Random Stuff

MetaTrader History ToolKit
My ATC2012 EA 
Heinz Engine
MetaResources 
Honda Steed Exhaust Mod
ResMagik
Frypan-Lamp Biquad Antenna
HTPC 001
Aerospace AL CPU Cooler 
PHP Pager
de-db 
Developer's Notes

Aerospace Aluminium CPU Cooler

Using RAW aerospace grade aluminium blocks as CPU heatsinks

In this new adventure, we (yes, you and me, you're part of the team) are going to explore and experiment with high quality aluminium blocks and we will determine if the material is good enough to be used as a passive CPU heatsink.

The materials in hand are a couple of solid blocks made out of high quality, aerospace grade (aeronautical and aircraft industries) aluminium alloy. I do not know which exact alloy they are: 7075, 2024 or 6061. I've been told that this aluminium is quite expensive and harder to get than other more common and domestic alloys found everywhere. These blocks are a gift from my dad when he worked many years ago in the aeronautical industry. These blocks are just leftovers from aircraft parts manufacturing and were going to be disposed.

We have two blocks

Block Nº1: The first one has a big planar area but has a slimmer body, perfect for low profile situations. This one has been machine cut and treated with precision tools to my own specifications and has that shiny finish, except for the bottom face that was left untouched to keep its brute flat surface from the foundry, this will maximize contact with the CPU heat spreader.

Dimensions:

  • Length: 8,75 cm
  • Width: 7,5 cm
  • Height: 2,9 cm

Block Nº2: The second block has a smaller planar area but has a thicker body and it is longer, it accounts for big mass. This one is untouched, it has its original height, surface and texture from the foundry. Its shape has been obtained only by machine sawing the brute aluminium wafer. No shiny finish.

Dimensions:

  • Length: 9,3 cm
  • Width: 6,6 cm
  • Height: 4,5 cm

And here is a picture of one next to the other so you can see their space ratio:

Both pieces do not have fins, this will theorically decrease cooling and heat dissipation performance, but they have one property that motivated me to make this experiment: Blocks are hard cold to the touch, all the time.

Additionally we have an auxiliary CPU fan. This is a Coolermaster, 60 mm, ball bearing fan capable of spinning up to 6.000 rpm (it hits quite a punch of airflow). We will include this fan in the benchmarks to get extra data with the help of active cooling.

Tests

We are going to do the following tests:

  1. CPU stress temps with default cooling solution (active).
  2. CPU stress temps with block Nº1 (passive).
  3. CPU stress temps with block Nº1 + fan (active).
  4. CPU stress temps with block Nº2 (passive).
  5. CPU stress temps with block Nº2 + fan (active).
  6. CPU stress temps with ultimate cooling taboo: Block Nº1 + block Nº2 + fan (hybrid).

Test bench

Our victim will be an old AM2 platform. The CPU is an AMD Sempron LE-1250, single core capable of reaching 2,2 Ghz but more importantly is rated at 45w TDP. The whole system is laid out in open air on top of a workdesk. Room temperature is about 24ºC. The motherboard is a custom Biostar one made for prebuilt systems. Only one stick of 2 GiB of DDR2 RAM will be used. No discrete graphics, just integrated one with the motherboard. Single HDD with a Windows 10 installation. Temps will be measured with HW Monitor and CPU will be put to use using CPU-Z's built in stress test. Nothing too fancy.

Test Nº1

CPU stress temps with default cooling solution (active).

We startup the system from cold boot and let it idle for a couple of minutes. Then we record its temp to get a reference:

Now we stress the CPU for 5 minutes and then we record its temp:

Tests results:

Test # Description Min Temp Max Temp
1 Default cooling solution (active) 26ºC 39ºC

Preparing the custom cooling solution

To proceed with the next series of tests, we need to remove the plastic bracket off of the standard cooling solution. This will leave the motherboard area around the processor cleared and our aluminium blocks will have more than enough space to fit on top of the CPU.

Test Nº2

CPU stress temps with block Nº1 (passive).

Of course we pre applied some thermal paste (Arctic Silver 5) for better contact and heat transfer. Again, we power on the system, wait for it to sit idle and carefully monitor its temp:

Once we are sure that the system is stable and that the experiment is viable, we stress the CPU for 5 minutes and then we record its temp:

The block remained cool for the first minute but became warm in time, up to the point it became scary hot to the touch. However the system behave pretty well and stable, we didn't notice anything unusual, the aluminium block did its job well.

