My point is that the pi doesn't really get hot and so doesn't need a fan or heatsink, infamous "oven test" and all that. In instances where the pi is situated in some place where it does get hot, eg: hotter than an oven. I don't think a fan or heatsink is going to help there either.[email protected] wrote:Keeping stuff working at temperature is a bit of an odd art... I've worked with stuff that has run hot. Very hot. Last year I had an industrial PC inside a cupboard that housed the exhaust stack of a cross country train... It got hot. Fortunately it was designed to get hot - inside a sealed box with no fans (you're not allowed to have fans or vents on a PC bolted to a train!) The hottest we recorded it getting to was about 80C for the processor (Intel Atom - designed to run hot), and 102C for the 3G/GPS chip. That was 20C above the spec. for the GPS. Amazingly the 3G connection kept on going, but the GPS was reading about 10m out...shalo wrote:I think if anyone is planning to run their pi in a super hot environment they probably need a different approach from heatsinks and fans. They must also have one hell of a story to tell.
Some folks I'm working with at the moment are putting radio gear on masts in the desert. -20C to +50C is not unusual. Sealed boxes again, and no fans there, either...
So it all boils (ha!) down to the design. There is no fan in your mobile phone or fondleslab, so the designers make doubly sure the chip keeps on going when it does get hot - and the SoC in the Pi isn't a million miles away from the sort of SoC in your phone...
Lego fans are fun though
Bit overkill having an ATmega processor doing the temperature sensing, but hey...
According to this thermal analysis (it's in spanish, machine translation), the LAN9512 can reach 65,1ºC with less than 27ºC ambient temperature. The LAN9512 have a top working temperature of 70ºC so is almost there.shalo wrote:My point is that the pi doesn't really get hot and so doesn't need a fan or heatsink, infamous "oven test" and all that. In instances where the pi is situated in some place where it does get hot, eg: hotter than an oven. I don't think a fan or heatsink is going to help there either.
It says somewhere in the Wiki that the SD card can get quite warm too.AndrewS wrote:If you're gonna go crazy, why not stick heatsinks on the USB and Ethernet connectors too?
Oh No you di'intAndrewS wrote:If you're gonna go crazy, why not stick heatsinks on the USB and Ethernet connectors too?
Besides the SD would not fit into my windows machine any longer. Of course I am going to buy a ten pack of each of the two sizes. It could look like a new wave porcupine!alexeames wrote:It says somewhere in the Wiki that the SD card can get quite warm too.
It sticks out, so you could put one on each side
If the USB connector is cold he's obviously not using a fast enough USB Flash driveLob0426 wrote:Not much use look at the thermal image:
http://hackaday.com/2012/06/23/checking ... pberry-pi/
Look at my sig I am using a USB HDD, that carries the heat away from the USB connector and uses my desk as a heat sump. It's all about the heat dissipation.AndrewS wrote:If the USB connector is cold he's obviously not using a fast enough USB Flash driveLob0426 wrote:Not much use look at the thermal image:
http://hackaday.com/2012/06/23/checking ... pberry-pi/
Assuming that's true (which according to JamesH it shouldn't be for the SoC at least), I wonder if the low total cost of the RasPi means that any "extra life" gained by the use of heatsinks is still a false economy outweighed by the cost of the heatsinks?Lob0426 wrote:It has always been my experience that removing heat increases the life of electronics.
Never! Always fairly chilly in the UK anywayIf @AndrewS comes over to the "Dark Side" I might send him a couple too!
Few mistakes in the above. I believe that the chips are baked until they fail or we run out of test time. Which I think is weeks. These oven tests are to accelerate chip failure, and lots of funky stats are used to then estimate chip MTBF. We also bake the production chips - there's not point in basing a lifetime test on the test of a prototype chip except during development to find errors early on. And we do find errors, sometimes in design, sometimes in manufacturing. And they get fixed.Lob0426 wrote:Ah, but that depends on more than just the cost factor. If you use it in a product that places it in a tough environment, such as under the dash or seat of a car. And you give a warranty and you would have to lose labor hours to replace it at your cost. Then it makes sense to hedge your bet a bit. The initial cost is only one element of the entire equation. As to @JamesH statement that it was baked, that is true. But for how long, certainly not until it quit. And that was probably a prototype chip. Not a production RasPi. I would like to see more details of the environmental test performed on a production RasPi not just the chip or an Alpha Board. the poly fuses would probably cause problems before the SoC. But that does not invalidate the fact the heat kills components. In my early computers the HDD's quit on me rather often, until I started paying attention to their temps. Since then I have lost very few HDD's at all. Since I started paying attention to the CPU and chipset temps I have had more stable trouble free systems. A little prevention ($1.50 or so) is well worth the effort in my book. One project I have is burying a RasPi into a router case. You can feel the heat outside that router. Putting heat sinks on the RasPi and a fan on the router case will not hurt a thing, but may save a lot of trouble. Your RasPi your call. A $35 RasPi is a cheap investment, but investing in 3 failed RasPii rather than a $1.50 is a stupid investment!
Only exception to that being the original "fat" PS3s that had the YLoD problem - fixable by reflowing the BGAs with a hot-air gun. But the PS3 runs much much hotter than the RasPiLob0426 wrote:Thermal Cycling was much more of a problem with the "pin" style CPU's than it is anymore. The BGA style packages seem to have mostly cured this.
And the RRoD on the XBox 360m and the GPU on the iBook G3 and G4 models, there's a couple of Dell laptops with the same issue as well, IIRC. Thermal stress can be really bad for BGA devices, and far worse than it ever was for "dead bug" chips.AndrewS wrote:Only exception to that being the original "fat" PS3s that had the YLoD problem - fixable by reflowing the BGAs with a hot-air gun. But the PS3 runs much much hotter than the RasPi :DLob0426 wrote:Thermal Cycling was much more of a problem with the "pin" style CPU's than it is anymore. The BGA style packages seem to have mostly cured this.
Ahem... (No, I know that's not what you were talking about)Lob0426 wrote:No one can attack a computer that is turned off
If you don't want to use a saw you can try this from SparkFunLob0426 wrote:I took a RadioShack heatsink and hacksawed a piece of it off for the LAN9512. I dug out my Artic Silver and stuck that on the LAN chip and the rest on my SoC. About a ten minute job including the sawing.
That gave me images of someone cutting up the kitchen sinkpygmy_giant wrote:that gives me an idea- could hacksaw down a sink for the RG2 next to the big capacitor that easily gets knocked off.