8600 bus error: possible causes

[beachycove]

My 8600/200 604e with 256k cache, which has an Adaptec scsi pci card and 9 GB 68-pin drive, gives a repeated "bus error" message on cold startup in 8.6. Essentially, it gets to the welcome screen and then crashes with the bus error message (freezing at that point, as the Restart button on the message doesn't work at that point). The Adaptec Control Panel is installed, as is Silverlining.

If I restart often enough, it runs perfectly and stably ("warmed up"?). If I start from cold with Extensions off, it also starts up fine.

What is the likely explanation?

 

[Charlieman]

Extensions Manager is your friend. Disable everything and then add it back in progressively. As a starting point, try turning off Virtual Memory (not just disabling the control panel).

 

[beachycove]

Turning off Virtual Memory seems to have cured it, for now anyway. Many thanks. I also turned off the memory check in the Memory Control Panel, which cured another problem whereby the machine occasionally stalled with a blank screen and needed the three finger salute to restart.

What is the deal with VM in this case? I have never had trouble with it on a ppc before.

 

[Dennis Nedry]

Turning off Virtual Memory seems to have cured it, for now anyway. Many thanks. I also turned off the memory check in the Memory Control Panel, which cured another problem whereby the machine occasionally stalled with a blank screen and needed the three finger salute to restart.
What is the deal with VM in this case? I have never had trouble with it on a ppc before.

It's probably an extension or control panel that crashes in the presence of virtual memory. By turning off virtual memory, you make the problem disappear.

Try disabling all control panels and extensions in extensions manager (Except Memory control panel if given the choice), then restarting with VM ON and see if that also cures the problem. If so, you can narrow down a different culprit and keep virtual memory.

If you haven't done this before, typically you add back half of them at a time. If it works, you know all of those are good and the problem lies in the other half that is still disabled. If it doesn't work, undo what you did and enable the other half. Then simply proceed to enable half of the ones that are left and repeat the process, dividing them all by 2 over and over until you figure it out. It's called a binary search, or in Mac techie terms, a "Split-Half Search".

 

[beachycove]

Adobe Acrobat 5, Photoshop 5, and Apple Color OneScanner software, among others, are on the machine, and this involves a goodly number of odd extensions required for their use. I have turned off all the base MacOS 8.6 Extensions (FireWire etc.) possible, but do not want to meddle further. The 8600 has nearly 300MB or RAM installed, so I will leave it without VM and hope for the best for now, methinks.

Thanks for the help. I had never before encountered a clear case of a conflict involving VM, though of course conflicts between Extensions are a commonplace known to us all. The upshot, however, is that I would never have thought to "Just say no" to VM. That is an important tip to squirrel away for future reference.

That sort of help/ information is why we hang around here, and in this case, is most appreciated.

 

[Dennis Nedry]

The 8600 has nearly 300MB or RAM installed, so I will leave it without VM and hope for the best for now, methinks.

Some PPC apps will run more efficiently with VM turned on, whether you actually use the extra memory or not. For example, if you had 256 MB RAM, and turned on VM to 257, you'd get little benefit from the extra 1 MB, but the increase in speed would make it desirable.

I'd say it's worth looking into an extension conflict if things start to seem slower than usual.

 

[beachycove]

I may investigate this further in due course, as I know that VM will speed up a ppc under many circumstances. Only not yet....

The particular 8600 in question here had previously given me trouble with a 2 x 604e, 200 MHz card in it (though from a different install of MacOS 8.6), and as I was having trouble with the single card, I thought the machine itself was possibly toast and would need to be replaced/ tossed. Having found that I could make it run stably with a 1 x 604e card in there merely by the turning off of VM, last night I replaced the single processor with the dual processor card that had been sitting idle.

Bliss. I don't want to change anything for a while.

Day to day applications are, of course, for the most part not able to take much advantage of the second processor (multimedia apps often do, however). This doesn't bother me much, because my day to day applications don't need a second processor. The kernel in 8.6, however, was rewritten to be more multiprocessor friendly in general, and I gather that the Finder in particular took greater advantage of the second processor.

Now I happen to have a PowerBook 2400c/180, which has a 603ev processor with 80-odd MB of RAM. It has MacOS 8.6 on it, and it is sitting alongside the dual 604e 8600/ 200, which also has MacOS 8.6 on it. I use the 2400c on a regular basis, so I know it well. Both machines have a 256 L2 cache, and both are of an era. However, if I try to do something like print or to copy a file on the 2400c, I'm effectively unable to do anything else until the first job is processed, the multitasking is so painfully poor. I have a lot of experience of the 604e and 604ev, and have always found them to be far better, and indeed, I prefer a 604 chip to a G3, in general. The dual 604e, however, is a revelation. Yes, there are inherent limits in the OS, but the dual 604e performs like I have never seen a classic Mac perform when doing several things at once.

I could make this my main machine....

 

[Dennis Nedry]

I'm not particularly familiar with multiple processors in pre-OS X, but here's my question. If you are multitasking, and have two separate, intensive programs running, wouldn't Mac OS automatically assign each program to its own processor? So even if the apps themselves aren't multithreaded, running more than one at a time could see a considerable benefit?

Or did it not work that way so long ago?

