mac mini GPU temperatures

[stevescivic]

I am wondering what others are seeing for GPU temperatures for their mid 2011 mac mini's running the i5 processor and the AMD radeon 6630M dedicated graphics GPU.

I've looked on the i-net for a while now and I can't seem to find the definitive on the AMD 6630M specifically.

The Mac Mini CPU as we all know it runs quite warm. I recently repasted my thermal grease with arctic silver 5 and botched it the first time. Too little on the GPU causing it to rise to almost 100C. I quickly shut it down and attempted a 2nd repaste

The 2nd time I applied the thermal grease and this time GPU hit 101C and the system shut itself down.

Now I'm left scratching my head wondering WTF did I do wrong. Thermal paste isn't supposed to be applied in gobbs and should be as thin as possible (but not so thin you can see exposed metal on the GPU or CPU). This time I opted to follow the ifixit application method using saran wrap on my finger to "dab" the paste on and I applied the GPU grease to be thicker (considerably thicker). No more GPU overheating but when I'm running Heaven 4.0 graphics test my GPU sits steadily between 79-81C at 1600 x 900 medium quality and no ambient occlusions.

Heaven 4.0 supposedly is quite graphics intensive and my CPU and GPU sit around the 80C mark with fans at about 2300 RPM. I ran this test for over 2 hours with the same consistent results.

Can anyone help chime in on what my temperatures should be? I was negligent and didn't note the GPU temps prior to repasting and only focused my efforts on CPU temps.

Fortunately my Mac mini survived 2 GPU overheats. Don't plan on overheating it again...

Some feedback would be great!

Thanks for your assistance,

 

[opinio]

I have used AS5 and Tuniq TX-4 on a 2011 (i5 and i7 quad). Had no problems. TX-4 is a little better with heat spikes and also is a little thicker than AS5. The temps run around 35'C on idle and 90-95'C when the CPU is maxed out.

I put paste of a small 'pea' size of one or two rice grains on the square GPU (about a 3mm ball) and a line about 2 mm thick along the rectangle CPU face. The line stops a few mm from the ends.

I do not use the spread method. It is no good because the heat sink surface is so poorley refined. I use the 'pea' method on the GPU and a 'line' method on the CPU.

I know the golden rule with paste is to use less, but because the heat sink is so bad (particularly on the 2012 mini) you need to use enough to fill/join the two surfaces. The 2012 i7 2.6GHz has about a 1mm gap between the sink and the GPU so if you use a spread method then there is still daylight betrween the two surfaces. There are spacers on the logic board that actually keep the gap at 1mm. It's incredible. After you do the paste you can shine a pen light in between the logic board and the sink to see the spread was reasonable or if it was too much. You might also want to use the light to check for a gap before you do the paste to see how well the suirfaces sit.

 

[stevescivic]

thermal paste

Yes I have read that tuniq TX-4 was better but I have a huge tube of AS5 so that is the choice of paste to use since it would be a huge waste to toss what I already own.

Thanks for the paste tip.

I will take it apart again tonight and clean the gpu/cpu again and then re-attempt to "repaste".

Yes now that you mention it there are standoffs on the logic board to heatsink so no matter how tight you screw the heat sink down it isn't going to make any tighter contact.

The only thing that I don't like too much about AS5 is the fact that it is slightly capacitive (not necessarily conductive) so it can theoretically short out the GPU or CPU should the paste spread to the little resistors and capacitors located just off from the heat sink surface. The factory paste clearly wasn't conductive b/c it was literally pasted and spread all over the place. It was messy and not done very well at all.

Just so I am clear then.

For the sake of simplicity of understanding:
1. Place a 3mm ball of AS5 paste in the middle of the GPU.
2. Place a thin line along the long side of the CPU stopping a few mm from the ends.
3. Bolt down heatsink and then use a flashlight to inspect spread and make sure it doesn't spill over to the electrical components to short the GPU/CPU out.

What if:
1. The paste doesn't quite spread and cover the entire mirror surface of the CPU or GPU?
2. The paste is uneven in it's thickness

Does "burn in or breaking in" of the paste really make a difference? I'm curious to know whether or not 200 hours is REALLY necessary as per AS5's instructions.

