VOGONS


Converted registered SDRAM

Topic actions

Reply 40 of 73, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie

Ok, try looking at this resistor on the top side:

32Chip_SDRAM_PossibleLE_PullUp.jpg
Filename
32Chip_SDRAM_PossibleLE_PullUp.jpg
File size
44.47 KiB
Views
1008 views
File license
Fair use/fair dealing exception

Measure the resistance from both sides to pin 28 on the register. I'm hoping that the side furthest from the edge connector reads close to 0. If that's true then the other side should connect to Vcc (pin 35 on the register). If those are both true (one side is !LE and the other Vcc) then that's the pull up resistor. If you can remove that, then recheck the resistance from !LE to Vcc and it should read much higher than before.

If it's not that resistor, check the ones nearby.

Without that pull up then it might (and this is very might) be possible to add a pull down by using a pencil to draw between the Y and GND pins in the previous photo. Use a soft pencil and make a wide trace, then measure the resistance between them. But it'd be better to try and fit the 10k resistor that you will have just removed. It's really not too hard to solder things this size, just give yourself time and patience, and have a pair of small tweezers and a magnifying glass to hand.

[edit: Just realised I'm being dense. The pull up has to be connected to the input of where the inverter would be, so it's only connected to !LE through that 0 ohm link next to the inverter pads. So removing that 0 ohm link will disconnect the pull up from !LE. At which point a weak pull down (like a pencil link) on !LE should work. So:
1) Remove the 0 ohm link. This should disconnect output Y (which goes to !LE) from both pin 147 REGE and from the 10k pull up.
2) measure the resistance from output Y to Gnd and to Vcc. Y to Vcc was about 10k, should now be much higher.
3) draw a nice heavy, wide pencil link from Y to Gnd and recheck the resistance. Make sure the line doesn't touch input A.

If this works, can worry about making a more reliable mod later]

[edit2: scratch the pencil idea. I just gave it a go and I can't get the resistance low enough when drawing on the smooth surface of the PCB (ends up around 200k, think it would want to be <50k). So it probably won't reliably pull the !LE low. So you'll need to either solder a wire or draw a line. I've never used a conductive pen so don't know what to recommend. I suspect that for this purpose any will do. But soldering a wire really wouldn't be that hard. Use a single strand of some multistrand cable, slightly over length, hold one end with tweezers and solder the other end to the furthest pad, then solder the nearest pad, then use a sharp knife to trim the wire to length. Or just use that 0 ohm link.]

Reply 41 of 73, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

I don’t see a 103 anywhere on the board.

I installed a 10k (103) and removed the link.

That’s not enough is it? I gotta put in a 1k don’t I?

Attachments

  • image.jpg
    Filename
    image.jpg
    File size
    959.23 KiB
    Views
    926 views
    File license
    Public domain

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 42 of 73, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

In anticipation of your answer that I screwed up I present to you: 1k resistor.

Does this look correct?

Attachments

  • image.jpg
    Filename
    image.jpg
    File size
    827.11 KiB
    Views
    926 views
    File license
    Public domain
Last edited by Sphere478 on 2021-12-25, 06:26. Edited 1 time in total.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 43 of 73, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

I set up a buggy motherboard I have, and did a test. It posted! And counted 512mb

Not 1gb 🙁

Probably no way around that huh?

Nothing left to do now except mod the other two!

Edit:

image.jpg
Filename
image.jpg
File size
1.91 MiB
Views
911 views
File license
Public domain

Unfortunately it seems that the bios gives up after counting 512mb in one bank and just reports 512mb total.

So unless there is a specific type of ram chip that I should be looking for this seems to not be the answer to 3gb. 🙁

Trying a non ecc non reg non buffered stick in bank 1 and a 1gb in bank 2 it shows 655360kb as I expected it would.

It’s like if the bios counts even one kb over 512mb it just gives up.

Also, wild that a ecc reg and a normal stick play well together.

So in theory I could use my two 256mb ecc reg sticks (after modding them) in bank 1,2 and a 1gb in bank 3 for a total of 1gb (since I don’t have any 512mb sticks)

Any ideas on how to get 3gb workin?

