Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[aBc]

Think I’ve another 2003 MDD J20 RAM slot problem here.

This 1.25 GHz SP MDD, loaded with four 512 MB RAM sticks would only occasionally fully boot from a cold start before freezing and then NOT completely restart (or fully boot) before immediately freezing. And by now don’t we all know to remove all but one stick of RAM and then test and retest with each stick of RAM alone, in order to possibly determine if there’s a bad stick. Right?

Well how many of us (like myself) usually test in this manner, leaving (or placing) each individual stick in the MDD’s J20 RAM slot? (In my case, thinking that the J20 slot is the first memory slot addressed by the machine.) So, when testing this machine in this manner… none of the four sticks individually would allow the MDD to fully boot before freezing. Was all of the RAM faulty?

NO.

Placing any of the four sticks individually, in other than the J20 slot… would then allow the MDD to fully boot. Seems like this is more than just a bit familiar to me.

See: https://macos9lives.com/smforum/index.php?topic=6285.msg47109#msg47109

Now I do wonder just how many MDDs might have somehow ended up e-wasted or otherwise possibly abandoned… simply because of this expanded  J20 RAM slot condition?

             

Sure, I could run just three, 512 MB sticks in the J21, J22 and J23 slots for a total of 1.5 GB... or try a 1 GB stick + two 512 MB sticks for 2 GB — and leave the J20 slot empty. (Or so I have been told.) Or I could even swap out the mobo with another mobo, without this J20 slot problem? But still, I wonder what is the root cause of this problem?

RAM placed in the J20 slot here seems to be a bit more loose in the slot and as mentioned in the link above… but this was remedied before with a slightly “thicker” stick of RAM that seems to now have solved that problem since 2022. (0.003 inch thicker!) And as much as I swap RAM in and out of machines I first thought that the “loose” J20 slot was a result of possible “wear”. But now I wonder if it might be a consequence of its location near the GPU and all of that heat emanating from that GPU heatsink over time? And perhaps especially… if the GPU has also been powering an Apple ADC monitor for most of its life?

GPU pictured above has a fan, whereas this MDD and the previous one here did not. Lot of cumulative running heat cycles to consider since 2003. And I found no weakened or faulty J20 solder joints back in 2022 with the other machine. So I assume that some heat related expansion of this slot (over time) might actually be the culprit?

I may swap out the mobo and retest all over again. And if that works, Mardeec might be getting this MDD back to use — or to possibly even sell on good ol’ flea-bay. Can anyone resist having at least one of these MDDs, especially one like this in near perfect condition? I think Mardeec’s original $1.00 offer to Forum members for this machine is no longer in effect.

https://macos9lives.com/smforum/index.php?topic=7630.msg59104#msg59104

Anyone running a 1 GB RAM stick in an MDD (or Quicksilver)? Do speak up please.
UPDATE:
Found a loose, thicker (0.053 inch) 256 MB stick of RAM here last night and inserted that into the J20 slot, which brought the RAM total up to 1.75 GB. It then booted and ran flawlessly, playing music CDs for 3 hours without fail. So… swap mobos or just accept the 1.75 GB as good enough (for now)?

[chrisNova777]

according to the numbering of the slots.. it would seem that the ram is intended to be installed in matched pairs, the two outer slots are one channel and the two inner slots are another channel, if its dual channel memory thats how the G5 memory has to be populated... i didnt know the mdd was the same way but according to that number scheme that would be what that means to me logically from lookin at it

[aBc]

Thanks Chris, I hadn’t considered that. It does seem to be working stable at present (even with all of the mis-matched RAM and the 256 MB stick now added in the J20 slot).

Seems that info about this is currently sparse and even posts in the “Apple Community” vary on the use of 1 GB sticks in the MDD. AND Google’s AI results seem to offer a bit of conflicting / confusing info about 1 GB RAM.

First, from the Apple Community (and Japamac’s response).
https://discussions.apple.com/thread/1307763?sortBy=rank (12/26/2007) 

Then… three different current screenshots from Google’s AI - also attached below.

*I really believe that the J20 slot has expanded over time.

