Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

2008 Mac Pro Heat Issues?

TheRealPorkchop

TheRealPorkchop

macrumors regular
Original poster
Mar 13, 2010
100
1
Whiteville, NC
2008 Pro, dual 2.8ghz quad cores. Having an issue that I'm not certain what's causing it, hopefully someone can help me get started resolving this issue.

I can turn it on, it boots fine and makes it to the desktop. Within a few seconds, the screen goes black, it actually stops producing video and the monitor goes into sleep mode. Within a few more seconds, every fan in the Mac goes wide open. I can usually hold the power, shut it down, restart it immediately and it will usually work for an undetermined time after that.

Sometimes, it boots to the desktop without issue. I can put it to sleep or let it go to sleep on it's own, and have no issues. But sometimes when it goes to sleep on it's own, it'll do the exact same issue. Just all of a sudden the fans take off into high gear and there is no waking it or video. Other times it'll do it right in the middle of whatever I'm doing, YouTube or browsing or whatever.

I thought it was the nVidia 8800GT since those cards are problematic, so I put a 5770 in it, same thing. Both cards are PC cards, if that matters. I've tried flashing the 5770 but it always takes the flash fine but has a scrambled Apple logo screen and sends the computer into the fan wide open state. Flash it back to the PC version, and it works fine except no boot screen. So I even tried a EVGA 640GT, same thing with all three video cards and that kinda makes me think it's not a video card issue.

I've tried several combos with the RAM, nothing changes. Also tried different hard drive, unhooking the dual DVD drives. The Pro is like new on the inside, clean as a pin. I've removed the heat sinks, the fans, the ram cage, basically the entire inside. I built this last summer and didn't have this issue but now all of a sudden it's happening. So I removed all this again a few days ago, reapplied compound and put it all back together, same thing. I used Tuniq the first time, Arctic Silver the second time. Cleaned both processors and heatsinks with Arcticlean by Arctic Silver before every application.

Now... attached is a screenshot of my Temp Gauge readings off my 2006 Pro that I'm typing this from. This computer has the dual 3.0ghz quad cores, running Yosemite, blah blah blah. The CPU core temps are way lower than that of my 2008 Pro. I can't get it to boot and stay up long enough right now to get a screenshot of it's temps but I will ASAP. What are the normal temps? Anyone have any idea what's going on? The temps on the 2008 are in the 70's... really high I believe.

Screen Shot 2017-01-22 at 1.03.21 AM.png
 
If you have not already done so, disassemble the front fan unit to gain access to the heat sinks. Using a small airblower clean them out. it's amazing how much dust and fluff builds up in those. Also give the machine a general dedustbunnying.
 
I'll have to look to see what processors I used. I built this machine last summer and don't remember which ones exactly I used.

The inside is hospital clean, including the heat sinks. It looks like a brand new machine inside.
 
Is it possible to provide a temperature capture of the problematic 3,1? If possible, 2 captures, one for idle, and the other one under stress (CPU).
 
I would say clean your cpu cooling glue or whatever they called and put a new glue it help a lot on my old mac pro 2,1
 
Please read my original post again. I've used two different ones, getting the same issue regardless.

If I can get it to boot and stay up, a capture of the temps will be posted soon.



7dc7e370b4cd2ed988e177635058ebb4.jpg


8e36e4b9bd83dde0fa605970ae24986a.jpg
 
Last edited:
As you can see, it was clean. Like I said previously, I haven't long built this machine and it hasn't been used much. I am in the process of redoing the heatsink compound for the 3rd time now. I will post temp stats if it'll start and run long enough.

This is arctic silver, this is where I just pulled the sinks and processors.

4254c67776443e23681b5d0805f6c407.jpg
8e8f8989f14f49cf417774dd7e6d265b.jpg
 
I've tried a little and a lot. Neither way changes. This is the 3rd time and I tried a totally different compound this time. It took forever to get it to boot and remain useable for a few minutes. Here is the temps at startup and idling.

dce580822fb8c8f1acc00fb2d30d99e7.jpg


Here are the temps with the fans manually turned wide open. They do not fall, they've remained the same.

395a756453c0d880746d87d9f69f99b8.jpg
 
Actually it only looks like a lot, I spread it with a razor blade, it was actually thin. It looks like half a tube in the pics but it really isn't. I have tried more and less though, I see no difference in the core temps.

Is it possible to just have bad chips? Maybe I bought crappy processors?
 
Those are really high temps. The heatsinks don't seem to be doing their job very effectively. In the second picture, the heatsinks are at 24C and 25C while the CPUs are 70C-87C.
 
Yea they appear to be copper. For the past hour I've been sitting here using it, no crash. The temps remain the same.

Is it possible the problem is in the logic board?
 
Try reading the same temps using Macs Fan Control.

Intel's CPU die thermal sensors are not absolute sensors; they don't give a value from 0°C or 0K. They give a value of how many °C below max Junction Temperature(TJ max). The software that gives you an absolute °C must calculate that temp by subtracting TJ max from the sensor value. If it improperly calculates this, or if it has the wrong offsets(these vary greatly among Intel CPU models and lines over generations), you'll get ridiculous numbers.

Temp reading software is not to be trusted blindly.

