AC86U Heat sink mod.... er hack

[Maverickcdn]

So while the network was down to upgrade firmware I decided to peel apart my 86u and check out the heatsink construction after other users complained of higher temps on 386...

To my surprise the construction of my 2018 build unit with PCB rev 1.5 was visually quite different than the article about this router on SNB. The large rf shroud around the wifi chips and cpu was certainly surprising

Anyway, to my dismay instead of machining a heat sink to close the gap between the heat sink and the chips they opted for 2.5mm thermal pads (typical for budget heat sinks)


I had a piece of 3mm copper stock that I probably could have thinned out more for a proper fit but I decided to wing it. Couple dabs of thermal paste and gentle reassembly and the results were, well.... substantial

Ambient 22

My 56 day uptime idle temp was 74-78deg on 384.19
1hr idle temp on 386.1 with copper mod.... 60deg
Added a 80mm fan above the lanports drawing air out of the case... well just look

Spoiler: Wowzer

Once 386 goes stable I may cut up more copper and do the wifi chips as well...


Update, disclaimer, read post 25

Remember this a warranty voiding MOD/HACK and like all physical mods/hacks it comes with its own inherent device killing risks. You've been warned.

 

[elorimer]

Interesting. My .19 temps at 20C ambient were 80C, just a little higher. I put a 30mm fan on blowing in, and the temps on 386.1b3 dropped to 68C. I didn't use a 80mm fan because so much of the back half is actually blocked by the solid diagonal mold of the back, and 30mm canted a bit is all unblocked. Since I used screws instead of glue I might reverse the fan and see if pulling air out is better. Ordinarily I would expect so, but the air flow is clearly designed to be in and up.

I was less ambitious in not taking the cover off. But wiring the fan to those jumpers just above is tempting.

 

[JWoo]

@Maverickcdn agree with you that the design of the shielding got quite a bit in the way of the heatsinking, so the actual heatsink has a transition thermal pad between it and the chip. Good discovery!!!

 

[Maverickcdn]

I was less ambitious in not taking the cover off. But wiring the fan to those jumpers just above is tempting.

Im not sure what voltage those headers provide, but honestly if you have an open USB port thats what I was going to use... till I noticed after reassembling one pin is fubar in the 2.0 port (can see it in the photo) and the 3.0 is in use. Ill fix it next time its apart...

And after flipping the fan to blow in above the lan ports (cpu side) now hovering 42-43

I used a 80mmx15mm I had laying around and wired it to a diy usb male plug then just two side taped the fan to the unit.... 30mm/40mm fans move virtually no air at desirable noise levels. 80mm at half voltage is near silent and still moves twice the air

@Maverickcdn agree with you that the design of the shielding got quite a bit in the way of the heatsinking, so the actual heatsink has a transition thermal pad between it and the chip. Good discovery!!!

As soon as I had the guts out of the plastics and saw the giant one piece heatsink I knew there would be thermal pads. They work great, thicker pads generally mean higher temps. 2.5mm is the thickest I personally like to see.

Its just too bad they already mill the bottom of the heat sink flat, they couldn't leave 3 plateaus to reduce the gap and maybe use 0.5mm pads... like they would in a high end cooler.

 

[heysoundude]

And after flipping the fan to blow in above the lan ports (cpu side) now hovering 42-43

I'm not brave enough (experienced enough with electronics) to open mine up like this. (you could open a custom shop...)
location and airflow direction - even without the copper heatsink - will undoubtedly bring temps down inside.

 

[jsbeddow]

Similar (but not exactly the same) to @heysoundude. I really want to try this, and I am just brave/stupid enough to try this, except for two things: one, that the whole family is fully reliant on a working internet connection, and two, I might still have some remaining warranty claim period left on my RT-AC86U. It was actually purchased from Amazon as a refurb/renewed unit, but I opted for an extended (3 year) 3rd party warranty that still has at least another year left on it.

 

[thecheapseats]

<snip>I was less ambitious in not taking the cover off. But wiring the fan to those jumpers just above is tempting.<snip>

that 4pin header (south of the ASUS silk-screen logo in the pic) is the internal serial port... gnd-tx-rx-vcc (not sure of the pin designations from the pic)... for others reading, be cautious hooking a serial connection into those with just 'any' rs-232 cable or usb-to-serial...

 

[Maverickcdn]

So to sum, was this necessary absolutely not... was it fun, yes but only cause I had the materials on hand while the whim of the idea was upon me. I've always done stuff like this, my Rpi3 with a huge oldschool Northbridge cooler overclocked to 1.5Ghz runs at 32-33deg all day.

I think users just have routers in stale air scenarios and heat soak is overwhelming their systems. Given the huge RF shield I discovered on my unit, its basically a giant tinfoil wrapper for CPU/Radiochips which WILL contribute to a larger heat soak temperature of the PCB etc... they only had foil spacers between the shroud and the heat sink, terrible for heat transfer, great for blocking RF, there is still places it contacts the heat sink though. The unit housing could use a few more cm2 of venting in my opinion.

