A Curious Problem with Electricity and an iMac: Updated

09/30/2011

I’ve had many problems with my recent 27″ iMac. First was a video problem, that was corrected with the 10.7.1 update, then came a problem with a burning smell coming from the computer. This smell only arises when I really push the iMac – converting videos with Handbrake, for example – and led to my contacting AppleCare for a solution. They had a repair center take the iMac, do tests, and, the first time, change the logic board, as the technician spotted a piece of scotch tape touching the processor.

After this repair, I got the iMac back and was happy. Until I converted another video; the smell occurred again. Back for repair, tests, yet nothing was found. So it was sent back to me.

All this time, I’ve been in touch with a level-2 AppleCare technician who has been very helpful, and who has tried to find a solution. But we’ve come across something extremely odd, and can’t figure out what could be the cause. Here’s the deal.

The burning is clearly caused by excessive temperatures. The iMac has a number of heat sensors, and I’ve recorded the temperatures when converting videos, to see how high they go. I won’t list all the temperatures, but rather the one that is the highest, the secondary heatsink. This is the inside of the back of the iMac – the aluminum back is a massive heatsink. In my tests, this sensor went up to 89 C. In the tests at the repair center, it didn’t exceed 66 C. And yesterday, I tested the same process – converting a video for an hour – in different rooms of my house.

First, in my office, it reached 89 C. Then I tried in my wife’s office, where it only hit 73 C. (That room was 2 degrees cooler, but those 2 degrees are not enough to make that much difference.) I then tested it in the living room, on the ground floor, where the temperature was the same as in my wife’s office. It hit 89 C.

So, the conundrum: what is causing this difference in temperature from one room of our house to another? I’ve ruled out voltage (I’ve tested in all rooms, during the tests, and it’s around 238 V.) It’s not peripherals, as the tests were done with nothing connected, only a wireless keyboard and Magic Touchpad. All the tests were done with the same power cable (and I’ve tried two different ones), with absolutely no variables other than the socket they were connected to. (All connected directly to the wall, not to power strips or UPSs.)

If anyone has an idea, I’d be grateful if you could post a comment. Thanks.

Update: I’ve found some very odd behavior on this iMac. The AppleCare technician asked my to do some tests using Snow Leopard. While I wasn’t able to get everything exactly the same (I booted off of an external USB drive), the discrepancies between Lion and Snow Leopard were more than surprising.

Using Handbrake for my tests, converting a .dvdmedia file, and using the same file both with Snow Leopard, then, later, with Lion, I found the following:

  • Temperatures were much lower with Snow Leopard than Lion. After an hour or so, the secondary heatsink was only 47, compared to as much as 89 C with Lion. There was a difference in temperature in my office; it was a few degrees cooler than during my tests the other day, but that is not where the difference lies.
  • Using the same file conversion with Handbrake, Snow Leopard draws between 16 and 17 W and between 16 and 17 A. With Lion, the same file, there was 36 W and 33 A being drawn.
  • With Snow Leopard, after an hour, Handbrake was converting the file at about 29 FPS. With Lion, Handbrake converts the file at 46 FPS. (Other files in my tests the other day were over 60 FPS, and I didn’t let the conversion run very long with Lion this morning; at times, the FPS rate increases during a conversion.)

Now, there are a few very small differences in my tests that would skew the results just slightly. My Lion tests today were with my standard account, so there were some background processes running. This would limit Handbrake’s access to the processors by a few percent, which could explain why I only got 46 FPS under Lion today compared to over 60 FPS the other day with a virgin account. But in all cases, the processor was being used to the max; all four cores were spiking, with total CPU usage around 390% or more at all times.

So, what does this mean? For some reason, Lion is using my processors much more; not only are they drawing twice as much power, in both watts and amps, but they are working harder (the FPS difference with Handbrake). I have an i5 processor which does not support hyperthreading, so that’s not how it’s working harder. It’s almost as if the iMac is overclocked with Lion. My thought is that this Mac was designed when Snow Leopard was available, so the temperatures and fan speeds are designed for that OS. But if Lion is doing more with the same processors, then the temperatures are naturally higher. It’s worth noting that there are a number of threads on Apple’s forums where people are experiencing higher temperatures with Lion than with Snow Leopard. I am curious to know what is going on; the temperatures I get when doing pretty much nothing (writing, web browsing, etc.) on Lion are higher than Snow Leopard temperatures when I’m maxing out the processors.

This does not explain my differences in temperature from one room to another; one friend suggested a grounding issue, which I’ll look into soon. But this difference in both temperature and performance is certainly surprising.