PowerMac 9600 startup problem

[Jaris]

Hi,

I have an old PowerMac 9600/300.  When it starts up for a first time it takes a long time before showing anything on the screen, then it plays a broken glass sound, and then everything seems to be OK. This problem doesn’t exist after restarting. I thought it may be a damaged RAM, but run TechTool and all 6 RAM memory sticks (64 each) passed the test, and they are recognised properly by the system. I tried also refreshing PRAM. It helps only once, meaning that when I shut down the computer, wait and power it again the problem comes back.

Do you have any idea what may cause this strange behaviour?

Thanks in advance!

Jaris

 

 

[novusgordo]

Put in a new PRAM battery and see what happens? The fact that it works perfectly on reboot suggests something like that.

 

[604ev]

Welcome to the Machine

 

[trag]

Replacing the battery is a good place to start. If that does not help, consider that tools that test the memory do not touch parts of the RAM at upper and lower address space, because the system is using those bits.

To thoroughly test the RAM put one DIMM in A (or B ) 1 and one DIMM IN A (or B ) 6 and however many other DIMMs you have in slots 2 - 5 of the same bank (A or B ). Now run your RAM test.

The problem at this point in the testing is that if it passes, you don't actually know if the RAM is good, because of the untested portion. However, the untested portion should be confined to the DIMMs in slots 1 and 6, so, if the test passes, you know that the DIMMs in the middle slot(s) are good. If the test fails it could be any of the installed DIMMs causing the failure, because portions of the outer DIMMs are involved in the tests.

Once you have found two confirmed good DIMMs, by passed tests while installed in middle slots, then you can use the good DIMMs in the outer slots as the bread of the test. Once you have known good DIMMs in the outer slots, then you know a failure is actually a bad DIMM in the middle slots.

If it passes, the RAM in the middle slots is probably good (did you read the Tech Tool manual about the different RAM tests and how long it would actually take to test *everything*?) but the RAM in slots 1 and 6 are still unknown. Move two of your now known good DIMMs to slots 1 and 6 and test the DIMMs previously in those spots in the middle (2 - 5) slots.

If it fails, rearrange your RAM and try again until it passes, possibly only putting one DIMM in the middle.

Once you have some DIMMs that passed the test in the middle slot(s) you can use these known good DIMMs as the outer DIMMs (slots 1 and 6) and from there on, if the test fails you know it was the middle DIMM that was bad.

We call this testing method the RAM Sandwich...

Personally, I prefer to run RAMometer (or RAM test in GuagePro, same thing) from NewerTech about 1600 iterations. It catches most RAM flaws within 15 iterations but the more subtle ones can take, in my experience, up to 1500 - 1600 iterations.

https://www.mail-archive.com/supermacs@mail.maclaunch.com/msg09225.html

 

[ziggy29]

Yep. The first thing I would do is start with the RAM, other tests aside, finding two sticks to put in the same interleaved bank that don't cause this (if that can be done). If so, there's probably bad RAM (or RAM sockets) and at that point it can be trial and error trying to locate the bad module(s) and/or socket(s).

Diagnostic software doesn't seem to identify RAM problems as well as the PM's startup sequence does.

 

[Jaris]

Thanks for all your comments. Unfortunately I didn’t have enough time to run all these tests. But I’ll try to do this during coming weekend. As for batteries I don’t have a new one for trying, so the only option would be to order it. On the other hand it doesn’t give an impression of being weak. Usually date and time is wrong then or something of this kind happens. But just to be sure I’ll buy a new one.

Meanwhile I checked again Apple System Profiler, and noticed that although it shows 6 DIMMs  interleaved, in location column slot B3 is missing. I mean it can see the memory stick is there but it doesn’t name the slot properly. Does it mean that the module in this slot is damaged or it’s not reliable?

As for RAM test I have only TechTool Pro 2.5.4. I run only default test as I can’t find any instructions for it. Default in this case means standard numerical test. There are other options though, but I’m not sure what to choose. There are other options like:  rotational, arpeggio, web, leap, minor march and major march. Also intensity can be standard or maximum.