Example, TJ max on an e5462 may be calibrated to 0°C offset at 85°C with a slope of 0.05°C negative measurement offset per 1°C and a zero point of 105°C. If your CPU is truly at 30°C, the sensor value will be something like 73. If the app is too dumb to calculate this properly and just spits out the sensor value, you see 73°C. That 73 value roughly means "your CPU is 73°C below overheating."

An easy way to test this is to see if the temperature drops when you put a load on the CPU. If the number drops as the CPU heats up, then the software is displaying the delta-TJ Max value. You can read that as "°C below max temp."

This paper explains it perfectly: https://pdfs.semanticscholar.org/c72e/ec75ada03c19580d569d52373ab1c8347544.pdf

Also, those sensors read TJ Max, which is often 20-30°C hotter than TCase Max(inside CPU die vs on top of heat spreader 'lid'). You can have 68-70°C TCase while TJ is 100-105°C, using Intel's reference thermal design. Both max temperatures are correct, just different temps. Think like a motor having 1500°F exhaust gas max and 230°F coolant max. The two are both correct maximums, but you have thermal resistance between the two. Another example, fire on stove burner may be 900°F right inside flame, but water in pot is 200°F.
 
Last edited:
  • Like
Reactions: h9826790
Screen Shot 2017-01-22 at 9.16.05 PM.png
I think we both kinda have the same problem but mine is not that high cause I'm editing 4k video while I'm monitoring the temp
 
Well Slash just blew my mind. My temps are scalding hot. I don't see them increase though regardless of what I'm doing. They're in the upper 80'sC which is crazy hot.

I'll try the different software. I personally am starting to think it's a hardware problem though.
 
Do something like handbrake transcode on very slow or placebo mode and see if the number decreases. If temp decreases under load then the software is trash.
[doublepost=1485143866][/doublepost]Any old gray paste should work acceptably. With stock clocks and voltage, anything higher $ than Arctic Silver 5 or Noctua's goop is a waste.

You're also using WAY too much thermal paste. Less is more in this case. You just want to fill the voids between the two metal surfaces. If you're doing more than that, the goop becomes an insulator. Copper and nickel-plated copper are many times better heat conductors than any paste. Having extra paste in the way is not helping.

Read this, or at least look at page 3: http://www.arcticsilver.com/pdf/appmeth/int/vl/intel_app_method_vertical_line_v1.1.pdf

Your RAM modules are the Apple-style ones with the giant heatsinks?
 
Last edited:
Ok, I got the Mac's Fan Control, attached the capture of what it says my temps are. Somewhat different from what Temperature Gauge says...

I've had it up and running since I posted the captures earlier. This Mac is outside in my HAM shack, so it doesn't see constant use. Just came outside and it's still up and running, the screen had been put to sleep, woke it up and it's still up and running. Odd.

Yes, the RAM has the Apple heatsinks on them. I've tried the thermal paste both ways, very little and too much. I'll tear it back down and redo it for the forth time, using much less this time. See if it'll boot back up afterwards and see what the temps are. Thanks!

Screen Shot 2017-01-22 at 11.15.06 PM.png
 
Yep. That other app is garbage.

You may still have another hardware issue causing your other problems, but it's not your CPU's overheating.

The core temps are terribly inaccurate at lower temps. They aren't meant to be used for what we're trying to do here. They're meant to indicate to the system a relative value below overheating. In that regard, 73°C below overheating vs 66°C below overheating doesn't really matter. They're fairly accurate as you get close to max temp; lower temps not so much.

There's a per-core offset as well that's registered inside the CPU, but not readable. Good luck with figuring that one out.

I've seen stuff as seemingly random as a cheap USB hub prevent waking from sleep. Unplug hub, and everything works again.
 
How is it possible to have those two temps showing lower than ambient? I understand that not all 8 cores are working during idle or surfing but still, how does it report those two as being so low?

If it's not an overheating issue, then what could it possibly be? I'd assume either a power supply issue or a logic board issue. I've removed the power supply and taken it apart, it had very little dust/dirt inside but I cleaned it anyway, the problem remained. I looked for bulged/swollen caps, none were found. Not that that means there aren't bad caps, I just don't see any that are obviously "bad".
 
The temp sensors are not accurate. They're not designed to be. They are only meant to detect approximate distance from overheating. When you're 70 from overheating, it doesn't mater if the sensor is off by 10 or even 20. They are calibrated to the max temp; when the CPU's get hot, they're more accurate in that range.

The machine is wanting to know ok, kinda warm, hot, danger, and overheating. The thermal controls don't care if it's "extremely ok" vs "ok." The on die sensors are meant for protecting the CPU from overheating. Intel never meant for end users to attempt to read them the way we're doing.

You should see how wacky they read when using phase-change cooling. I've run CPU's in PC systems at set points of -20°C. The on-die sensors will read anywhere from 10°C to -46°C all on the same CPU at the same time. Farther you get under max temp, the more inaccurate they are.

The sensors that Apple designs onto motherboard, etc are highly accurate thermocouples and give true, absolute readings. Those are generally within 0.5°C of reality. They are used by the system for fan control.

The on-die sensors are used to enable/disable thermal throttling, which is really an emergency mode to protect the hardware. Newer Intel chips also use those sensors to decide how far to go into turbo boost mode.

There's a slope to the sensor's error, where every 1°C below max you get the sensor is off by another few hundredths or tenths. When you're 50-80°C below max, that error can be as much as +/- 10°C.
 
Last edited:
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back