Raspberry Pi 3 - Idle/Load temperatures while running Retropie?

[dankcushions]

I use tape and it's fine. unless you've somehow applied it wrong but I would still not expect temps this high. I would be curious what temps it gets to outside of the case?

[hobnob]

Thank you everybody for all the input.

I've removed the heatsinks and now my temperatures have reduced by approximately 10C. Idling at the emulationstation fronted is around 55C and playing something is around 70C. The board is still within the provided plastic case and isn't throttling so far.

The provided heat sinks have thermal tape from the factory already applied to them. You simply peel the backing off and stick them on to the chips. The tape itself is clearly insulating the heat rather than dissipating it.

I wish I knew exactly what type of tape was utilized here as it's pretty obvious that it's not working as intended.

Hopefully this helps someone in the future who runs into heat issues with this kit too.

[caver01]

@hobnob Wow. I am glad you figured it out. It's pretty sad that the tape is having the opposite effect. Clearly, not all thermal tape is created equal. Case in point: @dankcushions says he's using tape successfully.

You could still recycle the heat sink--peel off that tape and clean any residue so it is absolutely clean and flat, then get yourself some paste/compound. My heat sink was cut from an old one I had (from an ancient Pentium CPU I think). I had to sand off the crusty residue, cut it to size and polish the bottom nice and smooth. The point is that a heat sink--any heat sink--can make a difference if it's applied correctly. Clearly, a corner was cut on that kit.

[tekn0]

Revive, I am having the same issue. On an RPI3 running FBA and Metalslug 1. The unit is hovering at around 83c. I tested with both heat syncs from that kit on and off. It holds around 80c on NES.

Both tests have the ctr-pi shader running in nearest neighbor.

[caver01]

@tekn0 the moment I added a heat sink to my Pi3 using arctic silver compound I have never seen my Pi throttle performance due to heat--not once. And my sink is just one I threw together by cutting a square off of a sink from an old Pentium CPU. I am really questioning the quality of the adhesive in these kits, as the heat sinks themselves look much better than mine.

[Concat]

Get a case + fan combo on ebay for like 10 bucks shipped. They work very well at keeping things cool... keeping it quiet is a different story, but oh well.

[mrbwa1]

The Pi3 runs hot, especially if you Overclock. Running a mild 1350Mhz OC with 500 MHz core and MEM bus speed, I can ply most 8 bit and 16 bit, but it's getting into the 70-75 degree range in a Pi Foundation Case with the lid off. I can't run Dreamcast, PSP or PSX with Enhance resolution for long without throttling though.

Once I figured out it was getting so hot, I rigged up a fan from a dead PC power supply and an old broken wall wart. It's not exactly quiet, but even going all out, temps rarely get above 55 degrees.

I would second the recommendation to get a case with a fan if you are seeing overheating. I tried just aiming a desk fan at my open case and even that kept temps around 60 degrees on the low setting (a lot less airflow that the loud PSU fan)

[Concat]

I'll add that a heat sink on its own really doesn't provide much cooling. If you research the topic you will see many example of people only getting 5 degrees cooler with a heat sink. Coupled with a small 5v fan running off the 5v out on the GPIO, however, you get 20 degrees cooler.

Any overclock setup should probably have a fan for the rpi3 in my opinion. Static air just isn't very good for heat transfer.

[mrbwa1]

Funny thing: I just logged into my Pi to make sure I shut down emulation station yesterday and noticed it is idling at 52 degrees (with Emulationstation running) sitting there doing nothing. That means it idles hotter than it runs maxed out with the fan added.

[Arcticade]

I have had my PI3 idling on emulation station AND transferring data over wifi for 12 hours right now. My CPU is 36 Celcius/96 F (very low). I have heat sinks and a kick ass fan installed. For those of you wonderinf if your heat sinks are working. Touch them. If they are hot, they are working. If your CPU is at a high temp and your heat sinks are just warm, they aren't.

[Drob10]

Just to chime in that I had this exact same issue running Pi3 from a canakit with the heatsinks attached. Pi idled around 65c and as soon as I removed the cpu heatsink the temps dropped at least -10c.
I just switched cases, which came with another heatsink. The pi, just running emulationstation with new heatsink, is at 49c cpu and 48.3c gpu.

[wolfgangrumpf]

For what it's worth, I have the same basic setup as OP described (RPI3 Canakit with thermal tape holding on heat sinks). My temperatures were getting VERY high (75 - 80C!) in the miniature NES case I printed up for it, so I went ahead and installed a fan at the bottom of the case that blows up directly onto the heat sinks - now my temperatures rarely exceed 55C. The fan makes a bit of noise but not enough to be distracting in a quiet room, really.....