On the other hand I was thinking whether hardware configuration may be the cause. There are 3 HDs and one extra graphic card (it’s Voodoo or something - didn’t pull it out yet). So maybe the SCSI chain isn’t OK, or PCI cards are not in proper order…dunno. In ASP control panel it shows the following configuration:

SCSI Bus 0 ID 0 - HD, ID 3 - CD-ROM

Bus 2 ID 1 - HD, ID 4 - HD

PCI Slot C1 - display card, Slot E 2 - SCSI card

The system is MacOS 9.1

Anyway I will run tests on this weekend, but if you have any other ideas please let me know.

Thanks!

 

[Jaris]

One more thought before I'll proceed with all those time consuming tests. If everything runs perfectly OK after rebooting (even RAM shows correctly under APS panel), what are the chances that the problem is caused by a damaged memory stick? Wouldn't the problem persist no matter what you do? What has rebooting to do with it?

 

[Compgeke]

Not necessarily, ram failures can be quite odd. Sometimes the ram will work perfectly fine for a while and other times fail right away. Whether that's heat or just ram being ram, who knows.

 

[trag]

Physically,RAM contains a multitude of little cells with electrical charge. Simple RAM flaws are when a bit is stuck at '1' or stuck at '0' and these are quickly detected by most RAM tests. However, there are also complex flaws in which a cell only fails when surrounded by a particular pattern of data in the surrounding cells. These flaws are quite difficult to test for exhaustively and are why RAM tests use changing data patterns.

In the complex case, the cells are not as well insulated from each other as they should be and the charge in some cells affects the charge in nearby cells.

I have copies of RAMometer and GuagePro here: https://www.prismnet.com/~trag/

 

[Jaris]

Sooo....I've tested all memory sticks using RAMometer (over 1600 iterations each) and they all passed. The chances are that the battery is not weak enough to cause major problems, but it is recognised during the startup sequence. Otherwise I don't know what else can cause such a strange behaviour, do you? I don't have a new battery yet, but should arrive in the end of the week. If you have any idea what else I could check, do let me know please.

 

[Jaris]

Now I have a brand new battery. Mac stopped playing this broken glass sound when booting, but it takes about a minute before the gray screen appears. I chuck out display prefs, reset PRAM again and pushed CUDA button. It helped, but only once. Next booting took very long again. It looks that it has something to do with some kind of display information which is being stored somewhere on the motherboard. Reseting PRAM helps temporarily, but a corrupted information is being remembered. My 6400 / 200 runs like a lightning when booting in comparison (in spite it has only 40 MB RAM). I don't know what else one could do. One other thought - there is no consensus as far as CUDA switch reseting is concerned. People describe it differently and I don't know which procedure is correct.

Anyway, I hope you have some other ideas...it would be very helpful.

Cheers 

 

[rsolberg]

The long delay on a cold boot is normal on systems with large amounts of RAM installed. The system doesn't initialise video output until the RAM test has completed. On a warm boot/reboot, such as when the PRAM is being reset, the system skips the RAM test. My Performa 6360 went from taking 5-10 seconds between the chime and video output with 40MB ram to about 90 seconds with 136MB installed. The smashing sound was the "death chime" indicating a hardware or configuration problem. I suspect it was a result of the amount of installed RAM changing- which the CUDA reset normally fixes.

 

[Jaris]

The long delay on a cold boot is normal on systems with large amounts of RAM installed. The system doesn't initialise video output until the RAM test has completed. On a warm boot/reboot, such as when the PRAM is being reset, the system skips the RAM test. My Performa 6360 went from taking 5-10 seconds between the chime and video output with 40MB ram to about 90 seconds with 136MB installed. The smashing sound was the "death chime" indicating a hardware or configuration problem. I suspect it was a result of the amount of installed RAM changing- which the CUDA reset normally fixes.

