User avatar
AlphaJack
Posts: 29
Joined: Mon Feb 29, 2016 4:10 pm

Heatsink question

Sun May 29, 2016 9:20 am

I would like to overclock my Raspberry Pi 2 to use it with Kodi on Raspbian (it looks slow and I have to wait 3 seconds for every operation). My currently setup is:

Code: Select all

boot_delay=1
initial_turbo=30
arm_freq=1050
core_freq=500
sdram_freq=500
over_voltage=4
temp_limit=70
gpu_mem=320
Idle temps are ~49°C and Kodi ones are ~56°C. Even if I know that these are pretty normal values, I would like to put two pieces of aluminium heatsink with a small fan (12V 0.07A) on it. Image
The high one is very good for the fan, as its height does't create problems for the fan placed on the top, while the small one is perfectly sized for the SOC (using the middle line), and the usb/ethernet chip (using a lateral line).
:?: Which one do you suggest me for cooling?
:?: Also, do I need a thermal adesive or only a thermal paste?
:?: If in your opinion I don't notice a significant temperature difference between the 2 types, will you suggest me one of them or a part of each one (for praticality)?
:?: I tried to connect the fan to the first 2 right GPIO (5V) but it didn't start, when if it is connected to my PC motherboard (12V) it starts, what do you think the problem is?
:?: Is my setup safe or would you change something?
I know that these are a lot of questions but I didn't want to open a lot of different threads only for me, and I'll be glad of any of your answers, thanks in advance
Last edited by AlphaJack on Sun May 29, 2016 4:31 pm, edited 1 time in total.

User avatar
davidcoton
Posts: 3946
Joined: Mon Sep 01, 2014 2:37 pm
Location: Cambridge, UK

Re: Heatsink question

Sun May 29, 2016 11:11 am

AlphaJack wrote:I would like to overclock my Raspberry Pi 2 to use it with Kodi on Raspbian (it looks slow and I have to wait 3 seconds for every operation). My currently setup is:

Code: Select all

boot_delay=1
initial_turbo=30
arm_freq=1050
core_freq=500
sdram_freq=500
over_voltage=4
temp_limit=70
gpu_mem=320
Idle temps are ~45°C and Kodi ones are ~55°C.
:?: Which one do you suggest me for cooling?
Neither. If you only reach 55°C cooling is unnecessary.
AlphaJack wrote: :?: Also, do I need a thermal adesive or only a thermal paste?
If you do use a heatsink, thermal paste is necessary to make it work properly.
AlphaJack wrote: :?: If in your opinion I don't notice a significant temperature difference between the 2 types, will you suggest me one of them or a part of each one (for praticality)?
Whatever fits best.
AlphaJack wrote: :?: I tried to connect the fan to the first 2 right GPIO (5V) but it didn't start, when if it is connected to my PC motherboard (12V) it starts, what do you think the problem is?
If you used the two 5V pins, they are connected together, both at the same voltage. You need one 5V and one 0V pin. A 12V fan is designed for 12V, so there is no guarantee it will start at 5V. With a heatsink on a Pi2, you will not need a fan. Without a heatsink, it won't do much anyway.
AlphaJack wrote: :?: Is my setup safe or would you change something?
AFAICT, quite safe. The Pi can run to 85°C without problems.
Signature retired

User avatar
Zalamel
Posts: 23
Joined: Wed May 08, 2013 11:25 am

Re: Heatsink question

Sun May 29, 2016 1:29 pm

The RPi 2 really doesn't need a heatsink, as David said. Heatsinks are more of a Raspberry Pi 3 thing. However, if you're planning on getting a RPi 3 eventually, these videos may be of some help as to how to cool it, and how effective each method is:

Raspberry Pi 3: CPU Temperature Tests & Heatsink

Raspberry Pi 3: Fan and Cooling Tests

Raspberry Pi 3: Extreme Passive Cooling

Raspberry Pi 3: More Extreme Cooling

Cheers! :)

User avatar
MarkHaysHarris777
Posts: 1820
Joined: Mon Mar 23, 2015 7:39 am
Location: Rochester, MN
Contact: Website

Re: Heatsink question

Sun May 29, 2016 1:45 pm

davidcoton wrote: ... quite safe. The Pi can run to 85°C without problems.
um, not true.