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 44 of 73, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

And finally these sticks

Edit: whoops, don’t forget the resistor T is on both sides.

All good now! These work also!! Woohoo

Attachments

  • image.jpg
    Filename
    image.jpg
    File size
    1.75 MiB
    Views
    899 views
    File license
    Public domain
  • image.jpg
    Filename
    image.jpg
    File size
    1.38 MiB
    Views
    899 views
    File license
    Public domain
  • image.jpg
    Filename
    image.jpg
    File size
    1.78 MiB
    Views
    906 views
    File license
    Public domain

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 45 of 73, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

Another update:

Previous tests were on a ga-5aa rev 2.2

Just tried on a p5a 1.4 and got some odd results.

It doesn’t post with the 256 sticks, and hangs on memory counting but does count 1024mb however not 2048 with 2 sticks. It just stops at 1024

Any idea why the 256 sticks aren’t posting?

Think it’s the T resistor?

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 46 of 73, by PC@LIVE

User metadata
Rank Oldbie
Rank
Oldbie

Today I discovered this interesting discussion, from what I understand you make some changes to RAM for Server to adapt them to PCs, in the latest writings I read 1GB RAM modified to adapt them to MB Socket7.
So maybe I too would have modifiable RAM, but I want to keep them for any MB server to be repaired, apart from that I have to congratulate you on the results obtained.
Going back to the 1GB PC100 or PC133 RAM, they exist but those for PC are very rare, while it is quite easy to find those for Server (even if they cost), I think they were for the first i845 (post RIMM). I have an AZZA 845S-AV PC with three PC133 DIMMs, I currently have 1.5GB of RAM, so three 512MB DIMMs, not having a 1GB one, I couldn't go further.
Here I thought that a similar MB could exactly recognize the modified RAM, trying with a 1GB RAM you can see how much it detects, or using two or three if it reads 2GB or 3GB, this could be useful to know if the detected capacity is a limit of the mb (or chipset).

AMD 286-16 287-10 4MB HD 45MB VGA 256KB
AMD 386DX-40 Intel 387 8MB HD 81MB VGA 256KB
Cyrix 486DLC-40 IIT387-40 8MB VGA 512KB
AMD 5X86-133 16MB VGA VLB CL5428 2MB and many others
AMD K62+ 550 SOYO 5EMA+ and many others
AST Pentium Pro 200 MHz L2 256KB

Reply 47 of 73, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie

What's the value of that resistor I circled in red in my last post? It doesn't actually matter since a 1k pull down with a 10k pull up should be fine anyway (voltage will end up around 0.3V, so will be seen as a '0'), and will make sure there aren't any large currents if anything does try to drive the REGE pin. Glad to hear it (sort of) worked. Sounds like sort of Yay! it works and Boo! not as wanted. So the 1GB sticks get seen as 512MB, and the board can still only see a maximum of 128MB+512MB.
I wonder if this is a BIOS limitation of some sort? What does memtest say for memory capacity? Or booting in to linux (I remember it could sometimes find hardware that had been disabled in the BIOS).

I think those couple of 'T' resistors seem to set the clock skew, so will change the stability of the RAM. For at least one of my sticks I found that the default position only worked at PC100, but fiddling with the 'T' got it working at PC133.

Sphere478 wrote on 2021-12-25, 06:26:

Also, wild that a ecc reg and a normal stick play well together.

Well, I think from the motherboard point of view they're both unregistered, just one of them has a slightly higher latency and lower load on the bus.