[DieHard]

according to the numbering of the slots.. it would seem that the ram is intended to be installed in matched pairs, the two outer slots are one channel and the two inner slots are another channel, if its dual channel memory thats how the G5 memory has to be populated... i didnt know the mdd was the same way but according to that number scheme that would be what that means to me logically from lookin at it

No totally different, not necessary to install in pairs, however AFAIR the MDD supported Memory Interleaving, so in theory, you should get better performance with matching pairs; neither configuration should introduce errors; also, it's a good idea to hit each slot with a crazy amount on canned electronic cleaner and also inspect tiny RAM claws to find any stuck any not sitting at the optimal height

[refinery]

MDDs will run fine with two 1gb sticks. I have been doing this for years to allow for a quiet heatsink on my Radeon 9800

[aBc]

Hey @refinery, thanks!

Which two RAM slots are your 1 GB sticks in?
Do we consider the J20 slot as number one?



OR… is J21 considered Slot 1?
[In which case, perhaps J23 is considered to be the number 3 slot.]

         



Again, it’s the J20 slot which seems to be the problem w/ this machine.
Mardeec had thoroughly cleaned it - and I did it again, twice afterwards.


Und NOTE:
For all that previously used the low-cost, CRC brand Electronic Cleaner,
that cost has risen from less than (or near) $5.00 — now to $8.48  USD.
*And Walmart now only offers it via online order.

[refinery]

I use J21 and J22, the two closest to the CPU heatsink. It provided the most clearance for the GPU heatsink. I actually had to remove the locking clips on J20 for it to fit right.
I know I probably sacrificed a little performance by not using interleaving, but having a quiet system was more important to me.

[aBc]

Thought there might be a better approach to test and compare… but QuickBench 2.0 did reveal a very major difference between two interleaved 1GB RAM modules and then same modules not interleaved. Such a difference that I find it very difficult to believe, even after testing twice. So I have ordered another matched pair of 1GB RAM to test all over again.

The first column is from Mardeec’s 1.25 GHz DP MDD (with the questionable J20 RAM slot). I found a “thicker” 256 MB stick of RAM that works in that slot, so all four slots were filled. Then (same machine with same SSD) was tested with and without interleaving. I did expect results to all be much closer. But if these results stand after another test… well there is a definte advantage to interleaving.



If the image above is difficult to see clearly you might try double-clicking the attachment itself, below. OR change your Forum theme to “Blu” temporarily. (You can change it back afterwards.) *Yes, if you’re using the Forum or Board Default theme OR the Mac OS 9 Lives 3.0 theme, it might be best to just drag the image file to your desktop and open it there.

The big blue numbers represent average of each individual test panel. Red numbers at the bottom are the overall averages of all three panels, of each individual test / configuration, in order to provide an overall one-number “rating”.

[*The first 1 KByte Transfer Size results were omitted, as those results often are noted in KB/sec.]

AND if testing with the different, newly ordered 1GB modules doesn’t provide much difference… there won’t be another test update from me on this topic. (Did I just hear a collective sigh?)

But really, a nearly one-third drop in overall performance?

[refinery]

That kind of doesn't make sense because 3x 512 + 1x 256 isn't an interleaved configuration yet the results are similar to the 1gb interleaved.
I will have some time over the next few days to test mine and I'll report back

[aBc]

That kind of doesn't make sense because 3x 512 + 1x 256 isn't an interleaved configuration yet the results are similar to the 1gb interleaved.
I will have some time over the next few days to test mine and I'll report back

Thanks refinery. I agree that it doesn't seem to make sense. Maybe it is because OS 9 only utilizes 1.5GB and as long as that’s available [especially when testing with QuickBench Classic (2.0) which IS OS 9 based] interleaving might not matter. Yet when one uses only the (2) 1GB sticks, interleaving can improve performance? Perhaps just as matched pairs of (4) 512MB can do?

The three 512MB sticks that I used (in J21, J22 and J23) were all matched and the same. The 256 of course, wasn't. And of course, depending upon just how your machine is configured or loaded up. This machine here is a bare bones sort of machine with a 128 GB SSD and no other internal cards, other than the GPU.

Before you get way out here in-the-weeds with me… if you’ve room to move the 1GB RAM in J22 to J23, you might then test with QuickBench Classic [2.0]. After moving the RAM to J23, boot up and zap the PRAM and then rebuild the desktop prior to running QuickBench. Otherwise you may only see results like what you might get from slots J21 & J22.