@alphajack, the PI 2B runs happily with passive cooling unless you over-clock it. The PI3B has been over-clocked from the get-go; it definitely requires 'extended' passive cooling at a minimum, and runs best with active cooling, particularly if its going to be cased, and particularly if you're going to do anything with it other than have it idling all day. If your PI idles most of the time you'll get by without extended cooling. If you start to use it for non trivial computing tasks (you use all of its cores to do graphics processing, number crunching, &c, your PI will exceed 85c and it will crash... many of us have proven this beyond doubt or dispute.

My personal cooling options have been posted here, and here...
marcus
:ugeek:

User avatar
davidcoton
Posts: 3946
Joined: Mon Sep 01, 2014 2:37 pm
Location: Cambridge, UK

Re: Heatsink question

Sun May 29, 2016 2:40 pm

MarkHaysHarris777 wrote:
davidcoton wrote: ... quite safe. The Pi can run to 85°C without problems.
um, not true.
@MarkHaysHarris777, please withdraw that. The Pi SOC protects itself at 85°C (default) by throttling the clock. I know there are probable issues with Pi3s, but we are talking Pi2. There MAY, possibly, in extreme circumstances on a Pi2 be cases where the throttling is not enough to protect the Pi, and the temperature gets even higher -- but physical damage will not normally occur until well over 100°C. So the normal protection provides a more than adequate margin. If you have a Pi2 that was damaged by overclocking and subsequent thermal runaway, something was wrong with the Pi. I would be surprised if you can demonstrate that a Pi was thermally damaged without exceeding 85°C. In this case we have evidence that it is running nice and cool without heatsink or fan under its intended load.
Signature retired

User avatar
AlphaJack
Posts: 29
Joined: Mon Feb 29, 2016 4:10 pm

Re: Heatsink question

Sun May 29, 2016 4:30 pm

davidcoton wrote:If you do use a heatsink, thermal paste is necessary to make it work properly.
Yes, but I asked if I should use a thermal adhesive or only simply a thermal grease
davidcoton wrote:You need one 5V and one 0V pin. A 12V fan is designed for 12V, so there is no guarantee it will start at 5V
Thanks for the clarification, but it doesn't start even if I use the second and the third pins of the RPi (they should be 5V and ground, as required). The fan works very well if attached to a mobo (12V) or to a molex (using the 5V red and a black holes)
Thanks for the other answers, I really appreciate your precision
This would be my ideal cooling solution, but from the thermal photos on the web it looks that the RPi 3 usb/eth doesn't heat as much as the RPi 2 one, so I would need another small heatsink, but a big one like the one in the video would be too much overkill for me I suppose
MarkHaysHarris777 wrote:My personal cooling options have been posted here, and here
This is what I always wanted to do, but my fan (12V 0.1A) doesn't start on the second and the third pin with the stock pin connector while your (5V 0,2A) seems to be perfect for the RPi. Would you suggest me to remove the plastic and use appropriate GPIO connectors, to use the big heatsink as passive cooling, or even to cut an old usb cable to join it to the fan wires (no problem for this one because I would eventually plug it to the usb hub where the RPi is connected to, but I must be sure that this will work)

ejolson
Posts: 3246
Joined: Tue Mar 18, 2014 11:47 am

Re: Heatsink question

Sun May 29, 2016 5:09 pm

AlphaJack wrote: Would you suggest me to remove the plastic and use appropriate GPIO connectors, to use the big heatsink as passive cooling, or even to cut an old usb cable to join it to the fan wires (no problem for this one because I would eventually plug it to the usb hub where the RPi is connected to, but I must be sure that this will work)
Since your goal is over clocking, how much cooling will determine how much over clocking. At some point there is little to be gained from additional cooling. This is, in part, dependent on the exact silicon sample in your unit. Some will over clock more than others.

