[DSL-AC68U] Temperture issues and how to solve them (project HEx-C)

[Zendilar]

Greetings to all, I have created this thread because I would like to share my discoveries and my modifications to improve the DSL-AC68U's temperatures.

I have 4 of these devises all of them are ruining GNUtonWrt (fork of AsusWrt Merlin), so as my good friend GNUton said his thread is about the FW not for the device's high temps from hardware design (the FW has nothing to do with the high temperatures) lets discus about that here.

I am not able to help him with his project because I can't write code but I am pretty good with hardware mods.

In Project~ HEx-C whitch stands for Project Heat Exchange I will try to advise and give you the cheapest and easiest solutions to reduce your device's temperature, and as well more expensive and advance modifications for anyone who likes to try.

Just to be straight from the beginning.

YOUR WARRANTY WILL BE VOID

and

I AM NOT RESPONSIBLE FOR ANY DAMAGE YOU MAY CAUSE TO YOUR DEVICE

if you think that you don't have the skills stay on the easy ones.

Now as for the high temperatures, I used to believe that it had to do with the FW but NO! there is nothing wrong with his fork, the official FW does not display the temperature of the cpu nor the ones from the wifi controllers. So when you install the GNUtonWrt (fork of AsusWrt Merlin) which displays them, it is natural to think that 80 C.-87C. and even 90 C. are too high.

And yes they are, and of course everything depends on the ambient temperatures but we are talking about a home device and I don't think that any one has below 20 C. in his home.

the temperatures in my country for the summer are close to 35 C. in a room without AC, and as I can recall there was no mention from asus that the device was designed for Antarctica, or that it always needs AC capability in the room.

The problem starts with the device's thermal exchange design which is wrong.

Its worth to mention that there are small differences between RT-AC68U and the DSL-AC68U, So don't post your RT's temps as a reference, because it has one device less (no DSL modem) and the heat sinks are different.

I don't know if it is revision related but all of my devises look like this
the DSL-AC68U is like this

the RT-AC68U is like this

The first one is in my office with ambient temp. 19 C to 21 C and the cpu temp. is 87 C.

The second one is at my home and with ambient temp. 25 C the cpu temp. had reached 97 C

For the moment and for testing purposes I have attached two 40x40mm fans at the back of the device with tape running at very low speed (no noise at all) and the temp is now 84 C to 85 C.

If speed is increased temp. can go to 75 C but there is a lot of noise.
This is by far the easiest and cheapest solution

But not for me, you see I don't want fans ,noise and dust in the device after two months, so I opened the device and removed the heat sinks to see if there was any improvement that could be done.

under the heat sinks are metal plates for shielding the components but only two small thermal pads.

the same goes for the cpu

 

[Zendilar]

Only one thermal pad

in some other thread they say to bend the heat sinks a little so that the contact is better...
well DON'T DO IT or the most possible situation is to end up with this like
Massimo93

that can be fixed but it needs a powerful soldering iron and some skill.
(the funny thing is that when it happend to me I had the exact same temperature 107 C.)

The problem is with the design, asus has put a bigger and more fat heat sink on the wifi socs that is why they never go above 60 C. , and a small thin heat sink on top of the cpu, and that is why you have a small first class mini oven on your desk.

the heat sink of the cpu has the same thickness as the one in the front, which is the heat sink for the components that are with the wifi socs, and it cools them thru the pcb, yes the components are on the other side, if you lift that heat sink there is nothing below it only the pcb (I wonder what they were thinking).

So I thought that if I make the contact surface bigger between the metal plate and the heat sink, may improve the situation.
I removed the pads from asus which by the way their quality is pretty good, they are just to small, and put in their place 120X20X1.0mm minus pad 8 from thermal grizzly. The cost was 6€ each and I used four of them to cover all the surfaces cpu and wifi socs, they are actually designed for ram.

After putting everything back and let the device run for 20 minutes there was a noticeable drop of the temperature from 97 C. to 87 C. with ambient temp. 25 C.
I thought, yes that is an improvement, so I decided to reopen the device, and replace the small pad inside the metal plate cover which is attached on the cpu with one bigger that covers all of it. After all I had some leftovers of my pads.

And yes after 20 minutes and with the same ambient temperature, the device's cpu temp. was now at 84 C.

(Every test was performed with cpu almost idle 3% to 6% usage, and I am not mentioning wifi temps. because they never even touch 60 C.)

