• Hello MLAers! We've re-enabled auto-approval for accounts. If you are still waiting on account approval, please check this thread for more information.

G5 adventures and the virtues of being able to tinker

On another thread, I wrote of my acquisition of a dual G5 2GHz PCI-X tower. "It" arrived yesterday, in the form of a dual 1.8GHz PCI tower with in-the-post case damage. The damage involved the feet being bent and the rivets at front and back being popped. A partial refund has been offered by the seller, which I am pondering. It would make the machine a very cheap purchase, even for parts (nearly as cheap as sending it back in the post, believe it or not, which in eBayland is at the buyer's expense).

Before accepting, however, I decided to see if the machine could be made to work.

The machine had problems, and turns out to have had very little use, judging by the lack of dust in the PS and on the heatsinks. I suspect that it had sat on a shelf or in a closet for some years. The problem with it was that one of the processors was misbehaving, resulting in the System putting the machine to sleep randomly and frequently due to a thermal runaway, with the processor core's temperature getting up close to 70ºC in near-instantaneous spikes. The fan was taking the usual G5 jumbo-jet take off run, and use of the requisite Apple utility (readily available on the web) only resulted in an error message to the effect that the software detected that the processor was out of thermal range, and that it needed to be replaced. It could not be thermally recalibrated. It was dead.

So, as the machine was toast anyway, I stripped it down and booted with one processor — worked fine. Booted with the other processor — worked fine (this has to be done in the upper slot). I then swapped the processors — worked fine. Each step required thermal recalibration of the processors, as a specific G5 processor is numerically tied to a specific slot in the machine, but the net result is that I reckon that the bad processor that had led to the machine being disused was just not seated properly — though there was no way of telling that by just looking. It is not at all difficult to do this, by the way, though you do need a half-decent toolkit to do the business.

Thus encouraged, and with both processors functioning more or less as Steve Jobs, the good Lord, and nature intended, I took another look at the core temperatures in Temperature Monitor and iStat, and noted that, though the machine was now working, one processor was often operating at 15ºC above the other. Out it came, to be taken apart, and cleaned up so that new Arctic Silver could be applied between processor and heatsink. Thus reconditioned, I put it back in, and it ran cooler than the "good" one! So I took "good" one out and did the same. The upshot is that I have balanced and very manageable temperatures in a machine that had been essentially non-functional. Cinebench, which uses a render in which both processors operate at 99% capacity, could scarcely make the fans spin up at all; iStat reported the core temperatures at approx, 49ºC. When I got the machine, by contrast, I could not even pop in a CD without the thing putting itself to sleep because of the thermal malfunction.

One additional effect of the final Arctic Silver treatment, by the way, is that benchmarking scores improved noticeably.

If I keep the thing, which I probably will now that I have it working, I will still have to repair the case, which could maybe be best done with clamps, some brute force, and judicious application of JB Weld. JB Weld is supposed to have a holding strength of something over 1000 psi, which ought to do the trick. The top handles are fine, so that would presumably give me a basis for fashioning some sort of wooden plug or mould that could be used to restore the shape of the feet.

A further frustration, mind you — as if I needed one in this purchase — is that I had sourced 8 sticks of 512MB PC3200 for the PCI-X machine (the machine, you will remember, that I thought I had bought). That 2.0 GHz model has 8 RAM slots; the 1.8GHz model has only 4 RAM slots. However, the guy I bought the RAM from locally also had numerous sticks of 1GB PC3200 available, so I'll see if he'll do a swap: 4GB of one for 4GB of the other, with $30-40 thrown in (1GB PC3200 RAM sticks are worth a good deal more than 512s). 4GB is, I believe, as much RAM as I would be ever needing in a G5, which is why I had bought the 8x512s in the first place.

So the moral of the story, children, is this: it pays to develop the capacity to tinker, and it pays to have the confidence to tinker. Tinker freely, therefore, because you never know when a dysfunctional G5 tower will come your way.

 
Nice conquest! :O

My friend just got a G5 with a bad PSU so I told him that instead of buying a $100 PSU for it, he could just wire it up with a ATX power supply. It always pays to tinker :)

 
A small update on repair of the G5 case, the feet of which had been bent and rivets popped in shipping. For the sake of posterity:

I briefly considered mechanical fastenings (screws), but thought I would try an adhesive in the first instance. For repair of the G5 case, therefore, I have been using a couple of pipe clamps, a crowbar, some scrap wood, and JB Weld. JB Weld is an epoxy with a metallic filler that comes in a standard two-part mix in small tubes, and that is readily available in most any hardware store in these parts.

To do the repair, the compound is mixed and spread with a knife or similar tool, and then the pieces pried into position and finally clamped together firmly. The clamps need to kept on for a minimum of 15 hrs in normal house temperatures; 24hrs is better. I have been doing the repair of the G5 with the side door in place, in order to ensure that it would be able to go into place once the epoxy had cured (those plates had been badly bent, so the side door, which was intact, amounted in my case to a template for the repair).

The forces involved in holding the bent metal together are far more than muscle power alone could manage, but the JB Weld does appear to hold, IF there is some purchase (something rough to hold) for it in the case parts rather than just smooth aluminum plate. Thus the rivets and the holes from which they were torn seem to be ideal; the JB Weld holds well there, and this is the area it ought to be mostly applied. Elsewhere on the case where there is no such purchase, e.g., at the seam of the cheese grater back and the side, it does not appear to hold (I has a small separation there and thought I would try it), but that is not the area that usually needs attention in G5 case damage.

The case, though not pristine, is now looking pretty darned good, given where I started; the JB Weld is a darker grey than the aluminum finish, and so has to be kept from view, but this is not hard to do. It can also be sanded/ cut away if any is visible on completion. Great stuff for these purposes.

 
Great work beachcove, can you post some pics of your repairs?

Used the same product (JB Weld) to repair a 12" PB a while back; the previous owner pried it open with a screwdriver to get access to the hard disk (normal people remove the 20+ screws and top case to gain access to it), once the metals were bent back it too came up nicely. Same goes with my Hackintosh in a G5 case - the JB Weld is proving a success mounting the entire PC motherboard.

I find for a reliable join, you still need to use a liberal amount around the join to gain greater surface area. On regions that need particularly strong holding power, making a small metal collar (eg. from a piece of bent aluminium) around the join helps. Otherwise, it will snap off cleanly within time.

JB

 
Back
Top