Cutting wires to plug the fan into a USB hub seems a waste. If you are going to that trouble, I would suggest rummaging around in the electronics recycle bin for a 9 to 12v power supply and use that. Check the adapters for powering thrown-away network switches, WiFi hubs, amplified speakers or charging older cell phones. Some will produce power in the 9 to 12 volt range.

Pithagoros
Posts: 580
Joined: Wed Nov 12, 2014 8:16 pm

Re: Heatsink question

Sun May 29, 2016 6:37 pm

As I (not encasing my Pis) don't use heatsink on RPI3, I DOUBT and DISPUTE that they are absolutely necessary. I would agree that they are a good idea though.

Just making an assertion on a web forum doesn't make it so.

User avatar
MarkHaysHarris777
Posts: 1820
Joined: Mon Mar 23, 2015 7:39 am
Location: Rochester, MN
Contact: Website

Re: Heatsink question

Sun May 29, 2016 6:46 pm

AlphaJack wrote:
MarkHaysHarris777 wrote:My personal cooling options have been posted here, and here
This is what I always wanted to do, but my fan (12V 0.1A) doesn't start on the second and the third pin with the stock pin connector while your (5V 0,2A) seems to be perfect for the RPi. Would you suggest me to remove the plastic and use appropriate GPIO connectors, to use the big heatsink as passive cooling, or even to cut an old usb cable to join it to the fan wires (no problem for this one because I would eventually plug it to the usb hub where the RPi is connected to, but I must be sure that this will work)
Yes, remove the stock connector from the fan cable, and connect the fan directly to the 5v & gnd of the board (I soldered directly to the GPIO pins (2 & 9) ; board numbering left to right top to bottom, with the ethernet port at the bottom, and the pins on the right, looking down at the board.

In fact, that little white plug on the end of the 5v fan will blow your PI !!! The connector is open on the side facing the first row of pins. If you plug it into pins( 4 & 6)--your only choice--the edge of the little white connector will put 5v directly against pin 3 !!! --- and no more PI for you !

I ordered my 5v brushless fans directly from China. Shenzen... took five weeks to get here. They were $3 each; they are perfect !--low current, low noise, high efficiency, cool chips!
Last edited by MarkHaysHarris777 on Sun May 29, 2016 6:59 pm, edited 1 time in total.
marcus
:ugeek:

User avatar
MarkHaysHarris777
Posts: 1820
Joined: Mon Mar 23, 2015 7:39 am
Location: Rochester, MN
Contact: Website

Re: Heatsink question

Sun May 29, 2016 6:56 pm

davidcoton wrote:
MarkHaysHarris777 wrote:
davidcoton wrote: ... quite safe. The Pi can run to 85°C without problems.
um, not true.
@MarkHaysHarris777, please withdraw that. The Pi SOC protects itself at 85°C (default) by throttling the clock. I know there are probable issues with Pi3s, but we are talking Pi2. There MAY, possibly, in extreme circumstances on a Pi2 be cases where the throttling is not enough to protect the Pi, and the temperature gets even higher -- but physical damage will not normally occur until well over 100°C.
I mean no disrespect, but your statement is simply not true. Water boils at 100c, at sea level. This is a matter of physics (science), not debate. Silicon chips should not be heated to those temperatures and doing so will drastically reduce their life (at this point, we don't know by how much yet).

I will not retract nor withdraw my statement... I and others have tested it empirically (this is a matter of scientific method, not debate, religion, nor opinion. I have driven both my PI2s and my PI3s to 85c and they both crashed... the problem is that they DO NOT reliably throttle back under all conditions. Rive has one condition (I have the other) where the PI will not only NOT throttle back, it will be damaged if you continue to push it... if you don't believe ME, then ask Heater... he was a non believer too... until he damaged his PI.

