Potentially Stupid RAM Hardware Mod Question

[jessenator]

Forgive my ignorance, but I had a question surrounding the hard-to-find single-bank EDOs for the Power Mac 4400 and the like:

Is it possible to take a double-banked 64MB EDO, like this one, and remove the chips from one side to turn it into a single-banked 32MB EDO?

Is that even possible? Are both banks integral to each other in that sort of arrangement, or do they operate in sort of parallel and can be segmented? O ris it more complex than that and this is another one of my pipe dreams :lol:  

Again, I'm just shooting in the dark here—just looking for a solution to maxing out my 4400  

 

[Gorgonops]

I was under the impression that you'd be fine plugging a double-banked DIMM into it and it'd just not see the other side, so there's no need to desolder anything, but I've never owned one of those systems so definitely don't take my word for it.

 

[jessenator]

I guess I could just try booting with the one I have and see. Worst that will happen is ...nothing

 

[Gorgonops]

I would swear there's been threads/pages where people claiming to have "192MB" in those 160MB-strapped machines based said claim on having three 64MB DIMMs installed, only to have their feelings crushed when it was pointed out that they really weren't getting everything they paid for. But those memories may well have been implanted by brain-eating Martians.

 

[jessenator]

Haha! Well at $5 a stick (if I wanted to ditch the extra tall EDO that's in it) $15 is not a bad price for an experiment.

I might wait for a day or two to let anyone with an adamant warning to not try testing the stick in the first slot have their chance.

 

[jessenator]

So I got impatient. Chimes every time and full ram visible:

Started with the 32 in the first slot, then added the 64, then the 16: https://imgur.com/a/L3SUaR3

Everything is seen...

Is there a danger to running it long term I wonder? Powerlimitations on the board? Shouldn't be a problem with my aftermarket PSU.

 

[Gorgonops]

 

[jessenator]

Ha!

I wonder if @omidimo or @Franklinstein have any experience on the matter.

I mean, really, the current running through the board is the only thing I can think of that could be problematic, now that the supposed "YOU CAN'T DO THAT!!1! RREEEEEEEEE—" guideline of installing a dual-banked EDO into the single-rank first slot has been debunked. Am I in danger of blowing up board VRMs or causing other potential collateral damage elsewhere?

My new PSU is 460 peak watts, more than double the original, so I don't think I'm in danger of overloading it.

 

[trag]

If you're seeing the full RAM you should be fine.   The currents involved are tiny.    They could cause signal integrity/ringing problems if not balanced properly, but they won't blow the board.

I suppose in a ridiculously sensitive machine, it might be possible to destroy the output buffers on the memory controller.

The bits of logic that output the actual signals to the DRAM modules are rated to output a certain level of current.   If you connect a whole bunch of extra RAM chips to them, then for example, the address output buffers might see too low of a resistance (too much current drawn) and this could in effect be the equivalent of being exposed to a short to GND or 5V.     But that's really far fetched.

The main thing I was concerned with, when I read your first message, is memory that seems to work, but isn't actually reliable.   This would only be a concern if your two bank memory was seen as 1/2 capacity.    If the board was truly one bank only, then it would be activating all the /RAS signals to a DIMM socket for every transaction.   On a dual bank board, half the RAM is connected to two of the /RAS lines and the other half of the RAM is connected to the other two /RAS lines.   The computer selects between them by selectively activating the /RAS lines.  Otherwise all the signals between the two banks are common.

So, when a machine accesses a double-banked DIMM/SIMM as if it were single banked, it is actually performing all the reads and writes to both banks simultaneously.    This overloads the data bus on reads (a little) but has the bigger issue that tiny variations in process (manufacturing process) might cause the RAM chips to try to drive the data bus at different levels, the contention can cause all kinds of signal noise.  In practice, I think this rarely happens.

 

[jessenator]

If you're seeing the full RAM you should be fine. 

Thanks for the detailed explanation, trag. One day I hope to understand it all   

Is there a test, say in MacBench 4, that I could run it through to see if I'm overloading the data bus? Or for that matter, any test that I can use to give the RAM a workout?

I did forget an attempt at popping the 64MB DIMM into the 1st slot, so I'll try that tonight. If that doesn't work, boo-hoo < /s >  I think at any rate I want to replace that extra tall DIMM, so I'll probably end up getting 2x 64MB DIMMs from that ebay listing at the top. 160 isn't bad, but I'll see if 192 is possible.

 

[trag]

Newertech's RAMometer which is later subsumed in their Gauges set has a continuous memory test.   Run it as long/as many iterations as you like.   I think it must alter the data patterns from one run to the next, because back when I was buying new DIMMs, I'd run it on new memory and usually faulty stuff would fail within 10 - 15 iterations, but there was some stuff that only failed around the 1200 - 1500 iteration mark.   And it failed consistently in the same place.    Cell leakage in capacitive memory cells based on data patterns in surrounding cells is a whole other topic...

I keep a copy here:

https://www.prismnet.com/~trag/Ramometer.sea.hqx

and

https://www.prismnet.com/~trag/NewerTech/

or maybe:

https://www.prismnet.com/~trag/gaugepro1.1.sea.hqx

https://www.prismnet.com/~trag/gaugepro.hqx

 

[jessenator]

Sweet. I'll give em a look see tonight/this weekend

 

[omidimo]

I never upgraded the ram in my 4400, side effect of having too many Macs, but on a Beige G3 a similar problem exists, and the machine only saw half of the ram. I never had a problem with that anomaly though. 

 

[jessenator]

So I tried the 64Meg chip. Normal boot, only sees 32 MB total. But it didn't blow up at least

I'll probably just get two 64 mb sticks then.

 

[EvilCapitalist]

Is that RAM usage correct?  I didn't think 7.6.1 was that RAM hungry, or at least it hasn't been on the machines on which I've run it.  It almost looks like what happens when you boot into 24 bit mode with more than 8MB of RAM on an SE/30 but Power Macs can't boot into 24 bit mode.

 

[jessenator]

Is that RAM usage correct?  I didn't think 7.6.1 was that RAM hungry, or at least it hasn't been on the machines on which I've run it. 

Could it be the updates? http://main.system7today.com/updates/76x_powerpc.html  I don't have a reference to off of, really… I also have ATi drivers for the Rage128 installed. And I probably have a good deal of other extensions and such I could go turn off. As a kid I learned the hard way about turning off extensions whose purpose I didn't understand :lol:

It runs really well and I don't have shortages… further optimizing wouldn't hurt I suppose.