Tests results:

Test # Description Min Temp Max Temp
1 Default cooling solution (active) 26ºC 39ºC
2 Block Nº1 (passive) 33ºC 51ºC

Test Nº3

CPU stress temps with block Nº1 + fan (active).

Before starting this test we had to shut down the system and wait for the aluminium block to cool down. Close to an hour later the block was still warm and concluded that it would take a lot more time for it to get back to its natural coldness. So, we decided to start the test with the block half-way warmed up as it was. This resulted in slightly higher idle temps, which in theory should be lower than those of the previous test:

And here are the further 5 minutes stress test results:

During the test, the temp normalized up with its maximum for the new configuration which indeed resulted in better values with the help of active cooling.

Tests results:

Test # Description Min Temp Max Temp
1 Default cooling solution (active) 26ºC 39ºC
2 Block Nº1 (passive) 33ºC 51ºC
3 Block Nº1 + fan (active) 32ºC 46ºC

Test Nº4

CPU stress temps with block Nº2 (passive).

For this test we waited for the whole system to cool down, removed the first block, reapplied thermal paste and squashed the second block onto the CPU. The blocks have different shape but have similar volume, the results should be around the same level:

The second block seems to be bigger in mass and that seems to help just by a couple of degrees. Let's see how it performs in the stress test:

The idle temps were somewhat better than the first block but that seems to have been a random measurement condition because when stressed, the maximum temp aligns tighter between the two blocks.

Tests results:

Test # Description Min Temp Max Temp
1 Default cooling solution (active) 26ºC 39ºC
2 Block Nº1 (passive) 33ºC 51ºC
3 Block Nº1 + fan (active) 32ºC 46ºC
4 Block Nº2 (passive) 27ºC 50ºC

Test Nº5

CPU stress temps with block Nº2 + fan (active).

Just like with the first block, the second one took a long time to cooldown, so after more than 30 minutes of waiting we again have decided to start this test with the block already warmed up. Let's see if this gets reflected:

Idle temp is indeed a couple of degrees higher due to previous warm up. Let's see if active cooling helps in the stress test:

As the maximum temp equals that of the previous block we can conclude that the differences and similarities in shape and volume account for very little in the long run. Active cooling does its part though.

Tests results:

Test # Description Min Temp Max Temp
1 Default cooling solution (active) 26ºC 39ºC
2 Block Nº1 (passive) 33ºC 51ºC
3 Block Nº1 + fan (active) 32ºC 46ºC
4 Block Nº2 (passive) 27ºC 50ºC
5 Block Nº2 + fan (active) 31ºC 46ºC

Test Nº6

CPU stress temps with ultimate cooling taboo: Block Nº1 + block Nº2 + fan (hybrid).

For the last stand we are going to stack this monstrous pile of aluminium plus the auxiliary fan. We added thermal paste between the two blocks. Blocks were warm from previous tests. Let's see how this experiment results. Idle temps first:

Idle temp is around the usual value. Let's see how this cooling monster does in 5 minutes of sustained work in the stress test:

Well, it does tend to keep temps down by a bit. That piled mass is able to transfer more heat into its body, relieving the CPU by a bit amount. The idea is not that bad but a better layout paired with a better fan might yield even better results.

Tests results:

Test # Description Min Temp Max Temp
1 Default cooling solution (active) 26ºC 39ºC
2 Block Nº1 (passive) 33ºC 51ºC
3 Block Nº1 + fan (active) 32ºC 46ºC
4 Block Nº2 (passive) 27ºC 50ºC
5 Block Nº2 + fan (active) 31ºC 46ºC
6 Block Nº1 + block Nº2 + fan (hybrid) 33ºC 45ºC

Conclusion

Aerospace aluminium does a great job as a passive CPU heatsink in my opinion. Though I can only recommend passive solutions for CPUs rated at 35w TDP or less. Currently there are passive solutions available in the market that offer good enough cooling at an affordable price. We've got great temps if we compare against commercial solutions, even when using higher TDP units than the recommended.
On the negative side of things, I think these blocks have great heat transfer but have difficulties dissipating all that energy into the air. It would have been interesting to see for example if putting a cold can of beer on top of the blocks would have produced some effects, that or some other experiment.
So, if it happens that you have a piece of aerospace aluminium lying around in your house (because why not) or you have a piece of unknown metal fallen from a UFO then I would recommend to give it a try if a silent PC is your thing.

I'll leave you with a good review of a commercial passive heatsink from a channel I appreciate a lot. I think those images are worth more than these words about this thing that makes no sound.........w?!