RoboNuggie I haven't tried the freezer but I did try using and old beer fridge putting the Pi inside and insulating the gap for the wires to come out , it made very little difference about 3°c all in all so just put beer back in the fridge and called it a day
lol, brilliant Mark...a win-win situation really: Pi = Cooler --> Drink Beer (shame to let it warm up) Pi= Not much cooler ---> Put back in fridge, drink beer later... Huzzah!
RoboNuggie So damn cool.. Am a software guy (CS) how can I start learning to program FPGS and use VHDL ... do I need to learn anything else before that .... Can I get into Arduino immediately without taking other courses
Just as an aside, the accessory care package that comes with Noctua stuff is part of the reason we who like them in our PCs buy them. That fan can be got for about £10 although £13 is more common. There's at least £5 of accessories included in the extension/low noise cables and those silicone mounts.
Put zip ties to secure the heat sink firmly to the Pi. Better contact will yield better temperatures. In practice, those can go all the way down to 10 degrees difference, but since you are already (relatively) so close to ambient under load (about 15 degrees delta), your gains won't be as big. Also, try putting thermal paste below as well as above the copper plate. Also, if possible, try to get a thinner copper plate. Would be a fun video if you could try the suggestions from the comments.
im now at this point in having to add a fan the cpu/gpu already has a heatsink. following your guide using pin 4 - 5v and 6 - gnd . When i power on the board it does nothing doesnt spin up. I have a noctua 40mm fan 5v 0.25w 0.5a. should it work without being programmed in
I don't think the aluminium heatsink will short the components as it is anodized: "The resulting anodized film is electrically non-conductive" source: www.chemprocessing.com/page.asp?PageID=1
I'm testing it with a 115W TDP Cooler that is supposed to cool i7s and Ryzens this weekend, to kick off the blog I'm launching after that. Try googling "Raspberry Pi 4 Arctic Freezer" mid August. Going to hook up a Pi with an SSD, an Arctic Freezer 7 Pro, Noctua NF-B9 and Thermal Grizzly Kryonaut paste, then bench the temps with and without overclocking in hardware hungry Retropie games like Perfect Dark (N64), Shen Mue (Dreamcast) or Monster Hunter Freedom (PSP).
People have done this already, it's not worth it, it's actually worse than a fan and a heatsink lol you can really only pull a .2ghz more out of it, so really what's the point? lol
Curiously I noticed you directed the airflow away from CPU with the cheaper fan but directed the airflow towards the CPU with the noctua. Was that deliberate? Does it make any difference? Am I asking a stupid question? Oh right, and thanks for another video.
ExplainingComputers sorry I was just curious. My Pi case's instructions showed the fan blowing air away from the CPU but I set it up to blow on the CPU, so I wasn't sure if there was some CPU cooling secret I wasn't privy to.
if you have an open air design, i find it better for your fan to blow on the heat sink and if you have a closed case design, for it to pull air off of it. in the end though, the difference is marginal - you're still using the fan to promote airflow and the main difference is where you're going to get dust buildup over time. You should check out PC building airflow videos for explanations on that.
josh blowing away may be more effective, blowing down on a heatsink is known to be inefficient because air will hit the flat bottom of the heatsink. Always make air move across a heatsink so the airflow is unopstructed
Nice review :) I like the small heat sink and Noctua fan combination it seems a versatile and fairly robust solution. At some point in the future would you consider revisiting this topic with the same solutions and use the stress or stress-ng utility to see how well the cooling continues. I appreciate the number of combinations that the stress tool can provide might be a little daunting, therefore configuring a load that sits at load average of 8 over 1 minute would work well. I use these tools occasionally to see how well our scheduler is supporting my Linux audio components to prevent audio drop outs.
When it comes to heatsinks, I always go by the theory of making sure to have copper touch the CPU/GPU and then lead that copper into some type of aluminum. In my experience it's usually some type of aluminum fins. I noticed a huge difference when I first did this with using an all aluminum heatsink and then switching to one where copper touches the chip and leads off into aluminum. There's actually some physics behind this as copper absorbs heat much quicker and aluminum's ability to spread out and dissipate heat. I think that's how it works anyway, but it definitely makes a huge difference with the type of metals you use in regards to touching the chip to be cooled and so on.
That's it, i'll build one for myself. My Pi3 started showing the temperature icon while using Ubuntu Mate. And as always, thanks for the awesome content!
And I did adapt an aluminum heat sink! I was surprised at how hot it gets while doing unattended upgrades just after booting Ubuntu Mate. I wish there was a way to attach a picture.
I'll leave here a version of Chris bash script if anyone wants to run some benchmarks too: #!/bin/bash clear echo -e "\e[32m~~~~Initial temps~~~~\e[39m" /opt/vc/bin/vcgencmd measure_temp for f in {1..10} do echo -e "\e[32m~~~~Starting run" $f"~~~~\e[39m" sysbench --test=cpu --cpu-max-prime=20000 --num-threads=4 run > /dev/null 2>&1 /opt/vc/bin/vcgencmd measure_temp done
It would have been interesting to see the results when using a Pi fan without the small heatsink. In other words, to find the minimal configuration to keep the processor below 70C at full power.
I discovered ExplainingComputers on UA-cam about a year ago when I was first interested in Raspberry Pi. Now I find myself trying to watch the backlog of 10+ years of Chris Barnatt's videos any time my wife and kids are busy binging on Netflix. Today I discovered a whole list of his books available on Amazon!!! Thanks Chris, for sharing your knowledge and experience! You really make the best videos on UA-cam!
The problem with big heatsinks is that you block the GPIO... do you know a solution of good cooling without blocking the pins? The problem is you cannot have good cooling + shields (or hats).