 

[Zendilar]

In cases like this where all components are very close together, and space is to small every thing maters.
so I decided to test some more things, right now the temperature is 28 C. and the device's temp touches 89 C. from time to time.

but after removing the metal shield on the top of the pcb that stands between the antennas, because I noticed that it blocks the upper vent of the device's enclosure.

the temperature drop to 86 C. and after cooling the room to 25 C. with the ac, it dropped to 77 C.

Conclusion the passive air flow of the device is not designed correct, and of course the metal shield is there for a reason, although I didn't notice any abnormal behavior from the wifi, but I would like to put it back.

The next step that I have in mind is to replace the cpu heat sink with one thicker like the one of the wifi socs have.
All I ask is to give me some time because it has a custom design and I would like to find something that does not require to mach modification to fit.

to be continued....

 

[Swistheater]

I have seen crazy solutions from attaching fans to the back (i think this has the most benefit for a space heater needs) and also attaching gpu fans to the heat sink area ( this method is alittle too complex).

 

[Butterfly Bones]

This one started ~17 hours earlier, with more solutions and more posts.
https://www.snbforums.com/threads/dsl-ac68u-temperature.57060/#post-497568

 

[Zendilar]

This one started ~17 hours earlier, with more solutions and more posts.
https://www.snbforums.com/threads/dsl-ac68u-temperature.57060/#post-497568

Yes I have seen it, no research , no photos,(I don't see any solutions) and almost all the posts are about the RT-AC68U not the DSL-AC68U, as I mention above it is not the same device.
That is the reason for opening a new thread, after speaking with GNUton.

Just wait for me to post everything I have done.

 

[Grisu]

I tried to make air-slots of backside housing bit larger, so I broke away all the thin plastic noses between them, much work but temps did not really go down, not worth, only to tell no need for you to try this too.
Yes removing the metal shielding on top helps a bit, I tested it too on an RT-AC68U with similar result.
Did you try to remove pads and use much thermal paste instead? Or thin metal plates instead pads, you can buy them as set with different thickness.

At the end I used a very thin noctua 5V inside fan connected to the 3.3V connector I found on the PCB to run more quiet.
So you dont get dust inside but better air flow within housing. Helps about 10°C down.

 

[Zendilar]

I tried to make air-slots of backside housing bit larger, so I broke away all the thin plastic noses between them, much work but temps did not really go down, not worth, only to tell no need for you to try this too.
Yes removing the metal shielding on top helps a bit, I tested it too on an RT-AC68U with similar result.
Did you try to remove pads and use much thermal paste instead? Or thin metal plates instead pads, you can buy them as set with different thickness.

At the end I used a very thin noctua 5V inside fan connected to the 3.3V connector I found on the PCB to run more quiet.
So you dont get dust inside but better air flow within housing. Helps about 10°C down.

thanks for the heads up, that was one of the things I had in mind to.
But tell me the temps didn't drop at all not even one or two degrees? because with all the mods added together you gain one from this, three from that, five from the other so at the end you are getting to the desirable outcome.

I will attach a bigger heat sink now to see what I can gain, and report back.

 

[Grisu]

thanks for the heads up, that was one of the things I had in mind to.
But tell me the temps didn't drop at all not even one or two degrees? because with all the mods added together you gain one from this, three from that, five from the other so at the end you are getting to the desirable outcome.

maybe 1-2°C, for me it wasnt worth to mention as really much work to break them all out.
As I can remember replacing pad with paste would need to adapt mounting of the heatsink, otherwise there will be a gap or you really need a lot of thermal paste running out some day.

If there is space enough inside best you can do is to use this small and relative silent high qualitiy fan inside with very low <50mA current for ~12€, I did it in my RT-AC68U connected to 3.3V pins: https://noctua.at/en/nf-a4x10-5v
When I had DSL68 there has not been GNUton to see temps, so I didnt know how high they has been.

Or if you can live with a partial open case you could make a whole on the backside exactly 'above' heatsinks, maybe this could help a lot too get heat outside.
Never thought about this much more as I allways wanted to sell them some day without signs of mod and upgrade to newer router.

 

[Massimo93]

in some other thread they say to bend the heat sinks a little so that the contact is better...
well DON'T DO IT or the most possible situation is to end up with this like
Massimo93

Actually it was too much termal pads, just 1 "leg" wasn't enough to handle the pressure so it just broke like 5 minutes after closing everything again.
After that i moved away from any commercial router and did picked a x86 mini pc with a separate business access point.