Reply 48 of 73, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

The circled one us a 0 ohm link

Maybe chkcpu has some ideas on why 1gb isn’t working.. my understanding is the chipset and cpu/cache support up to 4gb in my configuration so my guess is a bios error also.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 50 of 73, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
rasz_pl wrote on 2021-12-30, 13:34:

might be internal organization of this particular module not being supported (similar to 440BX SDRAM limits)

Tried it with a 16 and 32 chip stick

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 51 of 73, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie
Sphere478 wrote on 2021-12-30, 22:56:

Tried it with a 16 and 32 chip stick

Would you mind list which sticks you've tried, with front and back photos of them, and what the symptoms were. I've been having a read through datsheets of the M1541 and a few 1GB SDRAM sticks, and can't just see anything yet that would stop anything working, assuming the motherboard routes all the available signal lines. Also, some photos of the memory slots. I want to check if it looks like anything is connected to pin 126 (Address bit 12). That's used as the highest bit when setting the Row address and isn't used for Column addressing, so if it was missing, then maybe large sticks (using 13 bit row addressing) would break.

Actually.... Since I was thinking about pin 126 on the memory slot, I thought I'd just check the M1541 datasheet and check what pin it would be on there, just on the off chance that it was somewhere easy to check. That A12 line connects to MA[14] of the 15 bit memory address bus [14..0] on the M1541, which is on pin C23. And that's miles away from the rest of the MA bus, all neatly arranged down on rows J,K&L. It's also in the middle of the memory Data bus pins. I can easily imagine that line not getting routed out as it'd be a right royal pain getting the bus lengths matched with one trace starting from that far away, and might cause problems routeing the data bus nicely. And the only downside would be not supporting memory sizes that were only sort of available.

So that's now my current theory why 1GB sticks fail: pin C23 on the M1541 should connect to pin 126 on the SDRAM slots, and maybe it isn't because why spend a couple of weeks trying to route one signal that messes up both the address and data bus fan out.

[edit: MA14 is pin AD23 on the M1542
also, possibly interesting short discussion here: https://www.wimsbios.com/forum/ram-compatibil … n-t7262-15.html about reported size not matching installed memory, although with no actual resolution]

Reply 52 of 73, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
snufkin wrote on 2022-01-01, 19:55:
Would you mind list which sticks you've tried, with front and back photos of them, and what the symptoms were. I've been having […]
Show full quote
Sphere478 wrote on 2021-12-30, 22:56:

Tried it with a 16 and 32 chip stick

Would you mind list which sticks you've tried, with front and back photos of them, and what the symptoms were. I've been having a read through datsheets of the M1541 and a few 1GB SDRAM sticks, and can't just see anything yet that would stop anything working, assuming the motherboard routes all the available signal lines. Also, some photos of the memory slots. I want to check if it looks like anything is connected to pin 126 (Address bit 12). That's used as the highest bit when setting the Row address and isn't used for Column addressing, so if it was missing, then maybe large sticks (using 13 bit row addressing) would break.

Actually.... Since I was thinking about pin 126 on the memory slot, I thought I'd just check the M1541 datasheet and check what pin it would be on there, just on the off chance that it was somewhere easy to check. That A12 line connects to MA[14] of the 15 bit memory address bus [14..0] on the M1541, which is on pin C23. And that's miles away from the rest of the MA bus, all neatly arranged down on rows J,K&L. It's also in the middle of the memory Data bus pins. I can easily imagine that line not getting routed out as it'd be a right royal pain getting the bus lengths matched with one trace starting from that far away, and might cause problems routeing the data bus nicely. And the only downside would be not supporting memory sizes that were only sort of available.

So that's now my current theory why 1GB sticks fail: pin C23 on the M1541 should connect to pin 126 on the SDRAM slots, and maybe it isn't because why spend a couple of weeks trying to route one signal that messes up both the address and data bus fan out.

[edit: MA14 is pin AD23 on the M1542
also, possibly interesting short discussion here: https://www.wimsbios.com/forum/ram-compatibil … n-t7262-15.html about reported size not matching installed memory, although with no actual resolution]

The 32 chip sticks are posted above.

I don’t have the 16 chip ones anymore but here is a pic the only one I have

Asus p5a cache upgrade!
^Mobo pics^

Two behaviors observed.