Electronics on the Moon were cooled by heat sinks going to Octadecane and Eicosane wax... no air to move! That's per Ron Creel, LRV thermal control engineer... Something to try if I ever find a source...
Running Magic Mirror on the head, using a shined pre 1982 (1977) U.S. penny stuck to the cpu with some noctua NT-H1 in a case similar to this: www.amazon.com/KKmoon-Aluminum-Protective-Enclosure-Raspberry/dp/B07BFSVKFN I was able to get similar results to the fan and small heat sink. temp=33.8'C temp=49.4'C temp=51.0'C temp=51.5'C temp=51.5'C temp=52.1'C temp=51.5'C temp=52.6'C It should be noted that I have seen the temp down to 31.55556'C with just the Magic Mirror running
I glued a laptop fan onto a Pi case and put one of those little heatsinks on the Pi 3 B+ processor. My temps are: 25.7C, 37.6C, 37.6C, 38.1C 37.6C, 38.6C, 38.1C and 38.6C. No need to buy a new fan, just scavenge an old laptop for its fan and there you are!
I know this is old but i still reference this video to this day. Love all your vids by the way! I think i have been subscribed for officially 3 years now! i am 13 and im pretty certain i subscribed on this day when i was 10. I actually did not know that there was the GPIO pins that tap directly to the micro USB port. Thanks for the info! I salvaged an old PC fan and cut off some female ends of some GPIO wires and attached them to the positive and negative wires. Works great!
inspired by one of your video's I did something similar on a android box but used HC 910 thermal adhesive a very strong way to secure the heatsink cheers
I think I know what's next. An extreme overclock video that overclocks it to 1585 mhz. Because I overclocked my model B to 1385mhz and my ram to 635 mhz. I didn't overclock GPU. It didn't want to overclock.
Wow that's amazing that the idle temperature with no heatsink or fan is higher than the temperature the Pi had after the test with the LHS and the fan Absolutely amazing! :D
Just curious, if you tested the (3 rd) with the fan blowing on the heatsink as you did with the Noctua fan just to see how they compare. I noticed that in the video you had the small 30 mm fan blowing up instead of down on the HS. Great video on the other scenarios though . :) Edit)- I scrolled down a little further and seen someone ask the same question :). Thanks
If the server room is ”hot” then something isn’t right. Server rooms needs some kind of cooling installed, blowing around warm air doesnt’t do very much. But if by server room you mean a room at home where you keep a server that’s another thing :)
For a lot of digital electronics, engineers usually assume that the temperature rise (delta T) you see above your current ambient temp will apply at higher operating temps as well, within reason. So if you see a 15 degree rise in your system when ambient is 25C, you can be pretty confident you'll see the same 15C rise if ambient is at 50C, assuming your system is designed with reasonable airflow, all the parts are still in there listed operating range, and you don't have local hot spots likely to cause thermal runaway. This assumption works well 95% of time. Measured temps are usually within a couple of degrees of where you would expect them to be, and that difference may be due to slight changes in the test setup and the accuracy of the thermocouples. That doesn't absolve the engineers from testing the circuits at the top of their listed temperature range to insure an unexpected problem doesn't crop up. If you're building a power amplifier with large discrete transistors however, much of the power may be dissipated in individual devices. Bipolar transistors tend to increase gain with increasing temperature (which increases power), and you could risk a thermal runaway problem. As such, the constant delta T assumption should be checked more carefully.
Cynthia is right, measuring the ambient temperature is enough to express a delta over ambient temperature. A lot of publications use the Delta T to compare cooling solutions rather than the measured absolute temperature.
This is the most helpful stress testing video for the Raspberry Pi I have seen on UA-cam. Excellent and well thought out with clear repeatable information. Thank you.
I bought a Noctua NF-A4x10 5V fan (10mm high) and placed into a home made frame facing downwords just 2mm above a small heatsink as showed on the video and a small heatsink on the 2 Gb memory chip. The airstream is blowing out from the frame. Now the Raspberry Pi4b is hoovering around 49 to 50° C at full load and 35° C on idle with an ambient room temperature of 21.5° C
For Science, you monster. :) Also, why not? :P Yeah, yeah, I know.. just because you can, doesn't mean you should. So, let Christopher do it for you and learn something along the way? ;) Even if it turns out that there isn't much difference, its always good to see what is needed to keep it steady at a certain temperature. You know, for that "temperature sensitive" application that includes expensive crystals that throw out accuracy if it rises 0.1°C. Or not. It's academic at this point.
As an experienced overclocker, I do have to admit "just for the hell of it" is indeed a compelling argument. as for unstable crystals, well, a TCXO doesn't cost too much, or if you have access to good surplus an OCXO, or if you have a few extra bucks, Rubidium is cheap enough too ;) --... ...-- -.. . -.- -.-. ..... ...- -.. .---
😂Haha. When I saw baseline, petroleum jelly came to my mind (that's how it is called in Greek). Nice video. Also you can enable a taskbar object that measures the cpu temperature in real time, while doing the benchmark.
The large heatsink's passive cooling ability was impressive. I shall most likely use a similar setup for my Orange Pi PC+. Where did you get the copper? Thanks.