I just tested 2 different, new pairs of 1GB RAM in J21 and J23 and at first they all tested like the previous, non-interleaved results. I zapped the PRAM and rebuilt the desktop and retested after each pair change of RAM. QuickBench results then matched the previous interleaved results.

*QuickBench 4.0 runs only under OS X and thus, likely tests the full 2GB of RAM under OS X?


Also… just ran Apple Service Diagnostic (ASD v2.1.5 for the 2003 MDD) with one pair of the new 1GB RAM sticks installed and it PASSED every test. Like a big dummy... I didn’t select specific tests and it took over 35 minutes to complete ALL of its tests. I don’t like that I can’t simply screenshot the results & text - or all of the tests that ASD runs — but here’s a somewhat “reconstructed” iPhone image after only 15 minutes of runtime (it says 15 minutes, anyway).



Really like ASD v2.1.5 and wonder how it might test out with the original pair of 1GB RAM. (But not tonight.)


So again, likely more information than anyone really wants to know…

If a questionable J20 RAM slot wasn’t already enough to contend with here… a much closer inspection of the original (2) 1 GB modules provided yet another concern.



See that “Big S” on module #2? It’s in slot J23. But when I swap in into J21 — and module #1 into J23… I get a “Finder failure” message upon boot and reboot. (I’ll attempt to attach a SpecTek.pdf file below.) Swap the sticks back and no problem. Hmm?

And then upon a much closer examination of #2’s backside, what’s that? Missing TWO little capacitors? Yet it still seems to work fine, just as long as it remains in the J23 slot - with module #1 in the J21 slot.



Okay, that’s defintely more than enough for now!

[peeperpc]

For a long time, my MDD has problem at J20 and J23. So I configured it to use two 1GB memory sticks at the J21 and J22.

Last month my MDD couldn't boot. It kept beeping with power switch's light flashing which means no RAM detected. After trying to find out if I could run it with just 1 slot. The problem got worse. The fan would blow at full blast after turning on.

In my case, I suspect the problem is cracked solder joints under the RAM slots caused by GeForce4 MX video card. I think MDD owners with this card know how difficult it is to get the card's bracket's screw hole align with the chassis so that you can put the screw in. It's like the left-end of the card has to pull/lift the main board up somewhat to make it. Then the warp on the main board put stress on the solder joints of the RAM slots.

Just my theory.

For now, I disconnect the power to the main board and use the PSU to only power hard drives in there and use them through SATA to USB adapters.

[GaryN]

As a long-time MDD user, I have a little bit of perspective here.
I wish now that I had invested in Invidia then since I would be stinking rich today.
However, after with watching their video cards fail over and over I never thought they would become a leader at anything and certainly not the giant they are now.

Most (80-90%?) of those damn GeForce cards failed from long-term overheating. Those little fans they put on them were NOT just for show, rather they are the only thing that keeps those chips connected to the card without it raining little solder balls everywhere.

Unfortunately, in the MDD, they exhaust all that heat directly onto the RAM chips and slots. That can and does cause the same bad connection debacle on those as well.

Often the problem is/was exacerbated by MDD owners with self-proclaimed thermal engineering expertise who are certain there was no need for those noisy fans that Apple installed and so they replaced them with "modern" quieter ones without any real heat-probe testing and evaluation of the results. If the ambient temperature inside the chassis goes up, so does the air being moved by the GPU fan since it can only draw ambient air from around it and exhaust it, sadly, right into the RAM sticks.

The real issue here is that these MDDs were pushing the envelope when they were new and they're now 20-odd years old. All of those delicate solder connections have been stressed over and over until they're now barely hanging on.*

I run my studio with my MDD. I made certain to buy up enough units and parts back when they were really cheap to be able to repair any failures that should arise. I probably still have 4 complete computers and uncounted miscellaneous HDDs and various hardware on the shelf.

What I do NOT have is a single reliable working GeForce card. I DO have a couple of ATI 9600 G5 cards but I tried them and I don't use them anymore. I realized long ago that, even running two large monitors (even Cinema Displays) I could see NO significant difference in the performance of a 128Mb card vs. a 256Mb and there was simply no good reason to push that limit for no good result. That included Photoshop editing and such.
(To the 3 or 4 users reading this who are heavily into vintage gaming and/or animation, I understand that you may actually SEE and appreciate a visible difference and it's not necessary to remind me that there's always someone living on the bleeding edge even with obsolete hardware.)

