in 2005.
That more or less means that PcChips is still around.
.

A few other things I tripped over. May be well known already, I didn't look.
https://web.archive.org/web/19980425123639/ht … 19/919_001.html
https://web.archive.org/web/19970204001606/ht … com/4500faq.htm

"Q8" in the second link has info about EDO and the chipset version.

I don't understand why PcChips is "a thing" either. They gave bottom-end a lower bottom.

Yeah, in a nutshell.

I've run across most of these manuals over the years, but don't know how I missed that there was 5V and 3.3V cache modules. When I have time, I'll try to locate the spec sheets of my three m919 module IC's.

I took a look at my m919 SRAM modules. Looks like one of them is 3.3 V only, one is 3.3 V or 5 V, and for the last, I could not locate the datasheet. So looks like my Alliance module is the more useful one as it can swing both ways.

IDT 71V256SA15Y = 3.3V, Absolute maximum rating of 4.6 V

Alliance AS7C256-15JC = 5V / 3.3V, Absolute maximum rating of 7.0 V

HMC HM62H256AJ-15 = ? could not locate datasheet ?

I haven't found HMC HM62H256AJ-15 either.
I did find HM62256 and HM62256A which both have a 4.5-5.5v range.
I have not figured out what the H or J represent.

The IDT 71V256SA15Y one is marked M919-01S3 and 3.3v/4.0v okay but 5v CPUs are out_ -01S"3"

Alliance AS7C256-15JC one is marked M919-02S5 and okay to 3.3 to 5v _ -02S"5"

The HMC HM62H256AJ-15 one has the -02S"5" number so presumably it's at least okay for 5v CPUs.

We don't have much of a sample size and it's quite possible some of the original sticks floating around had the chips changed to the wrong kind at some point so we will probably never know for sure, but: so far the last digit in the PN indicating the intended voltage seems plausible to me.

I like your theory on the 02S5 and 01S3 markings. It is probably valid.

I've had these sticks for around 15 years, so it seems less likely that some enthusiast went changing the chips.

Just wanted to contribute to this thread since I got so much help on my own M919 thread: I accidentally bought a PC Chips M919

My cache module is a 02S5, and it came with an Amptron DX9700. The theory that the "5" means that the module is okay for 5v may still be correct, but there was a sticker on the back that says 3v. I'm going to stick with 3.3v CPUs when using this board just to be on the safe side.

As someone without soldering skills, I've been monitoring projects trying to snag one but haven't seen any movement.

Well, an eBay alert fired off for a new clone. I was able to order one, and I'll post back after I receive it. So, everyone, that's waiting. There is hope.

@Boxpressed The datasheet for the chips on that stick suggests they run off 5v for power, and will interpret anything above ~2V and up to 0.5 volts above supply voltage as logic-high, so they should work with no problem with either 3.3v or 5v I/O. At the very least if it works at 3.3v, I don't think it'll release the magic smoke at 5. I don't think the chips themselves can run off any supply voltage besides 5v +/- 0.5v.
Only other thing on the board (looks like) are capacitors, and I don't think those will restrict the voltage.
Datasheet:
http://read.pudn.com/downloads105/ebook/43530 … %8C/UT62256.pdf