The copper came from Amazon; I searched for "3mm copper plate", and managed to get a piece 50x50mm. Searching today on Amazon I first found a 100x50mm piece; amzn.to/2rUkwG8 But there may well be smaller bits. I looked at this in more depth -- cutting it and such -- in this video: ua-cam.com/video/1AYGnw6MwFM/v-deo.html
Thank you , Dr. Barnatt, for this interesting, ambitious, practical, and useful research effort you have done for us. It has indeed helped me to decide how to best cool my various RPi contraptions. It would also be a truly wonderful world if all those commentators who are suggesting to you that you undertake to do many alternative research variations, would themselves instead do them for us, and then offer up their professionally-presented results to the rest of us, entirely free-of-charge as you do. In that case I could continue to enjoy the benefits of all those efforts without actually having to do anything myself. A wonderful world indeed!
The fan is a good solution. Im using it for the SBC projects, mostly from small netbooks, and Im also using the small heat sink. Last one was for my godson, when I fitted the SBC in a toy car, and the fan on the roof. He is using it for gaming purposes, and the temperature never raises above 60 C. Anyway, the final combination looks impressive. If you ever do decide to put such a massive radiator on the Pi, consider overclocking the CPU, and testing the results how far you can push the Pi limits
Delid the pi, and use liquid metal with a Peltier chip! Do pi-bong cooling, hook it up to a water cooling loop from your well, I love cooling videos! Especially on odd components.
There's no comparison to a boxed-in product like an XBox here -- in an open rig the fan is simply moving air through the vanes. And I know from tests -- there is no temperature difference from push or pull.
Hi Christopher, I really like your Raspberry PI videos, 2 months ago I got my first PI and I am now waiting for my 4th to arrive, I especially liked your Devastator Robot Videos as I would like to make a FPV Explorer Vehicle. Keep up the good work :)
watched a vid where a fella added a water cooling set up to his pi.. kind of defeated the mobility of the small board with a radiator that was way bigger than the board and with the reservoir and line made the area needed about 3 times as big... He called himself a "hacker" (I do not know what that is supposed to be these days) But he was obviously proud of himself.. A bit overboard I thought.. but this was a helpful experiment thanks Chris...
There are many possibilities with this, imagine the servers you could make, because if you can keep it that cool, making 24/7 NAS servers or Web servers is possible. Pretty awesome video.
Very , very Interesting .Since one hour , i was looking for information on the heat dissipation of the RPI and it was stressful to read debilitated forums. Nothing like an excellent demonstration video pedagogical. Thanks a lot
Do you know what the maximum load on the 5v gpio pins are? I am considering running fans on a seperate power supply as the pi would be on full load and running 24/7
The 5V pins are connected directly to the micro USB 5V input, before the voltage regulator or any other electronics on the Pi. So you can draw as much current as your PSU (minus what the Pi draws) will allow. Well, OK. there will be a limit to what the PCB tracks on the Pi can take. :)
Once again another brilliant video , I've personally gone down the water cooling route at a cost of £120 for a pre made kit but only on the RPi 3 and so far I had practically the same results as you had on the large passive cool system you got with the new Pi , I'll let you know my results when I finally get around to upgrading
Hey Chris you're videos are great man. I really appreciate what you're doing here. Anyways I just wanted to say that I'm not sure why you say that people have trouble getting their Pi 3B+ up to 1.6GHZ as long as you have great cooling. I must've hit the silicone lottery with mine because I have been running mine full tilt at 1.62Ghz for weeks now. The response time and loading time are a 1/4 of what they were at the beginning thanks to your videos.
I wish these tests also included the execution time of the benchmarks. It would help demonstrate how long the Raspberry Pi spent in the throttled state.
I hope you still reply to these older videos, but here goes. How loud was your PiFan? I have a couple of those and they've been very noisy to a point where I've needed to hide them to a sound"proof" box. I've also been planning to make them temperature controlled (or at least on @65C and off @50C or similar) by using a transistor as a switch and some Python, though this has proved to be difficult as for my lack of knowledge on electronics.
I am sorry, but did you put the pi-fan in pull configuration and the Noctua fan in the push configuration. Shouldn't you put both in push or pull configuration for consistency in results. I think both would have produced the same temperature results. Please correct the testing methodology/video!!
You are right, the first fan was the "wrong" way up. Though as you say, the results would probably be the same, esp. in an open rig. As I am sure you are aware, there is no way to change or "correct" a UA-cam video after upload, and it really does not matter! :)
You should take into consideration the time it takes to perform a heavy task, because if the cooling is poor then not only will the temperature change, but it will also begin to throttle, taking a longer time to complete the task.
Could you recommend a tutorial on how I could make the fan turn on and off when the CPU reaches a certain temperate level or maybe control a relay to switch the fan on/off again based on the CPU temp, perhaps this could be one of your future projects to show us how ? Many thanks
Will there ever be a day when c.p.u's ( and computers ) are designed to take in localised heat ( including self generated ) and turn it into power which then can be used to cool in a efficient way
I think "fan & shs" result may be much better if you can place fan not strait on the top of heatsink. Because center of fan is just rotating and doesn't move any air below! :^) There just cylinder of static air that cooled down by air convection arround, and only then cooled static air cools heatsink. I'm sure that heatsink cooled down directly by air convection (with fan installed aside) will perform better. =) [not my native language, in case of horrible errors]
You are right -- but I did anticipate this, and neither fan was directly on top of the heatsink for the reasons you mention (well, except in the last test with the Noctua screwed in the top of the large heatsink).
I wonder if you would get a better result from the 30mm fan (fan & SHS), with the fan flipped to intake instead of exhaust, just like how the Noctua fan was turned? I think it could be comparable to the Noctua results.
The 30mm fan was indeed, by accident, the "wrong" way up. But that will make almost no difference to the results (esp. in an open rig) as the airflow through the heat sink vanes will be pretty much identical. The Noctua delivers massively more airflow, so will always win! :)
Next time you should also add a timer so that it is possible to see if it is thermal throttling/how much it is thermal throttling. The less it thermal throttles, the faster it will complete the test.