So, all those who "can't understand" why their 20+-year-old computers won't perform like new, don't ever forget that you almost certainly bought that computer used and it was not only second-hand, but possibly third, fourth or …? hand. You don't know where it's been and/or how hard it's been ridden. You must do whatever you can to ease the stress on these old-timers and paramount among that is to keep them as cool as possible.

*This is where you begin to understand and appreciate the parameter MTBF.

[aBc]

Once again, GaryN’s silver hammer!

The ONLY thing that I might even remotely question is the “3-4 users” estimate of those “heavily into gaming and or animation”. Thinking that there are far more gamers around here now than we might suspect.

My Tiny Brain Fails = MTBF? (I know, I know.)

DIfficult post to follow - but I already had one ready to go… below.

And now, not unlike GaryN… there are 5 fully functioning MDDs here.
(We won’t count all of the contingency, parts machines.)

Onward!

[aBc]

Read GaryN’s post here… https://macos9lives.com/smforum/index.php?topic=7637.msg59356#msg59356


"Curiouser and curiouser!" -Lewis Carroll

After last Wednesday night’s post, ran another quick series of tests. This time using the original RAM (as pictured above) and using Apple Hardware Test (AHT 1.2.7) because it’s faster than the extensive, full-gamut ASD 2.1.5  tests.

Anyway… first with the Kingston Elixir-Chipped module in slot J21 and the SpecTek module in J22…



Then tested with the Kingston SpecTek stick in J21 and the Elixir in J22 (seen below) … even though I thought the SpecTek stick was inferior (missing capacitors?) as it would not boot the machine without the Finder failure message — the Elixir then still reported bad in the J22 slot while the SpecTek (in J21) did not. Maybe attempt to boot with only the SpecTek alone, to see if that Finder Failure error still pops up? (It did not!) Tested the SpecTek 1GB stick alone in J21 slot and surprisingly, it tested good. Rebooted with that single stick and no more Finder failure message upon cold boot. SO… the Elixir-Chipped stick was problematic!

Finally decided to use the 512 + 1GB + 512 (J21 J22 J23) configuration… using the matched, interleaved approach. With the two matched 512’s and it now seems to boot even faster! Also ran QuickBench 2.0.
   



So now I’ll run this MDD for a while, configured just as it is.


“For every complex problem there is an answer that is clear, simple, and wrong.” - H.L. Mencken

While I was confident that this particular 1.25 GHz SP 2003 MDD had / has an expanded (or otherwise compromised) J20 slot, I now justifiably question the two original 1GB sticks of RAM as well… because over time, they too can fail … and perhaps fail in such a gradual manner as to possibly go undetected? It was most definitely a surprise to me here with this machine, that one stick tested bad, while the other that I had suspected of being bad, was not (yet).

Certainly I understand peeperpc’s theory of the torque-twisted + heated mobo, yet I also wonder if maybe a new stick or RAM might allow his MDD to boot and then possibly run AHT or ASD… in order to then test his original RAM (and ma in various slots? Even trying a single 512MB module (in various slots) might be worth trying if it would boot? AND maybe dremel out that GeForce4 MX video card screw hole slot just a bit to take some tension off of the mobo?

ASD will allow testing beyond an encountered error, while AHT stops with an ERROR CODE (but that might only be when running the AHT Quick Test). AND again, make certain to download, burn and run the appropriate versions of ASD and AHT for your specific machine, if you can. (There are two different versions of AHT & ASD for the MDDs.)

Thank you very much peeperpc… for your post and its contribution to this topic. I was beginning to wonder about just how many MDDs were still “out there”. There’s still a loose, untested FW400 mobo here if you might be interested in that (for shipping / mailing cost alone. PM me.) It’s doubtful that I will use it here to replace this one.

It cannot be stressed enough… the importance and use of AHT and / or ASD, whenever & wherever they can be used.

*AND with all of this in mind, I do wonder if some of the failed, only gray-screened, single RAM slotted G4 Mac minis (recently, such as Bolkonski’s) might have a failed 1GB RAM module… as opposed to say a possible GPU or CPU failure? *There’s also two different versions of AHT and ASD for the G4 Mac Minis.

Good night Irene.