[Capeman]

The pi isnt designed to need a heat sink. If you look at a typical X-86/64 processor versus an ARM processor, you'll notice the material housing the processor on an X-86 is metal and really smooth - that design is intentional - the metal cap conducts heat to the surface which transfers to thermal compound then to the heat sink which is then blown away by a fan.

An AMR processor you'll notice is the same material as a standard IC, its plastic capped with a textured surface and printed markings. The textured plastic is terrible for heat sinks. Tape will leave microscopic air bubbles between the plastic and the tape, air is the best insulator. The last thing you want between a CPU and a heat sink is any layer of insulation. Only a really squishy tape OR thermal compound will have any sort of effect if well worked in... but because of the basic material housing of the ARM processor, it is not designed to be heat sinked.

If the pi was meant to have a heat sink, the foundation would have included one or at least put a mounting solution in for one.

[caver01]

@capeman I agree with this to a point. The assessment of the materials, the texture-induced air pockets, and most importantly, the failur of “tape” to properly conduct heat are all important points. I think the take-home is that if you are serious about getting the most out of a sink, you should be using a thermal compound, not self-adheasive tape to apply it.

Yet, once applied properly, even a sink without a fan can do wonders. As for the Pi Foundation not intending us to use a heat sink–this is true, but only because they have the throttling mechanism built-in. If we simply allow throttling, sure, no need to ever install a heat sink. But most of us want to push the heat envelope and maximize performance.

One of the problems with all of the anecdotes about only getting 5 degrees of effective heat reduction without a fan etc. is that most of us are using different solutions. We have different heat sinks, different application methods, different cases, and so on. It’s apples and oranges in many cases. Still, what we do know is that there are paths to success. Some use fans. Some don’t. I don’t, but I have a good sink with Arctic Silver. The details really matter.

[Capeman]

Given the negligible differences in degrees with and without a heat sinks that most people report, id really like to see somebody test a heat sink on the ARM versus just pointing a fan directly at the ARM with no heat sink to see if simply moving air across the bare board would reduce heat in a similar way. I'm almost willing to bet that it will.

Since most people get those cheap sinks with tape backers, I'm almost sure only about 1 in 10 people are applying these things in a way that would be useful (meaning thermal compound and pressure with a fan)... because simply sticking these things on with the included tape is more likely to actually raise temps versus helping lower them. But they definitely look cool haha

[caver01]

@capeman I absolutely agree. The test you suggest would be interesting. I’d like to see someone do the test you describe, try a heatsink using the included tape, then try again with thermal compound. It is the combinations of all of these options that is giving everyone a different result I imagine, on top of the fact that we all have different interests, so the tests are not even the same. Where I might show performance throttling in Golden Tee (arcade) someone else is testing with PSX.

[Capeman]

@caver01 my guess is that the properly seated sink with thermal compound and a fan would definitely perform the best, next would be the bare board with a fan blowing on it. I guess that the cheapo tiny sinks with tape or adhesive backing would actually raise heat in most cases, most of those seem pretty cheaply made and appear to be more for style than anything else.

[Hex]

There is active development going on to address this issue.

https://retropie.org.uk/forum/topic/11304/testers-needed-power-saver-features-pr-172

[Capeman]

@hex Very interesting thread and awesome progress! I must agree that nobody reads those stupid descriptions and it's casualty worth sacrificing, i use a batch method to create gamelists that don't even include it.

[steptoe]

I used thermal heatsink glue, basically heatsink paste with epoxy already mixed in. Left overnight the heatsink is solid

I bought a big heatsink of eBay and cut it down to just fit over the main CPU leaving a little bit over overhand and also mounted a very small fan on top (superglued on but it did the job). Also the memory chip undernaeth is mounted on a heatsink as I'm led to belive that can get pretty hot when overclocking

These are the settings I used with an RPI3 :

#Overclock Settings
arm_freq=1350
over_voltage=6
sdram_freq=500
temp_limit=80
core_freq=500

disable_splash=1
force_turbo=1
boot_delay=1

#GPU Based
h264_freq=333
avoid_pwm_pll=1
gpu_mem=450
v3d_freq=525
gpu_freq=400

#Ram Overclock
sdram_freq=575
sdram_schmoo=0x02000020
over_voltage_sdram_p=6
over_voltage_sdram_i=4
over_voltage_sdram_c=4

Temperatures barely move over 45c regardless what heavy stress tests I used and left for a good 20-30 minutes testing it to the limits. With the fan off it climbs to about 10-15c more so the fan does make a difference. Its not in a case though as its mounted inside my desktop arcade cabinet

Maybe once I build a media player/KODI mounted in a case attacged to the back of our TV on a mount I might have to think differently