Knuckles the Echidna Dude, add a fan. My Nespi with heat sinks and 30mm fan runs around 47 c. Even under load. If noctua made a 30 mm fan, I’d be using one. The pifan is a usable fix. A bit loud and needs proper mounting, otherwise it will vibrate and sound like a weed whacker. Also get a copper heatsink for the bottom chip.
does c mean Fahrenheit or Celsius because in Celsius the final temperature was 70.9*C. I you convert that to Fahrenheit that 159.62*F. You can die from the heat in that temperature.
C means celsius. A human being would indeed be greatly harmed if heated to 70.9 C. I'm not quite sure how why this is of note here though! :) Lots of things in the world -- like cookers and fires and kettles -- are heated to high temperatures on a very local basis.
I think the main problem with small heat sink and big fan was that there was death zone under fan motor similar in size to the size of small heat sink itself. You could try puting fan not directly above heat sink but a bit to the right or left. Also to stabilize big heat sink you could just use another DIY plastic frame that would tight fit the heat sink.
Ok, out of curiosity have you considered something like this: www.amazon.co.uk/Enzotech-Corp-MOS-C10-MOSFET-heatsink/dp/B004CL89D8/ref=sr_1_22?ie=UTF8&qid=1526824983&sr=8-22&keywords=small+heatsink+copper Or even go wild and try to fit this: www.amazon.co.uk/SVF152C29M-Processor-Heatsink-Cooler-3VHK9TMN010/dp/B07D5MW85P/ref=sr_1_30?ie=UTF8&qid=1526824983&sr=8-30&keywords=small+heatsink+copper
Liquid Nitrogen should get you there, failing that stand in a walk in freezer ;-)
Great video as always Chris. Love your SBC videos.
RoboNuggie I haven't tried the freezer but I did try using and old beer fridge putting the Pi inside and insulating the gap for the wires to come out , it made very little difference about 3°c all in all so just put beer back in the fridge and called it a day
lol, brilliant Mark...a win-win situation really:
Pi = Cooler --> Drink Beer (shame to let it warm up)
Pi= Not much cooler ---> Put back in fridge, drink beer later...
Huzzah!
i would just dual sided thermal tape to put directly the big heat sink on the rpi.
with noctua fan.
would be perfect.
RoboNuggie
So damn cool..
Am a software guy (CS) how can I start learning to program FPGS and use VHDL ... do I need to learn anything else before that ....
Can I get into Arduino immediately without taking other courses
RoboNuggie hahaha or sent it to space 😁🤣
Just as an aside, the accessory care package that comes with Noctua stuff is part of the reason we who like them in our PCs buy them. That fan can be got for about £10 although £13 is more common. There's at least £5 of accessories included in the extension/low noise cables and those silicone mounts.
Good point. :)
Put zip ties to secure the heat sink firmly to the Pi. Better contact will yield better temperatures. In practice, those can go all the way down to 10 degrees difference, but since you are already (relatively) so close to ambient under load (about 15 degrees delta), your gains won't be as big. Also, try putting thermal paste below as well as above the copper plate. Also, if possible, try to get a thinner copper plate. Would be a fun video if you could try the suggestions from the comments.
nice. they must modify the pie to have 4 small holes for more cooling mounting ideas.
Such holes would be so good!
No freaking space
You sure you're not gonna go all the way and add a water block? Haha!
I would have liked to see how this effects how long it takes to do the test
Fair point.
im now at this point in having to add a fan the cpu/gpu already has a heatsink. following your guide using pin 4 - 5v and 6 - gnd . When i power on the board it does nothing doesnt spin up. I have a noctua 40mm fan 5v 0.25w 0.5a. should it work without being programmed in
I don't think the aluminium heatsink will short the components as it is anodized: "The resulting anodized film is electrically non-conductive" source: www.chemprocessing.com/page.asp?PageID=1
Not something I'd like to try -- just one scratch . . .
Cant wait for raspberry pi 4B 4gb ram to be tested with extreme cooling like this :D
That will happen!
Hypetrain is rolling
Please try a peltier 😀
I'm testing it with a 115W TDP Cooler that is supposed to cool i7s and Ryzens this weekend, to kick off the blog I'm launching after that. Try googling "Raspberry Pi 4 Arctic Freezer" mid August. Going to hook up a Pi with an SSD, an Arctic Freezer 7 Pro, Noctua NF-B9 and Thermal Grizzly Kryonaut paste, then bench the temps with and without overclocking in hardware hungry Retropie games like Perfect Dark (N64), Shen Mue (Dreamcast) or Monster Hunter Freedom (PSP).
@@ExplainingComputers now the 8 GB!!
Pimp my pi :)
Frankenstein's monster toward the end haha !
great job on the cooling!!
Nice
Ask Linus tech Tips to do liquid cooling on Pi.
They don't care about anything that costs less than a gtx 1080.
People have done this already, it's not worth it, it's actually worse than a fan and a heatsink lol you can really only pull a .2ghz more out of it, so really what's the point? lol
@@bobbybologna3029 i hope the foundation next gen pi will not come with factory locked processor.
@@minepro1206 unless it's droppable.
Bobby Bologna actually water cooling PCs works better than fans
47.2
47.2
47.2
47.2
47.2
47.2
47.2
OCD approved
Curiously I noticed you directed the airflow away from CPU with the cheaper fan but directed the airflow towards the CPU with the noctua. Was that deliberate? Does it make any difference? Am I asking a stupid question?
Oh right, and thanks for another video.