One: it counts all 1gb of memory and hangs (p5a does this)

Two: it counts 512 and gives up. If 1gb in slot 3 it will count sticks 1 and 2 so long as not 1gb sticks in those slots. So 1,2,3 256 256 1gb equals 1gb and 1gb works but if 1gb is in slot 1 it skips 2 and 3 and you get 512 ga-5aa does this

Attachments

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 53 of 73, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie

Ok, so maybe the ga-5aa doesn't route MA[14] and stops counting when it hits an error. And the p5a does route MA[14], but isn't happy with the timings. What is the bus frequency set to? If you haven't already, might be worth booting with a working stick, going into the BIOS settings and setting a bunch of safe defaults (disable caches, quick POST, and BIOS RAM shadowing, set all the SDRAM timings slow) then save those settings, then swap out the working memory for the 1GB stick.

Also, what does it count up to with multiple 1GB sticks? Or does it just count to 1GB and hang?

I'm curious on the ga-5aa what memtest86 shows when testing a 1GB stick (that the BIOS sees as 512MB), or what linux would find (it seems to bypass the BIOS for some things).

Reply 54 of 73, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
snufkin wrote on 2022-01-02, 12:18:

Ok, so maybe the ga-5aa doesn't route MA[14] and stops counting when it hits an error. And the p5a does route MA[14], but isn't happy with the timings. What is the bus frequency set to? If you haven't already, might be worth booting with a working stick, going into the BIOS settings and setting a bunch of safe defaults (disable caches, quick POST, and BIOS RAM shadowing, set all the SDRAM timings slow) then save those settings, then swap out the working memory for the 1GB stick.

Also, what does it count up to with multiple 1GB sticks? Or does it just count to 1GB and hang?

I'm curious on the ga-5aa what memtest86 shows when testing a 1GB stick (that the BIOS sees as 512MB), or what linux would find (it seems to bypass the BIOS for some things).

On mine it counts to 1gb and hangs but others have gotte. 2gb to count on unknown ali v board

Memtest sees what bios sees

Dono about cpuz and linux.

I don’t have any of these systems in service at the moment. I’ll try to revisit this later though.

I only have those ecc reg 1gb stciks that are modified to work on normal mobos atm

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 55 of 73, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie

Ok, I don't have an Ali chipset to test any of this on, but when you have either board out again, might be worth a quick check to see if there's continuity between pin 126 on all the RAM slots, just to check if it's at least routed between them all.

Part of me is now wondering what happens if I try to put 4GB in a KX133 system. The KA7 I have says it supports up to 512MB in each slot, and nearly all the other KX133 boards seems to have either 256MB or 512MB maximum stick size. But the VIA reference design VT5249 only has 2 slots, with apparently 1GB supported in each slot. There's only a partial summary datasheet for the VT8371, and that specifies a maximum of 2GB. But there's no details about what the cause of that limit is, and I can't just find if anyone has tried breaking the limit before.

The whole thing is plainly ridiculous though. I'm pretty sure I was happy when I scrabbled together enough sticks to get me up to 128MB, which was a step up from the 32MB on the 430VX board I had before that. So trying to bung in 4GB seems a bit out of keeping. But there is a UK seller on ebay currently with some 1GB sticks, similar to yours, but with the adjustable 'T'. I could get a couple, to get up to the apparent 2GB limit, see if it works, then add anything in slot 3 just to see if it counts past 2GB. But it'll be about £35 for a couple of them, which is currently more than my level of curiousity will cover. Hmm.

I've got an old laptop currently in pieces that I ought to finish fiddling with first (so far I've managed to make a broken flat-flexible keyboard cable into a brokener FFC). I'll get that back together first, then maybe come back to this.

Reply 56 of 73, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
snufkin wrote on 2022-01-02, 21:26:
Ok, I don't have an Ali chipset to test any of this on, but when you have either board out again, might be worth a quick check t […]
Show full quote

Ok, I don't have an Ali chipset to test any of this on, but when you have either board out again, might be worth a quick check to see if there's continuity between pin 126 on all the RAM slots, just to check if it's at least routed between them all.