What is the max and average GPU temps you're seeing when playing graphically intensive content/applications/games?

CPU temps are a given that they're going to be 90+C. Seems like that is the way Apple has designed their systems.


Thanks!

 

[opinio]

The thin line on the CPU face goes up the middle length of the rectangle. Not the side. You can see the effects of the spread using this technique on you tube. It works far better than the normal spread.

It will spread when the sink is screwed down.

You might see a very slight over run of paste when looking between the sink and logic board with an led light. This shows that the spread is pretty much just right. When I say over run I don't mean squeezing out all over the place, I mean just a very slight line around where the sink surface and CPU/ GPU surface meet.

I really only measure CPU temps with istat menus. When using handbrake and the CPU is running at 100% I get about 90c on a 2011 i5, 95c on a 2011 i7 and 100 on a 2012 2.6 i7.

These temps are totally normal. Apple designs the mini this way. I have run the 2011 quad for two or so years now 24/7 doing many handbrake codes, sometimes for days on end.

No problems.

Personally I think the 200 hours does make a difference to AS5. Probably about 2-3c at peak temps.

What does make a huge difference to temps with older minis is cleaning the fan with a soft brush to get old dust off. Use a vacuum too. The dust weighs down the fan and effects the efficiency of the airflow. I see a 5c difference in peak temps after cleaning a dirty fan and outlet channels on the sink.

 

[MJL]

Arctic Silver is good for when the CPU / GPU closely fits against the heat sink. Tuniq-4 is good for when the heat sink is not closely fitting. Different animals.

Did you check how well the heat sink fits against the CPU / GPU? It is pathetic - it fits very, very badly and as such the Tuniq-4 is the stuff to use, no question about it.

I call this "penny wise and pound foolish", risking an expensive computer trying to save a few pennies for Tuniq-4.

There is also another issue you have not thought about with Arctic Silver: It is conductive. With the parts exposed at the side of the chip (and on the same sub-board as the chip) you do not want to use conductive thermal paste. Not always will it short out there but there will be some capacitive influence and may make functioning unstable. Longer term there may be some reaction with the solder flux that is encapsulating the parts and then you get some short and that's the end of it. You could have been a bit heavy handed or used some chemical cleaner and damaged/stripped the protective layer away from the solder joints.....

I just did the 2012 model - the chip next to the CPU/GPU (some controller chip) that needs to be cooled has a gap of at least 1mm - I suspect some thin thermal foam was initially specified but then that may have been too thick. Tuniq-4 is the only stuff to use there. (non conductive too)

 

[stevescivic]

GPU temps

So that being said, what are your guy's graphics processor temperatures looking like?

I am not concerned with my CPU temp, strictly GPU...

Thanks,

 

[opinio]

So that being said, what are your guy's graphics processor temperatures looking like?

I am not concerned with my CPU temp, strictly GPU...

Thanks,

What app are you using to measure GPU temp and what are you running to load it up?

 

[stevescivic]

GPU temps and the loading I use to test it

Mac Mini Specs:

i5 2.5 GHz dual core
8GB RAM
500GB Seagate Momentus XT hybrid drive
OSX 10.8.5
Video Test: Heaven 4.0
1600x1200
Ambient occlusion turned off
detail level medium

GPU temperatures typically hover between 79-82C

The question is: Is that too hot for the GPU and what are other folks seeing for temperatures on their AMD Radeon GPU in their mac mini.

Thanks!

 

[opinio]

Mac Mini Specs:

i5 2.5 GHz dual core
8GB RAM
500GB Seagate Momentus XT hybrid drive
OSX 10.8.5
Video Test: Heaven 4.0
1600x1200
Ambient occlusion turned off
detail level medium

GPU temperatures typically hover between 79-82C

The question is: Is that too hot for the GPU and what are other folks seeing for temperatures on their AMD Radeon GPU in their mac mini.

Thanks!

What are you using to measure the GPU temps.

 

[stevescivic]

measurement tools

Istat Pro and Temperature monitor