It is a very good question, and I think an error on my behalf with the smaller fan. :(
ExplainingComputers sorry I was just curious. My Pi case's instructions showed the fan blowing air away from the CPU but I set it up to blow on the CPU, so I wasn't sure if there was some CPU cooling secret I wasn't privy to.
if you have an open air design, i find it better for your fan to blow on the heat sink and if you have a closed case design, for it to pull air off of it. in the end though, the difference is marginal - you're still using the fan to promote airflow and the main difference is where you're going to get dust buildup over time. You should check out PC building airflow videos for explanations on that.
josh blowing away may be more effective, blowing down on a heatsink is known to be inefficient because air will hit the flat bottom of the heatsink. Always make air move across a heatsink so the airflow is unopstructed
not that you'd want to but could you re-run the cheaper fan again I'm curious as to how that'll play out against the noctua
Nice review :) I like the small heat sink and Noctua fan combination it seems a versatile and fairly robust solution. At some point in the future would you consider revisiting this topic with the same solutions and use the stress or stress-ng utility to see how well the cooling continues. I appreciate the number of combinations that the stress tool can provide might be a little daunting, therefore configuring a load that sits at load average of 8 over 1 minute would work well. I use these tools occasionally to see how well our scheduler is supporting my Linux audio components to prevent audio drop outs.
When it comes to heatsinks, I always go by the theory of making sure to have copper touch the CPU/GPU and then lead that copper into some type of aluminum. In my experience it's usually some type of aluminum fins. I noticed a huge difference when I first did this with using an all aluminum heatsink and then switching to one where copper touches the chip and leads off into aluminum. There's actually some physics behind this as copper absorbs heat much quicker and aluminum's ability to spread out and dissipate heat. I think that's how it works anyway, but it definitely makes a huge difference with the type of metals you use in regards to touching the chip to be cooled and so on.
I love the experience you make us live. Thanks.
You always make such "cool" videos.
Thanks. :)
That's it, i'll build one for myself. My Pi3 started showing the temperature icon while using Ubuntu Mate. And as always, thanks for the awesome content!
And I did adapt an aluminum heat sink! I was surprised at how hot it gets while doing unattended upgrades just after booting Ubuntu Mate. I wish there was a way to attach a picture.
Thanks for the experiments. Superb, crystal-clear photography as usual.
I'll leave here a version of Chris bash script if anyone wants to run some benchmarks too:
#!/bin/bash
clear
echo -e "\e[32m~~~~Initial temps~~~~\e[39m"
/opt/vc/bin/vcgencmd measure_temp
for f in {1..10}
do
echo -e "\e[32m~~~~Starting run" $f"~~~~\e[39m"
sysbench --test=cpu --cpu-max-prime=20000 --num-threads=4 run > /dev/null 2>&1
/opt/vc/bin/vcgencmd measure_temp
done
It would have been interesting to see the results when using a Pi fan without the small heatsink. In other words, to find the minimal configuration to keep the processor below 70C at full power.
Ah, I missed that one. Would indeed be interesting to try given the heat spreader on the Pi 3 B+.
I discovered ExplainingComputers on UA-cam about a year ago when I was first interested in Raspberry Pi. Now I find myself trying to watch the backlog of 10+ years of Chris Barnatt's videos any time my wife and kids are busy binging on Netflix. Today I discovered a whole list of his books available on Amazon!!! Thanks Chris, for sharing your knowledge and experience! You really make the best videos on UA-cam!
Would the 30mm fan not be more effective if it were blowing towards the cpu ?
It would make very little difference in an open rig like this -- probably none! :)
Maybe even more effective as it'll be drawing air through the heatsink and pushing it away ! @@ExplainingComputers
The problem with big heatsinks is that you block the GPIO... do you know a solution of good cooling without blocking the pins? The problem is you cannot have good cooling + shields (or hats).
Lookout for my next Pi cooling video in a few weeks with the answer to this . . .
Electronics on the Moon were cooled by heat sinks going to Octadecane and Eicosane wax... no air to move! That's per Ron Creel, LRV thermal control engineer... Something to try if I ever find a source...
Running Magic Mirror on the head, using a shined pre 1982 (1977) U.S. penny stuck to the cpu with some noctua NT-H1 in a case similar to this: www.amazon.com/KKmoon-Aluminum-Protective-Enclosure-Raspberry/dp/B07BFSVKFN I was able to get similar results to the fan and small heat sink.
temp=33.8'C
temp=49.4'C
temp=51.0'C
temp=51.5'C
temp=51.5'C
temp=52.1'C
temp=51.5'C
temp=52.6'C
It should be noted that I have seen the temp down to 31.55556'C with just the Magic Mirror running
My Favorite CS teacher is Mr Christopher Barranet.
I like Counter Strike also!
I glued a laptop fan onto a Pi case and put one of those little heatsinks on the Pi 3 B+ processor. My temps are: 25.7C, 37.6C, 37.6C, 38.1C 37.6C, 38.6C, 38.1C and 38.6C. No need to buy a new fan, just scavenge an old laptop for its fan and there you are!
Overclocking? 1.5ghz! Hmm mini water cooler! I must try :)
I know this is old but i still reference this video to this day. Love all your vids by the way! I think i have been subscribed for officially 3 years now! i am 13 and im pretty certain i subscribed on this day when i was 10. I actually did not know that there was the GPIO pins that tap directly to the micro USB port. Thanks for the info! I salvaged an old PC fan and cut off some female ends of some GPIO wires and attached them to the positive and negative wires. Works great!
inspired by one of your video's I did something similar on a android box but used HC 910 thermal adhesive a very strong way to secure the heatsink cheers
Seriously, didn't know noctua made those fans, lol.. PC-master race..