The 14-18 deg drop I saw in the highly un-scientific testing I did between 384.19 and 386.1b3 with the copper mod is all a testament to copper vs thermal pad (so glad I didnt crush the cpu die). Sorry I didnt do more testing of copper vs pad and 384 vs 386 but I try and keep my network up as much as possible. The design of the heat sink all though sharing 3 heat sources is large and thick enough with a reasonable amount of fins/thickness to keep the cpu well within comfortable lifetime temps (sub 80) given SOME air flow around the unit, Im quite certain the Asus engineers have a bit of experience in this area.

In the end a second 80mm @ 5v blowing in

Ill probably remove both fans as it runs about 56 with no active cooling

And Ill chalk it up to the placebo effect as between 386/active cooling my entire network feels just a touch snappier (including WIFI).... I know its all just in my head

 

[thecheapseats]

" copper vs thermal pad "... that copper-shim sink will perform better every time... gotta' love a nice hack...

 

[maxbraketorque]

Does the AC86U open just like this:

 

[elorimer]

I think users just have routers in stale air scenarios and heat soak is overwhelming their systems.

In my case the router is in the open, in a basement heated to 19C. On .19 I was at 80C day in, day out; on .1B2 I was at 84C, and on B3 as is I was at 96C, with the wait mod at 92C. That isn't just a stale air scenario.

WIth the 30mm fan pushing in I'm at 68C for B3 with wait mod on. So I think this was worth doing. I have it wired to a USB plug; the fan draws 200ma at 5v and is silent.

This upright design I think is a better thermal solution than the 87U but neither is amenable to being put on top of a laptop cooler.

 

[Maverickcdn]

Does the AC86U open just like this

Yes I think all the stand up Asus routers have the same construction.... As soon as you peel the label though, warranty void

In my case the router is in the open, in a basement heated to 19C. On .19 I was at 80C day in, day out; on .1B2 I was at 84C, and on B3 as is I was at 96C, with the wait mod at 92C.

Now I'm curious, what do your radios report for temps? I don't use 2.4 so I don't have a reference. And what year\revision of 86u?

And how long did you let it rest after flashing? I found 386 with copper at 60 on passive at flash but 24hrs later passive idle set around 56

 

[elorimer]

A

Now I'm curious, what do your radios report for temps?

After two days up, the radios are at 43C(2.4) and 50C(5g), which is about 3C cooler than before the fan. The radios have been pretty constant.

2019 Rev 1.5

 

[Maverickcdn]

A

After two days up, the radios are at 43C(2.4) and 50C(5g), which is about 3C cooler than before the fan. The radios have been pretty constant.

2019 Rev 1.5

Interesting... seems pretty normal

Theres a multitude of reasons why you might be seeing the high temps, from poor heatsink pressure, degraded thermal pad, and even maybe the firmware to some extent. Being its from 2019 they may have even changed thermal pads to a lower quality which degrade quicker which will result in a logarithmic increase in temperatures as the oils/silicons separate in the pad.

Ive never been a fan of pads on dies, chips sure, dies should have liquid/paste/grease TIM between it and a cooler. The pad I pulled from the 86u not only being 2.5mm thick, seemed to have a low oil content (it was 3yrs old now) which means more air permeates (bad for heat transfer) and had already partially baked to the CPU die.

If you're out of warranty and up for it you could peel it apart and mod it
copper stock
Or you could just add thermal compound to the existing pads, but given the age and the temps they've seen youd be better off replacing them with new and using liquid compound

I still think you should at least cool the unit to ambient, reflash 386.1b3 and give it 24hrs and re-evaluate, I've always found temps after flashing to be a few degrees warmer for a few hours.

 

[Maverickcdn]

Had to reboot last night so the uptime is wonky. But 2 days with 386.1b3... its a little cooler here today

 

[maxbraketorque]

Yes I think all the stand up Asus routers have the same construction.... As soon as you peel the label though, warranty void

Thanks. I will likely do this to at least one of my AC86Us that is located in a closet that routinely gets up to 85F in the summer. I already have a fan on it now, but I'd like to do a heat sink mod too. I'll perhaps take the easy route and just smear thermal compound on the heat transfer pad. I guess these pads are some kind of wire mesh, perhaps like a brillo pad?

 

[maxbraketorque]

Also, how easy is it to break the clips that hold on the front cover? Do I need to go to great lengths to carefully pry it, or do the clips spring free rather than easily break?

 

[Maverickcdn]

Thanks. I will likely do this to at least one of my AC86Us that is located in a closet that routinely gets up to 85F in the summer.

85F in the room not the CPU temp right??

I guess these pads are some kind of wire mesh, perhaps like a brillo pad?

No its the pink thing in the first post. Its a silicone pad, somewhat spongy (assuming it isnt baked hard)

Also, how easy is it to break the clips that hold on the front cover? Do I need to go to great lengths to carefully pry it, or do the clips spring free rather than easily break?

I had to start it with a small flat had screw driver then used an old credit card to complete. The PCB assembly just pulls out once the case is apart. Edit. Sorry I think you have to remove the back heat shroud and 4 mounting screws, then it pulls out

If you opt for a 3mm piece of copper like I did , the actual clearance from factory was more like 2.3mm so you have the risk of crushing the die. Copper heat sinks require higher mounting pressures vs pads which is why I only filed/sanded my piece of copper and gently remounted the heat sink screws.

 

[maxbraketorque]

Thanks much. I'll try this soon.

 

[Darkje]

@Maverickcdn could i use a copperpipe and cut and flatten a small piece?