Part of me is now wondering what happens if I try to put 4GB in a KX133 system. The KA7 I have says it supports up to 512MB in each slot, and nearly all the other KX133 boards seems to have either 256MB or 512MB maximum stick size. But the VIA reference design VT5249 only has 2 slots, with apparently 1GB supported in each slot. There's only a partial summary datasheet for the VT8371, and that specifies a maximum of 2GB. But there's no details about what the cause of that limit is, and I can't just find if anyone has tried breaking the limit before.

The whole thing is plainly ridiculous though. I'm pretty sure I was happy when I scrabbled together enough sticks to get me up to 128MB, which was a step up from the 32MB on the 430VX board I had before that. So trying to bung in 4GB seems a bit out of keeping. But there is a UK seller on ebay currently with some 1GB sticks, similar to yours, but with the adjustable 'T'. I could get a couple, to get up to the apparent 2GB limit, see if it works, then add anything in slot 3 just to see if it counts past 2GB. But it'll be about £35 for a couple of them, which is currently more than my level of curiousity will cover. Hmm.

I've got an old laptop currently in pieces that I ought to finish fiddling with first (so far I've managed to make a broken flat-flexible keyboard cable into a brokener FFC). I'll get that back together first, then maybe come back to this.

I mean, it’s totally possible to use all that ram when opening up the right programs but sure, this is all for fun 😀

Any chance you can show me which pin that is?

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 57 of 73, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie
Sphere478 wrote on 2022-01-03, 02:30:

Any chance you can show me which pin that is?

I've had a go at marking it, although I've got a history of being unable to count past 1. I'm pretty sure it's one past the middle separator.

5aa_Pin126_SDRAM_A12.jpg
Filename
5aa_Pin126_SDRAM_A12.jpg
File size
159 KiB
Views
684 views
File license
Fair use/fair dealing exception

[edit: forgot to ask... Given the 1GB sticks currently don't work as intended, but also didn't blow anything up, where are we on the heresy front? Can it be partially heretical? Is it in some sort of quantum heresy state? Also realised I haven't yet said Happy Christmas (still valid for a few days yet) and New Year to all here]

Reply 58 of 73, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
snufkin wrote on 2022-01-03, 11:36:
I've had a go at marking it, although I've got a history of being unable to count past 1. I'm pretty sure it's one past the mid […]
Show full quote
Sphere478 wrote on 2022-01-03, 02:30:

Any chance you can show me which pin that is?

I've had a go at marking it, although I've got a history of being unable to count past 1. I'm pretty sure it's one past the middle separator.
5aa_Pin126_SDRAM_A12.jpg

[edit: forgot to ask... Given the 1GB sticks currently don't work as intended, but also didn't blow anything up, where are we on the heresy front? Can it be partially heretical? Is it in some sort of quantum heresy state? Also realised I haven't yet said Happy Christmas (still valid for a few days yet) and New Year to all here]

No connection between slots on that pin on p5a 1.04

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 59 of 73, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie
Sphere478 wrote on 2022-01-04, 01:48:

No connection between slots on that pin on p5a 1.04

Huh. Ok. Either I've miscounted and that's not pin 126. Or it's not routed on the motherboard. In which case how does it count to 1GB when it's missing A12, so half the Row addresses are missing?

Maybe it doesn't know that it can't, and it's actually wrapping around. If it reads and believes the SPD then it knows that it needs to test 1GB (128M locations * 8 bytes at each location). If it just does a simple test (write value, read value back) then it won't be able to tell if the location wraps around when it gets past half way. It'd need to do a test where it first wrote non-repeating data to all the addresses before reading back. So it could be doing a simple test and passing, then it writes some actual data to RAM, overwriting something important, and crashes. Maybe the BIOS on the ga-5aa stops at 512MB not because it gets an error, but just because it's programmed to not count past 512MB because it can't address it.

Can you measure the resistance from that pin 126 to pin 127 (Ground) and to pin 124 (Vcc)? Maybe they just pull it high or low.

Still, if that is pin 126 and it's not routed, then that'd explain why 1GB won't work. Which is annoying. Getting the A12 line out from the 1541 would be tricky.