That's a horrible haircut.
I think I know what's next. An extreme overclock video that overclocks it to 1585 mhz. Because I overclocked my model B to 1385mhz and my ram to 635 mhz. I didn't overclock GPU. It didn't want to overclock.
Try overclockimg sdcard.
I did that. 100.25 mhz. The sd card READER. Yeah, it now performs full speed.
30G Put a heatsink on the RAM, take it to 700 Mhz. Also you can overvolt the CPU to get better OC.
The question is now, what are the performance hits between 38c to sub 70c (before the throttle). would the Pi perform the same @38c than at 54c?
Wow that's amazing that the idle temperature with no heatsink or fan is higher than the temperature the Pi had after the test with the LHS and the fan
Absolutely amazing! :D
And a great observation -- I wish I had spotted that in the video!
Thanks! xD
what about trying the noctua fan on top of the small heatsink, but pulling air UP instead of pushing down? Can you test it? Thanks :)
Just curious, if you tested the (3 rd) with the fan blowing on the heatsink as you did with the Noctua fan just to see how they compare. I noticed that in the video you had the small 30 mm fan blowing up instead of down on the HS.
Great video on the other scenarios though . :)
Edit)- I scrolled down a little further and seen someone ask the same question :).
Thanks
I was waiting for the liquid nitrogen. I am disappointed. Thumbs down.
good vid, thanks for this, it wd be interesting to vary the ambient temp too, to simulate being iin a hot server room...just a thought
Nice idea -- noted for a future video if I can figure out how to control ambient.
If the server room is ”hot” then something isn’t right. Server rooms needs some kind of cooling installed, blowing around warm air doesnt’t do very much. But if by server room you mean a room at home where you keep a server that’s another thing :)
For a lot of digital electronics, engineers usually assume that the temperature rise (delta T) you see above your current ambient temp will apply at higher operating temps as well, within reason. So if you see a 15 degree rise in your system when ambient is 25C, you can be pretty confident you'll see the same 15C rise if ambient is at 50C, assuming your system is designed with reasonable airflow, all the parts are still in there listed operating range, and you don't have local hot spots likely to cause thermal runaway.
This assumption works well 95% of time. Measured temps are usually within a couple of degrees of where you would expect them to be, and that difference may be due to slight changes in the test setup and the accuracy of the thermocouples.
That doesn't absolve the engineers from testing the circuits at the top of their listed temperature range to insure an unexpected problem doesn't crop up.
If you're building a power amplifier with large discrete transistors however, much of the power may be dissipated in individual devices. Bipolar transistors tend to increase gain with increasing temperature (which increases power), and you could risk a thermal runaway problem. As such, the constant delta T assumption should be checked more carefully.
Cynthia is right, measuring the ambient temperature is enough to express a delta over ambient temperature. A lot of publications use the Delta T to compare cooling solutions rather than the measured absolute temperature.
For those who do not pay their bills in $ or £ but in €: the Noctua fan is about 14.90 €.
This is the most helpful stress testing video for the Raspberry Pi I have seen on UA-cam. Excellent and well thought out with clear repeatable information. Thank you.
Thanks. Look out for the follow-up using a Pi 4B -- today! :)
I bought a Noctua NF-A4x10 5V fan (10mm high) and placed into a home made frame facing downwords just 2mm above a small heatsink as showed on the video and a small heatsink on the 2 Gb memory chip. The airstream is blowing out from the frame. Now the Raspberry Pi4b is hoovering around 49 to 50° C at full load and 35° C on idle with an ambient room temperature of 21.5° C
Totally want the desktop PC coolers onto the RPi. Also want to see if cooling to ambient is even possible :)
Cooling to ambient is always possible. There is the one caveat that comes with that: WHY?
For Science, you monster. :)
Also, why not? :P
Yeah, yeah, I know.. just because you can, doesn't mean you should.
So, let Christopher do it for you and learn something along the way? ;)
Even if it turns out that there isn't much difference, its always good to see what is needed to keep it steady at a certain temperature. You know, for that "temperature sensitive" application that includes expensive crystals that throw out accuracy if it rises 0.1°C. Or not. It's academic at this point.
As an experienced overclocker, I do have to admit "just for the hell of it" is indeed a compelling argument. as for unstable crystals, well, a TCXO doesn't cost too much, or if you have access to good surplus an OCXO, or if you have a few extra bucks, Rubidium is cheap enough too ;) --... ...-- -.. . -.- -.-. ..... ...- -.. .---
What if you used arctic silver instead of the shitty 2 sided tape on the small cooler
This would cleary improve matters! :)
😂Haha. When I saw baseline, petroleum jelly came to my mind (that's how it is called in Greek). Nice video. Also you can enable a taskbar object that measures the cpu temperature in real time, while doing the benchmark.
@Julius Songling Ναι!
well i have raspberry pi 3 b+ it is so hot without the heat sink and fan. when the fan is on it's so noisy maybe i have to change the fan...
The large heatsink's passive cooling ability was impressive. I shall most likely use a similar setup for my Orange Pi PC+. Where did you get the copper? Thanks.
The copper came from Amazon; I searched for "3mm copper plate", and managed to get a piece 50x50mm. Searching today on Amazon I first found a 100x50mm piece; amzn.to/2rUkwG8 But there may well be smaller bits. I looked at this in more depth -- cutting it and such -- in this video: ua-cam.com/video/1AYGnw6MwFM/v-deo.html
Thanks for that. (A Mr Scissors fan.)
i would just dual sided thermal tape to put directly the big heat sink on the rpi.
with noctua fan.
would be perfect.