 

[trag]

My 8600/200 604e with 256k cache, which has an Adaptec scsi pci card and 9 GB 68-pin drive, gives a repeated "bus error" message on cold startup in 8.6. Essentially, it gets to the welcome screen and then crashes with the bus error message (freezing at that point, as the Restart button on the message doesn't work at that point). The Adaptec Control Panel is installed, as is Silverlining.
If I restart often enough, it runs perfectly and stably ("warmed up"?). If I start from cold with Extensions off, it also starts up fine.

What is the likely explanation?

Have you tested the RAM in the machine? The other suggestions are good ones, but a flawed RAM module is also consistent with the problems that you are having and would explain why changing VM settings affects the symptoms.

Dennis, don't you belong over on the Larry Niven group? What are you doing over here?

 

[Dennis Nedry]

Dennis, don't you belong over on the Larry Niven group? What are you doing over here?

I'm not sure exactly what that's supposed to mean. Are you just kidding about my username, or are you also telling me you'd rather I wasn't here?

 

[beachycove]

My 8600/200 604e with 256k cache, which has an Adaptec scsi pci card and 9 GB 68-pin drive, gives a repeated "bus error" message on cold startup in 8.6. Essentially, it gets to the welcome screen and then crashes with the bus error message (freezing at that point, as the Restart button on the message doesn't work at that point). The Adaptec Control Panel is installed, as is Silverlining.
If I restart often enough, it runs perfectly and stably ("warmed up"?). If I start from cold with Extensions off, it also starts up fine.

What is the likely explanation?

Have you tested the RAM in the machine? The other suggestions are good ones, but a flawed RAM module is also consistent with the problems that you are having and would explain why changing VM settings affects the symptoms.

I have tested the RAM using TekTool Pro v. 3, and though I had one failure, it has passed maybe a dozen times. The documentation says that there can be occasional failures and not to worry, but it seems to me that this would be worth pursuing through other utilities. I have some software options available and will try them out over the weekend, along with installation of different RAM sticks.

 

[trag]

Dennis, don't you belong over on the Larry Niven group? What are you doing over here?

I'm not sure exactly what that's supposed to mean. Are you just kidding about my username, or are you also telling me you'd rather I wasn't here?

No, it was a friendly joke/recognition because I thought I recognized your name as regular participant in the Larry Niven email group. However, my memory being what it is, perhaps I was mistaken. Most likely mistaken, as it sounds like you don't have any idea which email group I'm referring to. My apologies for the confusion.

Later edit: Nope, I checked. There is definitely a Dennis Nedry on the Larry Niven email list. That's not you?

 

[trag]

I have tested the RAM using TekTool Pro v. 3, and though I had one failure, it has passed maybe a dozen times. The documentation says that there can be occasional failures and not to worry, but it seems to me that this would be worth pursuing through other utilities. I have some software options available and will try them out over the weekend, along with installation of different RAM sticks.

I would recommend using RAMometer of the RAM testing facility in GaugePro which is basically also RAMometer. These were free utilities from NewerTech. They may still be available on their site. If not, I think I have a copy of one or the other or both on my web space: http://www.io.com/~trag or http://www.jeffwalther.com.

I also recommend you use a method called the RAM Sandwich. If you search on that term in the archives for LEMList's SuperMacs list you should find a (several) detailed description(s) of the procedure. The issue it solves is that portions of a DIMM are occupied by the operating system, so a normal memory test does not test all locations in all of your DIMMs. The RAM Sandwich method overcomes this problem. The last discussion was probably four or five years ago at a guess. I'm not sure if that was before or after LEMLists switched to Google Groups.

 

[beachycove]

I'm not particularly familiar with multiple processors in pre-OS X, but here's my question. If you are multitasking, and have two separate, intensive programs running, wouldn't Mac OS automatically assign each program to its own processor? So even if the apps themselves aren't multithreaded, running more than one at a time could see a considerable benefit?
Or did it not work that way so long ago?

I am not yet quite to the bottom of this, but it is a good question and it appears that the answer is that it did not work that way, unfortunately. There is only one primary processor per machine, to which applications are in the first instance assigned. Applications (e.g., Photoshop) which are then MP aware can use the second processor as well. That, of course, is also true in OSX (not sure about Snow Leopard), except that in X, the OS can allocate one program to one processor and another program to another processor. I think....

However, it appears that some of the operations of OS 8.6 itself are using both processors, as I get partial use from the second processor just from, say, opening a file folder. I do not, however, yet have a good way to measure how this works in action, beyond some fragmentary information provided by one of the (very basic) apps in the stock Multiprocessing Folder, which tell me that the second processor is working, say, at 15% or some such.

Not ideal, but it still feels snappy.

 

[beachycove]

I would recommend using RAMometer....

Very helpful discussion, and lead on the Gauge Utilities, which I have used for many years but without knowing this particular virtue. Gauge's Ramometer identified a problem within seconds, and was able to find at least 50 more before I gave up counting and quit the test.

Given that the error is a gross one rather than the kind of subtle one that would take all night to find, I think I will try a quick swap of a spare 32MB chip for the six or so I have in there, one by one, running the utility each time. The RAM Sandwich looks a good deal more substantial, but less inviting.

 

[beachycove]

Confirmed: Bad RAM.