The trouble is that the PI does NOT throttle back enough, nor SOON enough. 85c is 'hotter' than the stove top temp I use to fry my eggs in the morning. Now, if you don't want to use active (nor even passive) cooling on your PIs, that's up to you... but against empirical evidence to the contrary you should not be telling anyone that its safe to run the PI at 85c without damage when you clearly don't know what you're talking about.

,respectfully
marcus
:ugeek:

Heater
Posts: 12972
Joined: Tue Jul 17, 2012 3:02 pm

Re: Heatsink question

Sun May 29, 2016 7:27 pm

MarkHaysHarris777,

I'm not sure what the boiling point of water has to do with anything. Apart from being a temperature reference.

I'm of the opinion that silicon transistors will run at 100C for a long, long time. However I also imagine that is not what happens.

Given the amount of power being pumped in and the tiny mass of the Pi SoC chip I'm inclined to think parts of the chip rapidly
get to exceed a lot more than 100C before the temperature sensor, wherever it is, notices. Damage can occur before any shut down can be triggered.

As it stands, I'm not totally sure I did damage my Pi 3 when stress testing it. Certainly it crashed or produced wrong results when stressed.

Unfortunately, I have yet to find a wall wart USB supply that can hold the voltage up enough to keep a Pi 3 happy under load. I have just acquired a 2.6 amp wall wart but have yet to try it.

More unfortunately I'm pretty sure the heat sink I have on the Pi 3 will not be sufficient when I get meet the power demands.

Experiments will continue but it's clear one cannot be so care free about the P3 power supply or thermal management.

User avatar
MarkHaysHarris777
Posts: 1820
Joined: Mon Mar 23, 2015 7:39 am
Location: Rochester, MN
Contact: Website

Re: Heatsink question

Sun May 29, 2016 7:38 pm

Heater wrote: I'm not sure what the boiling point of water has to do with anything. Apart from being a temperature reference.
Yes; its a reference, mostly for the English speakers on this side of the pond... who still for the most part don't think in c. The idea is that this reference is 212F (that's an F bomb!)
Heater wrote: I'm of the opinion that silicon transistors will run at 100C for a long, long time. However I also imagine that is not what happens.
Exactly right !
Heater wrote: Given the amount of power being pumped in and the tiny mass of the Pi SoC chip I'm inclined to think parts of the chip rapidly get to exceed a lot more than 100C before the temperature sensor, wherever it is, notices. Damage can occur before any shut down can be triggered.
Couldn't have said it better myself ! ... yesss. One of the things that a heatsink does (other than dissipate the energy) is to spread the energy out reducing it dramatically from a singularity...
Heater wrote: As it stands, I'm not totally sure I did damage my Pi 3 when stress testing it. Certainly it crashed or produced wrong results when stressed.
Well, and good. ... and from your standpoint the jury is still out on part of this ...
Heater wrote: ... unfortunately I'm pretty sure the heat sink I have on the Pi 3 will not be sufficient when I get meet the power demands.
Probably not.
Heater wrote: Experiments will continue but it's clear one cannot be so care free about the P3 power supply or thermal management.
And that, sir, is really my main point. Being cavalier about thermal management on the PI seems a bit more than at-your-own-risk to me! (especially if one is inclined to advise others!)
marcus
:ugeek:

paulv
Posts: 558
Joined: Tue Jan 15, 2013 12:10 pm
Location: Netherlands

Re: Heatsink question

Sun May 29, 2016 8:00 pm

Interesting discussion.

A very long time ago, (in RPi terms) I designed a PWM driven fan for the Model (1) B, that really didn't need it, but it may help with the model 3. It's fairly simple and can be changed easily. Please note that the installation I described is outdated by now, but the application itself will still work.
viewtopic.php?f=35&t=62438&p=463458#p463458

I also have a design with a dedicated fan controller chip I use for other (non RPi) applications, but that may be overkill. If there is interest though I can post the link.

Enjoy!

