So, this is bad, right?

Dear Lazyweb,

My iMac (2008 24" 3.06 Core 2 Duo) has been making a lot more (which is to say, "any") fan noise for the last few days. Resetting NVRAM and SMU didn't change it. Vents are clean, room is cool, load is 0.17.

    25°C Ambient Air
    39°C CPU A Heatsink
    44°C CPU A Temperature Diode
    46°C Graphics Processor Chip 1
    44°C Graphics Processor Heatsink 1
    98°C Graphics Processor Temperature Diode
    58°C Hard Drive Bay 1
    43°C Memory Controller
    50°C Optical Drive
    41°C Power Supply Position 1
    59°C SMART Disk Hitachi HDS722020ALA330 (JK1130YAH94N1T)
    37°C Wireless Module
So, that's bad, right?
Tags: , , ,

28 Responses:

  1. pnendick says:

    Yeah, that ain't great.

    You could try to stave off death by arbitrarily amping the fans up a bit with smcFanControl

  2. g051051 says:

    98 and 59 don't look like good numbers. Everything I've seen says HD max should be between 55 and 60, and you're right on the edge. GPU seems really hot, unless you've got something running that you didn't mention. What program are you using to show those numbers? I'll check my iMac at work for a comparison. It's idle in a climate controlled lab at work.

  3. laptop006 says:

    Honestly, the drive's a little hot, but I'd bet that that GPU sensor's just gone, some googling does support that theory.

    Ah well, that's what applecare's for. (Not that I'd buy it, I usually break the warranty within the first five minutes)

    • hatter says:

      You know their warranty says, approximately "if you open the machine and break stuff, that's your problem; if you open it, close it, and something unrelated breaks, we'll fix it" ? Think this was partly as a result of some court cases where manufacturers got slapped for unreasonable behaviour, but I've only noticed apple actually mentioning it.

      the hatter

  4. enochsmiles says:

    One way to quickly confirm the "the GPU's heat sensor is dead" theory is to measure the temp of the GPU yourself; if you have one of those needle thermometers used for measuring the inside of an oven while cooking (the ones with the lead to a digital display...) you could affix the thermometer to the backside of the board with the GPU on it, close up the case, and see if your thermometer reading is within, say, 15°F of the diagnostic reading you're getting. If it winds up being somewhere in the 40°-65° range, my money's on the blown diode. At that point, you can fire up the soldering iron, or you can turn to Apple. (My choice would be made based on whether I had a warranty or not.)

    • skreidle says:

      If it can be measured with the case open (and line-of-sight), an IR thermometer would work too--pick one up at an auto parts store for $10-$15 if you don't have one already. Point from a distance, click, read.

    • supersat says:

      There are other GPU temperature sensors listed that are much lower.

      At 98°C, I doubt the GPU would even work.

      • fleshmeatstg says:

        You might be surprised at how hot things can get and continue working. When Firefox does that fun little thing where it turns into a zombie on quitting and consumes all my CPU time, my CPU A can get up to roughly 200°F. When I force-quit Firefox the temperature drops back down to normal and the machine seems none the worse for wear. Granted this has only happened overnight (quit Firefox, go to bed, wonder why my computer sounds like a jumbo jet when I walk in with my coffee the next morning) but my MBP seems to have not suffered any permanent damage from it.

        98°C comes out to about 208°F so I'd say it's within the realm of working-but-potentially-shortening-lifespan.

        I'll add that since everything else is much cooler I agree it's probably a dead sensor; I'm just saying that getting a chip that hot nowadays doesn't equal instant computer death.

        • Lots of semiconductors are rated to work fine at 100 C or hotter, though 85 C is more usual for something that doesn't have an automotive or military rating. It totally wouldn't surprise me if 98 C were in the "it'll probably pretty much work, but we can't guarantee it" range for a GPU.

          The bad-sensor theory seems like a likelier one though.

  5. enochsmiles says:

    (Also, you might boot to console in single-user mode and let it sit for a while; if the GPU temp is pushing 100°F even when you don't have any heavy graphics lifting going on (even Aqua or X), that'd be a pretty sure indicator that the sensor is bullshitting you.)

    ⌘-s will do it.

  6. This:

    46°C Graphics Processor Chip 1
    44°C Graphics Processor Heatsink 1
    98°C Graphics Processor Temperature Diode

    says your diode has gone insane and is telling the fans "HOLY SHIT I'M ABOUT TO BURST INTO FLAMES!"

    • neacal says:

      I agree. Two out if three sensors concur that there is no unnatural warming.

      The problem now is how to fix it. I'm not aware of any software that lets you change the way the sensor data is converted to fan RPM...

      So either, you can give the machine to Apple, to replace the diode (if they can do such things), or you can use SMC Fan Control to override the "natural" fan speed. The second option has the potential to cause your machine to fry, if it really does need high fan RPMs and you're not allowing that, the former option is based on the assumption that the diode is a discrete element, and not built in on the GPU.

      • jwz says:

        I downloaded smcFanControl 2.2.2 and it claimed to not work on my model of iMac.

        • It looks like the 24" iMac has the graphics on a MXM II daughtercard at least. So I would think that it shouldn't be too difficult or expensive to replace either through Apple or by doing it yourself. This is of course if the diode is more of a pain in the ass to replace than swapping out a card.

      • emeb says:

        As you suspected, the temp sense diodes are generally integrated onto the die of the chip in question. There's no replacing them.

  7. g_na says:

    When my Mac Pro's fan started making noise, it turned out it was the fan on the graphics card. It had clogged up with dust. Try checking/cleaning that if you haven't yet. (My card eventually died and had to be replaced.)

    • jwz says:

      I had the case open in Feb to upgrade the drive, and it was shockingly clean inside, so that's not it.

      It takes hours to disassemble/reassemble these things, because the back no longer comes off: you have to pull the glass and monitor to get to anything.

  8. ryanlrussell says:

    What are you using to dump the list of temperatures?

  9. gryazi says:

    Don't hold it that way.

  10. etfb says:

    You're living in one of the three remaining countries that don't use the metric system (the other two are the Independent Duchy of Grand Fenwick and the Demoblurgic Splorple's Filchpublic of Gibberia) so I feel I should warn you that 98 degrees is just 13.7 degrees short of one metric shitload, which in the imperial system is 3.7 milliArmageddons.

    HTH. HAND.