For me it was 17 days since my last hub update/reboot. The telltale sign was the cpu usage spiked from average 4-6% to 22-25% that I’m assuming was increased garbage collection/freeing up memory. If only the hub had another gig of memory….
I haven't looked at CPU usage but I forgot about the heat(related?). @brianwilson pointed out his findings and my temps peak daily at about 194F. Right now my hub is 167F with an ambient temp of 69. It is vertical and free standing. Same place it has always been. I have 2 temps sensors near it and the ambient temp runs from about 68 to 72 this time of year. But, I haven't had any real slowdowns. Just the crash. I use Hub Watchdog and the numbers are fine. The last readings before the crash the other day were
Virtual 04-27 01:29 - 0.172, Zigbee 04-27 00:30 - 0.363, Zwave 04-26 23:59 - 0.258
The only oddity there is Zigbee is usually faster than Zwave but those numbers are miniscule. That aside though, it would appear to be a memory issue. I was in the UI the day before and didn't notice that it was slow or sluggish.
My hub watchdog numbers had actually gotten faster. I mostly rebooted to see if it was a memory issue causing the CPU usage and temp increase. I figured that a reboot would fix it if it was. I think I would start to twitch if my hub temp was 194F.
I'm on day 18 running my C4 stick on my C5 (only 11 days shown on graph because I'm now graphing the data out of Hub Information device vs. old way). No GUI slowdowns, memory is decreasing but it's dropped below 150k and didn't seem to impact anything. No CPU spikes that I can tell. If I was still running the same devices and stick on my C4, it would have crashed several days ago.
Only thing "new" on my C5 is the Hub Information driver as I was afraid the device was too "chatty" to put on my C4. But I'd rather graph the data from the Hubt information device via Maker-API/node-red and account for the "time" to get the information vs. querying URL's on the hub itself which doesn't account for load and such. Perhaps that's not a valid argument since Hub information is doing just that...
And my top apps for reference.
1 Like
Curious, how are you measuring the temperature? Is there a hub command or are your measurements external?
I have a C7 and admittedly there is not much load on the hub, but:
- My ambient is 70°F and the bottom of the hub is ~ 84°F.
- My hub is horizontal on a shelf. The top of the hub is cooler as expected.
I don't care if it catches on fire. As long as it puts itself out and keeps working. 
4 Likes
What @672southmain posted. But there is a line command for it. Sorry, don't know it off the top of my head but it's in the forums somewhere. Following many threads the C7 runs considerably cooler.
I rarely use a Dashboard, but I do always set them up/adjust them for new devices. 
Here's my display of the set of 4 hubs that are my Production hubs:
From the top down, there's a C-7, a C-3, a C-4 and a C-7. The C-3/4 series display much hotter temperature values than the C-5/7 series.
1 Like
The temperature on my main hub seems to spike during a reboot, but I've never seen it go above 149°. It averages around 113°.
Is this a C7? My current C4 is at 155, ambient 66.
Mine is a C5 with the ambient temperature at 70°.
Hmm... Maybe time to upgrade. I hate hot running electronics
1 Like
Yep mine (C-4) is a bit less @ 66 C
There probably isn't a "real" difference... I'm thinking it's a difference of a) sensor or b) proximity or c) CPU version. Very slight design differences/choices in how to manufacture the PCB. Perhaps one CPU has a temp sensor in the die, while another CPU is designed for it to be external.
They will sit in a somewhat narrow range however.
C-3's and C-4's have a heat sink on the CPU, but it's coupled through a piece of conductive material. The heat sink is screwed to the lid and the PCB is on the bottom piece of the case. Between is a blue heat conductor that is somewhat squishy. When I initially saw the temp differences in the reports between the C-3/4 and C-5/7 the image of that blue conductor came to mind. 
A C-4 from oh so long ago. The blue stuff (circled) is the material that transfers the heat from the CPU (square) and into the aluminium heat sink in the cover. (Maybe it's clear, but the CPU's impression in the blue material is offset. It isn't centered on the CPU. When I put it back together after taking these pics, I put the blue stuff on the CPU first and then jammed the lid back on. This particular hub is "retired" at the moment... I may never find out if adjusting the blue stuff improved anything. )
This C-5 has none of that
and yet it runs 'cooler'.
1 Like
I noticed that when I opened my C4 as well. However the purple pad is really thick and the "heatsink" is not aluminum. So any cooling will not be all that effective.
The thought came to mind that the "heatsink" wasn't for thermal management but a weight to keep the unit from being overwhelmed by the cable weight.
I don't know about the C7 internal cooling, but on my C7 bottom is warmer that the top (87 vs 83 °F)
I think a die temp of 194°F (90 °C) is well out of the expected range.
If one wanted to, a simple test would be:
- Turn the hub off for 2 to 3 hours,
- Restart reading the temperature as soon as possible.
- Temp should be near or at Room temp
Long ago, someone else mentioned the squishy pad and that caused me to open mine and take a look. I remember that post mentioning the name of the material, but I haven't been able to dredge it up from memory.
You could easily be correct regarding the value/use of that chunk of metal. I just popped it open again and it is ferris so not aluminium. 
Not great as a thermal sink, but better than air
Can't measure junction temps, but 90C die temp (if that's the reading we're getting) leaves some headroom...
All semiconductors have some specified safe upper limit for junction temperature (TJ), usually on the order of 150°C (sometimes 175°C ).
Likely dragging this our too far ..... but FWIW
Yes but, the temperature limits are largely determined by the thermal stresses in the die/package. An engineer from International Rectifier told me their MosFets will run a long time at 175°C (not near the temperature when the semiconductors start to change back to a chunk of silicon.)
What causes the failures (for the most part) is thermal cycling. This is where the stress is, especially in a power semi.
1 Like