User avatar
Gerd
Posts: 66
Joined: Wed Mar 16, 2016 10:48 am
Location: Europe

Re: Heatsink question

Mon May 30, 2016 11:06 am

As physics was mentioned before in this thread, let's have a look at it. I have compiled a little excerpt from the datasheets of a heatsink producer. It lists different heatsinks together with their thermal resistance diagram.
On the x-axis of the diagram you find the air velocity in meters per second. Here v=0 means cooling by natural convection (no fan). Values v>0 are for forced convection (fan).
On the y-axis the thermal resistance of the heatsink in Kelvin per Watt is shown. The lower the thermal resistance, the cooler your Pi will run.
Lets look at the first heatsink. This is the one often sold in a kit with a Pi. It is 14mmx14mm and fits the cpu. At v=0 it has a large thermal restance of 30. As this may not be suficcient for your usecase, you decide to get a fan. What you see is a huge initial win. Lets say you fan blows with v=2m/s. Thermal resistance drops to 15. Not enough? Get a more powefull fan with say v=4m/s, this yiels R(th)=10.
Now lets jump to the penultimate heatsink in the compilation, the 31,34mm x 34,5mm one. Look at the thermal resistance at v=0, it is 10 K/W. But this is exactly the value you got for the small hs with a powerfull fan!

So two Pi sitting side by side will run at the same temperature, but one is noisy and eating more power.

Please draw your own conclusion, but as it is said: You can't beat the laws of physics.

To answer one of the questions of the OP: Convection, natural and forced, happens at the surface of the heatsink. So you should take the hs with the larger surface. From the photo it is not clear for me which of your 2 hs has the larger surface; I guess it is the left one.

PS.: All these heatsinks are available at a big german electronis shop and their nearly all cost the same. Prices range from €0,77 for the top one to €1,20 for the large one at the bottom.

Edit: typos
Attachments
Heatsink comparison small.png
Heatsink comparison small.png (49.63 KiB) Viewed 4171 times

User avatar
AlphaJack
Posts: 29
Joined: Mon Feb 29, 2016 4:10 pm

Re: Heatsink question

Mon May 30, 2016 8:01 pm

paulv wrote:viewtopic.php?f=35&t=62438&p=463458#p463458
I also have a design with a dedicated fan controller chip I use for other (non RPi) applications, but that may be overkill. If there is interest though I can post the link.
Nice job, but there isn't a way to control the fan (or at least read its speed) without buying anything?
Gerd wrote:Lets say you fan blows with v=2m/s. Thermal resistance drops to 15. Not enough? Get a more powerful fan with say v=4m/s, this yiels R(th)=10.
What is 2m/s in CFM? I've always seen fans classified for their airflow, not for the airspeed. Is this converter valid? Because in this case my fan (a 92mm Titan case fan, running at 5V) should blow at 0,5 m/s, compared to an Arctic F12 @12V (4,31 m/s).
Gerd wrote:Now lets jump to the penultimate heatsink in the compilation, the 31,34mm x 34,5mm one. Look at the thermal resistance at v=0, it is 10 K/W. But this is exactly the value you got for the small hs with a powerfull fan!
How about the materials (copper and aluminium, which one is the best for small size cooling) and height, comparing your last one with my hs?
2tall4u.jpg
Gerd wrote:To answer one of the questions of the OP: Convection, natural and forced, happens at the surface of the heatsink. So you should take the hs with the larger surface. I guess it is the left one has the larger surface.
Indeed, I was planning to cut the chosen one with the measures of the SOC and the USB/LAN (square 1,4mm and square 0,9mm): how can I let an heatsink bigger than the SOC to stay on the board without preoblems?

Also, from the 48°C IDLE of yesterday I'v got this:
2cool4u.PNG

User avatar
MarkHaysHarris777
Posts: 1820
Joined: Mon Mar 23, 2015 7:39 am
Location: Rochester, MN
Contact: Website

Re: Heatsink question

Mon May 30, 2016 8:21 pm

AlphaJack wrote:
Also, from the 48°C IDLE of yesterday I'v got this:
2cool4u.PNG

Now, drive it! Also, watch this YouTube. (11 minutes)

PS, copper is better than aluminium
marcus
:ugeek:

Return to “General discussion”