August 28th, 2006 • 4:08 pm
Unfortunately, I am still in the process of identifying which RAM module(s) is(are) causing my G5 Quad to crash on a regular basis. As indicated in my last post on the subject, my G5 Quad started freezing on a regular basis about 10 days ago, and I was quickly able to determine with near certainty that the problem was with my third-party RAM modules.
In addition to the two 256 MB RAM modules that came with the G5 Quad, I have four third-party 1 GB RAM modules. In the G5 Quad, RAM modules have to be installed in pairs, so I tried running with only two 1 GB modules in addition to the two 256 MB modules.
I got a freeze within 48 hours, which led me to suspect that one of those two modules was the culprit. I took them out of the machine and numbered them “1” and “2.” I then installed the other two modules, numbered “3” and “4.”
I ran three passes of the test suite. It took a long time, but eventually the installed RAM was given a clean bill of health. Just to make sure, I decided to continue running with these two modules for a few days.
Unfortunately, after three or four days of stability, I got a freeze while I was away from the computer, and then a kernel panic the next day while I was working at the computer. (It’s hard for me to tell whether the freeze was actually a kernel panic or not, because it happened while I was away from the computer and the monitors were asleep. When I came back to the machine, the fans were running at full speed and it was impossible to wake the machine up. As far as I can tell, these symptoms are the same whether the machine freezes or it has a kernel panic. In both cases, after a few moments, the fans kick in and stay on until you do a hard reset.)
What do I make of this? I really don’t know. Either I have more than one defective RAM module or the problem is not with the RAM modules themselves, but rather with the RAM bus or something.
Or maybe I didn’t spend enough time running the machine with only the 512 MB of RAM that came with it. When the machine started freezing, 10 days ago, I would get at least one freeze per day. So when I removed all four 1 GB modules and used the G5 Quad with only the 512 MB of RAM and didn’t get a freeze for a couple of days, I thought that was sufficient proof that the defect was in the 1 GB RAM modules. But now I am not so sure anymore.
So the bottom-line is that I have gone back to running the G5 Quad on only 512 MB of RAM. And I am going to stay that way for a few days.
Now, as indicated in my earlier post on the subject, running Mac OS X 10.4.7 on a G5 Quad with 512 MB of RAM was pretty painful the first time I tried it. I would get lots and lots of hard disk activity, and many operations were extremely slow, to the point that the machine was barely usable.
At the time, I speculated that the problem might be with the fact that I have two large monitors (a 30″ display and a 23″ display) connected to the G5 Quad. Even though the video card has enough RAM to support both displays, I thought that maybe the calculations involved in using such a dual-monitor set-up were still rather taxing on the system’s RAM as well.
Today, I spent about half a day trying to use the G5 Quad with 512 MB of RAM and then decided that it was pretty much intolerably slow, and decided to do some further testing. instead of suspecting the displays, I started suspecting the applications that I was running.
I am someone who does a lot of multi-tasking, especially since I am still on dial-up and Internet-related operations are painfully slow. I typically start loading a page in Safari or sending an e-mail with an attachment in Mail, and then switch to something else without waiting for the process to end, because, on dial-up, it’s a painfully slow process.
So I switch applications a lot, and I also tend to keep quite a few applications open at the same time: Mail, Safari, NetNewsWire, Speed Download, iTunes, Pages, Microsoft Word, Microsoft Excel, and a number of utilities. Given the RAM restrictions, I tried to reduce the number of applications open at the same time, but even with a reduced number, I would still experience enormous amounts of disk activity and huge waits when switching applications.
So I started suspecting specific applications, especially Mail and Safari. I tried using Camino instead of Safari, and it helped somewhat, but things were still very sluggish. And I simply cannot function without Mail open, so I didn’t have an option there.
But then I thought that maybe it wasn’t the applications, but rather the handful of utilities that I always have running in the background, including Default Folder X, Butler, DragThing, LaunchBar, and Spell Catcher X. The reason I started suspecting them is that switching applications with LaunchBar was particularly painful. It sometimes took a minute for the LaunchBar bar to drop down and register my keystrokes. It was effectively unusable.
So what I did was that I logged out. (I timed the process of logging out with all these applications open, and it timed out twice because both Mail and iTunes took far too long to quit. In total it actually took 10 minutes!.) And then I logged back in with the Shift key down, which prevented all the applications from starting automatically, except for Spell Catcher X, which launches anyway because it’s selected as the input method—and I cannot live without Spell Catcher X anyway.
Well guess what? Without the utilities, things are enormously better, even with several applications open at the same time. Right now, I have Mail, Safari, NetNewsWire, Speed Download, Pages, BBEdit, Word, and Spell Catcher X—and there is no background hard disk activity while I am typing this in BBEdit. I can easily switch from one application to the next without triggering a minute or two of constant hard drive activity.
So the conclusion here seems rather obvious: Mac OS X 10.4.7 is quite usable on a G5 Quad with only 512 MB of RAM, as long as you don’t run these little utilities in the background. I never noticed that the utilities were eating up many CPU cycles in the Activity Viewer, even when the machine was horrendously slow—but then, the problem was not with CPU cycles, it was with hard disk activity.
I do not know which of the utilities is the culprit here. Is it LaunchBar, Default Folder X, Butler, or DragThing? I might try launching them one at a time and see what happens. If I can isolate the problem that way, I will definitely post an update. But for now I will simply do without them. (I am obviously very relieved that Spell Catcher X is not the culprit, but then I never really suspected it, because it’s not typically a resource-hungry application.)
It is of course annoying to be without LaunchBar, Default Folder X, Butler, and DragThing. But it’s not the end of the world. I can survive without them for a few days if that’s what it takes.
All I can say is that, right now, my G5 Quad is quite usable, and I can reasonably consider using it with only 512 MB of RAM for a few days, which is really what I need to do to make sure that the problem is definitely with the third-party RAM modules.
If the G5 Quad does not freeze or have a kernel panic this week, then I will definitely conclude that the third-party modules are the culprits. But then I’ll have to come up with some testing procedure. memtest did not identify any problems with modules 3 and 4, and yet I still experienced two freezes/kernel panics. Of course, the documentation does say that RAM problems can be intermittent, which is why you have to run more than one pass of the test suite. But I ran three passes. Maybe I’ll have to run five. The trouble is that it is a very time-consuming process, and the machine makes a horrendous noise while running the tests. (All the fans are on in single-user mode.)
I have to face the reality that memtest might not be 100% reliable in detecting RAM problems. Or maybe there’s something wrong with some of my RAM slots… Unfortunately, the G5 manual says that you should install the RAM modules in pairs “from the center outward,” meaning that I have to use the same slots that I have been using since the beginning—unless this is only a recommendation for optimal performance and things might still work OK if I install the extra RAM in the other slots farther from the center. I might try this at some point, depending on how things go.
This whole process is rather painful. It is very time-consuming to try and identify the culprit here, yet I know very well that I have no other option. If third-party RAM is involved, then I have to eliminate that potential source of problems first before I can try to get any help from Apple.
So that means a few days with 512 MB of RAM only. Fortunately I have finally found a way to make things usable even with only that relatively small amount of RAM. The machine is not blindingly fast and I will be careful to avoid running several RAM-hungry applications at the same time. But at least I no longer have to wait for several minutes each time I try to switch applications!