Very helpful video! Applying these results to my MISTer!
Thank you , Dr. Barnatt, for this interesting, ambitious, practical, and useful research effort you have done for us. It has indeed helped me to decide how to best cool my various RPi contraptions. It would also be a truly wonderful world if all those commentators who are suggesting to you that you undertake to do many alternative research variations, would themselves instead do them for us, and then offer up their professionally-presented results to the rest of us, entirely free-of-charge as you do. In that case I could continue to enjoy the benefits of all those efforts without actually having to do anything myself. A wonderful world indeed!
The fan is a good solution. Im using it for the SBC projects, mostly from small netbooks, and Im also using the small heat sink. Last one was for my godson, when I fitted the SBC in a toy car, and the fan on the roof. He is using it for gaming purposes, and the temperature never raises above 60 C.
Anyway, the final combination looks impressive. If you ever do decide to put such a massive radiator on the Pi, consider overclocking the CPU, and testing the results how far you can push the Pi limits
Delid the pi, and use liquid metal with a Peltier chip! Do pi-bong cooling, hook it up to a water cooling loop from your well, I love cooling videos! Especially on odd components.
Thanks for the video, i am waiting for my b+ board and try these cooling tips. #log2pcmac
Towards the end in fact I wondered how far you would go 😄😄😄 and there you are!
i have the 3'rd config small fan and small heat sink and it's enough for what i need
I could not stop LOL that big cooler on that little Raspberry Pie
thank you
Would have been interesting to see the results of the first 30mm fan mounted the same as the Noctua.
Exactly, I see a ton of RPI3 builds with fans mounted to blow the air outside, you'd figure after the xbox 360 dissaster people learned their lesson.
There's no comparison to a boxed-in product like an XBox here -- in an open rig the fan is simply moving air through the vanes. And I know from tests -- there is no temperature difference from push or pull.
Hi Christopher, I really like your Raspberry PI videos, 2 months ago I got my first PI and I am now waiting for my 4th to arrive, I especially liked your Devastator Robot Videos as I would like to make a FPV Explorer Vehicle.
Keep up the good work :)
Thanks for this. I will return to the Devastator Robot project fairly soon.
watched a vid where a fella added a water cooling set up to his pi.. kind of defeated the mobility of the small board with a radiator that was way bigger than the board and with the reservoir and line made the area needed about 3 times as big... He called himself a "hacker" (I do not know what that is supposed to be these days) But he was obviously proud of himself.. A bit overboard I thought.. but this was a helpful experiment thanks Chris...
Go for mineral oil, and theme it as a aquarium lol
With fish? You certainly wouldn't need a heater!
interesting to see the difference. thank you from Spain.
There are many possibilities with this, imagine the servers you could make, because if you can keep it that cool, making 24/7 NAS servers or Web servers is possible. Pretty awesome video.
It's perfectly possible to keep a NAS or web server 24/7 as-is, with no cooling at all... Trust me, the SD card will be the thing that dies sooner.
Very , very Interesting .Since one hour , i was looking for information on the heat dissipation of the RPI and it was stressful to read debilitated forums. Nothing like an excellent demonstration video pedagogical. Thanks a lot
Should have tried liquid cooling it.
Do you know what the maximum load on the 5v gpio pins are? I am considering running fans on a seperate power supply as the pi would be on full load and running 24/7
The 5V pins are connected directly to the micro USB 5V input, before the voltage regulator or any other electronics on the Pi. So you can draw as much current as your PSU (minus what the Pi draws) will allow. Well, OK. there will be a limit to what the PCB tracks on the Pi can take. :)
Wow! the noctua fan actually works... interesting! thank you for this vid i'm learning everyday.
the noctua fan is probably quieter too, they're extremely good fans.
rickster4k Yeah! Noctua stuff works great. If only they'd offer better color choices...
Of course a noctua fan works.
Hey 1974 color schemes can't be beat!
they do offer better colour choices...
Add a Peltier cooler to the tests. Thanks for the demonstrations.
Once again another brilliant video , I've personally gone down the water cooling route at a cost of £120 for a pre made kit but only on the RPi 3 and so far I had practically the same results as you had on the large passive cool system you got with the new Pi , I'll let you know my results when I finally get around to upgrading
I am glad not to be the only one here trying these kinds of things! :)
Hey Chris you're videos are great man. I really appreciate what you're doing here. Anyways I just wanted to say that I'm not sure why you say that people have trouble getting their Pi 3B+ up to 1.6GHZ as long as you have great cooling. I must've hit the silicone lottery with mine because I have been running mine full tilt at 1.62Ghz for weeks now. The response time and loading time are a 1/4 of what they were at the beginning thanks to your videos.
You have done well in the silicon lottery to achieve 1.62. :)
I wish these tests also included the execution time of the benchmarks. It would help demonstrate how long the Raspberry Pi spent in the throttled state.
Fair point.
I hope you still reply to these older videos, but here goes. How loud was your PiFan? I have a couple of those and they've been very noisy to a point where I've needed to hide them to a sound"proof" box. I've also been planning to make them temperature controlled (or at least on @65C and off @50C or similar) by using a transistor as a switch and some Python, though this has proved to be difficult as for my lack of knowledge on electronics.
The Noctua fan here is audible, but not that loud. For the airflow it provides it is quiet. But if you want a silent room, this will be noisy for you.
we now need the 2 fans the other way to see the difference
I am sorry, but did you put the pi-fan in pull configuration and the Noctua fan in the push configuration.
Shouldn't you put both in push or pull configuration for consistency in results.
I think both would have produced the same temperature results. Please correct the testing methodology/video!!