Hmm...you maybe right, but why it shows a grey screen almost immediately during first booting after PRAM was reset then (I am talking about the situation, when PRAM was reset, Mac shut down, and then started up again after some time, not after normal restart procedure)? Also I have a PM 8600 / 200 and it takes about 10 sek to boot with 128MB RAM. My Performa is fastest as I said but it may be due to small amount of RAM installed as you noticed. In general you may be right that there is a correlation between booting time and amount of RAM, but don't you think that waiting the whole minute to see only a grey screen (before loading extensions) is a little bit too long? Those machines had 12 RAM slots! If one filled them all with memory sticks it would take ages to boot, no?

 

[Compgeke]

A minute to grey screen is odd. I have my 9600 with 448 megs ram and even after a PRAM reset or cold boot the grey screen comes up in about 8 seconds. My Supermac S900 with a gig of ram and the same general architecture is only around 10 seconds and that was with no battery.

Here's a quick test: If you turn it on and it's taking a while to bring up the screen, if you turn it off after about 5 seconds and turn it back on again does the screen come up quick? I have noticed with some hardware changes (like internal SCSI drives) mine will need to be shut off then turned back on again before getting the quick graphics output. Never let it sit long enough to see if it comes up after a minute or so.

 

[Jaris]

Here's a quick test: If you turn it on and it's taking a while to bring up the screen, if you turn it off after about 5 seconds and turn it back on again does the screen come up quick? I have noticed with some hardware changes (like internal SCSI drives) mine will need to be shut off then turned back on again before getting the quick graphics output. Never let it sit long enough to see if it comes up after a minute or so.

The only way to do it when booting is to plug Mac out. And yes, it helps, but I can't do it each time. Anyway this is not a normal behaviour and there is a cause it doesn't work properly. I am trying to find it. 

Now, I am beginning to suspect the graphic card. It's Village Tronic MPDD, and there is no original Mac card inside. MPDD occupies PCI slot C1 at the moment, but I am not sure if this is correct. Do you know which slot is normally meant for a graphic card? This card is better than a standard Mac card (it has a big choise of available resolutions), but maybe it causes this delay at start up, who knows? Anyway it is not a matter of SCSI disks. I have 3 of them, and after unplugging 2 the problem still exists. Do you know how to check whether graphic card is the cause (other than taking this out and putting in another - I don't have an original one)?

 

[rsolberg]

Since replacing the PRAM battery, does the system keep time when the power cord is unplugged? If not, then you might have a faulty replacement battery or a bad connection. I just remembered that my experience with my Performa 6360 is likely based on the fact it has no PRAM battery and I don't leave it plugged in between uses.

 

[Jaris]

Yes, it does. The battery is brand new. I measured it with voltmeter and it's absolutely perfect.

 

[Jaris]

I tried another graphic card...same problem. Also I read service manual. I've run through all steps in case of display problems including reseting logic board. After this step checking monitor cable or monitor itself is advised. So now I am beginning to think that there is probably some kind of incompatibility between graphic card and modern monitor (it's Samsung 17''), or graphic card is not installed in a proper slot (I couldn't find information where it should be), in which case system may be looking in a wrong place first for a card and then finding my card after checking all PCI slots (since delay). It's just a hypothesis. So if any of you know anything about display problems with modern monitors, please let me know.

 

[Compgeke]

Never had issues with monitors and the PCI video cards for macs. Onto PCI slots, on the 9600 there're 6 slots and the upper three and lower three are on separate PCI buses. There wouldn't be a real reason to run through all the slots with the card, only one in each section for testing. In any case I've had video cards in both the upper and lower slots with no problems, various cards from Voodoo 3 2000 to Mac Radeon 9200. 

Here's another thing you can try: Mactest Pro. It does a bit more than just RAM tests and might throw something useful. The ram test will also tell you which specific slot and address has an error. Here's a link to that: http://archive.compgeke.com/Programs/Apple/MacTest/Mactest%20Pro%20PowerPC/

Edit: Picture of a bad ram error. It seems my own 9600 has an error I wasn't ware of.