You are right, the first fan was the "wrong" way up. Though as you say, the results would probably be the same, esp. in an open rig. As I am sure you are aware, there is no way to change or "correct" a UA-cam video after upload, and it really does not matter! :)
You should take into consideration the time it takes to perform a heavy task, because if the cooling is poor then not only will the temperature change, but it will also begin to throttle, taking a longer time to complete the task.
True.
Could you recommend a tutorial on how I could make the fan turn on and off when the CPU reaches a certain temperate level or maybe control a relay to switch the fan on/off again based on the CPU temp, perhaps this could be one of your future projects to show us how ? Many thanks
Will there ever be a day when c.p.u's ( and computers ) are designed to take in localised heat ( including self generated ) and turn it into power which then can be used to cool in a efficient way
Saxon , I like how you think
I think "fan & shs" result may be much better if you can place fan not strait on the top of heatsink. Because center of fan is just rotating and doesn't move any air below! :^) There just cylinder of static air that cooled down by air convection arround, and only then cooled static air cools heatsink. I'm sure that heatsink cooled down directly by air convection (with fan installed aside) will perform better. =) [not my native language, in case of horrible errors]
You are right -- but I did anticipate this, and neither fan was directly on top of the heatsink for the reasons you mention (well, except in the last test with the Noctua screwed in the top of the large heatsink).
Excellent video... but what about liquid nitrogen? ;-)
Uhm, now I wonder if I can get some on Amazon! :)
ExplainingComputers that would be awesome!! ;)
I too like the noctua fan, but I’m populating the GPIO pins with an mmdvm. So I’m using the 5v pins to power the module. Any suggestions?
Some kind of break-out board from the GPIO pins, perhaps.
With the large heat sink covering the wifi chip, did you see any connectivity issues? I would think it could possibly interfere with signal strength.
I did not notice an issue, but did not explicitly experiment with the WiFi to find issues. Good point.
I wonder if you would get a better result from the 30mm fan (fan & SHS), with the fan flipped to intake instead of exhaust, just like how the Noctua fan was turned? I think it could be comparable to the Noctua results.
The 30mm fan was indeed, by accident, the "wrong" way up. But that will make almost no difference to the results (esp. in an open rig) as the airflow through the heat sink vanes will be pretty much identical. The Noctua delivers massively more airflow, so will always win! :)
Next time you should also add a timer so that it is possible to see if it is thermal throttling/how much it is thermal throttling. The less it thermal throttles, the faster it will complete the test.
True.
Keep it up, amazing content!
You always make great videos, my Pi must be overheating with it's tiny heatsink and NESPi case.
The NESPi case has mounting for a 30mm fan as I used here.
ExplainingComputers think I'll have to invest in one! Poor Pi, has to work quite hard running emulators.
Knuckles the Echidn
Knuckles the Echidna Dude, add a fan. My Nespi with heat sinks and 30mm fan runs around 47 c. Even under load. If noctua made a 30 mm fan, I’d be using one. The pifan is a usable fix. A bit loud and needs proper mounting, otherwise it will vibrate and sound like a weed whacker. Also get a copper heatsink for the bottom chip.
does c mean Fahrenheit or Celsius because in Celsius the final temperature was 70.9*C. I you convert that to Fahrenheit that 159.62*F. You can die from the heat in that temperature.
C means celsius. A human being would indeed be greatly harmed if heated to 70.9 C. I'm not quite sure how why this is of note here though! :) Lots of things in the world -- like cookers and fires and kettles -- are heated to high temperatures on a very local basis.
Really interesting video, would be nice to see the test with a overclocked Raspberry.
Overclocking makes a HUGE difference when you're using kodi, from stuttering 1080p30 h.265, to no stuttering at all.
Very nice video! I love all your vids! Thank you for taking your time and doing these vids, I really do appreciate them. 👍
Thanks.
Liquid Cooling? xD
This guy looks like an Oblivion character
Bucket of mineral oil........
Very interesting experiment ! Thanks for sharing !
I think the main problem with small heat sink and big fan was that there was death zone under fan motor similar in size to the size of small heat sink itself. You could try puting fan not directly above heat sink but a bit to the right or left. Also to stabilize big heat sink you could just use another DIY plastic frame that would tight fit the heat sink.
In both cases the fan was not directly over the CPU -- in part due to mounting practicalities, but also to avoid the death zone as you note.
Ok, out of curiosity have you considered something like this:
www.amazon.co.uk/Enzotech-Corp-MOS-C10-MOSFET-heatsink/dp/B004CL89D8/ref=sr_1_22?ie=UTF8&qid=1526824983&sr=8-22&keywords=small+heatsink+copper
Or even go wild and try to fit this:
www.amazon.co.uk/SVF152C29M-Processor-Heatsink-Cooler-3VHK9TMN010/dp/B07D5MW85P/ref=sr_1_30?ie=UTF8&qid=1526824983&sr=8-30&keywords=small+heatsink+copper
whats with the little hole on the cpu is it for the bugs to get out of XD
Very cool, very cool.
No puns? Aaaaw. Very nice video though.
Plz draw some plot.......
Sf2590 yes, pls give us the plot
Arctic Cooling and Arctic Silver are completely different brands. You had Arctic MX4, that's not Arctic Silver :)
Yes, I've realized that now. My brain sometimes get confused under movie lights!
Noctua fans are a sham.. no better than any cheaper brand.
temp=30.6'C
temp=37.0'C
temp=37.4'C
temp=38.4'C
temp=37.9'C
temp=39.4'C
temp=38.4'C
temp=39.4'C
thank you
Excellent results.
You could have Made a custom LOOP😂