Page 1 of 1

opinion about heat problems

Posted: Thu Apr 16, 2015 7:46 am
by chourizo
Hello,

I am having some problems with heat with some Netburner Nanos. I am running them in sealed enclosures, and not heatsinking them. I know the specified temperature range is up to 85C, but I guess it also depends on the app that you are running.

When the enclosures get hot (50-60C) I start having network problems. The ping reply time goes to 150ms (<1 in normal conditions), and in general the whole network (not only the Nanos) is seriously affected.

I have seen that happening in two units with different software, but the same conditions, a hot sealed enclosure (50-60C). As soon as you open the enclosure and cool it down, it works fine again.

Has somebody seen that before? How do you deal with heat? I am considering two options:

1. heat sink them. Not easy, but I'll do it if I have to.
2. underclock. Never have done that, not sure if it will make a real difference.

Any comment will be appreciated.

Re: opinion about heat problems

Posted: Thu Apr 16, 2015 9:54 am
by rnixon
Is the enclosure metal or plastic?

Re: opinion about heat problems

Posted: Thu Apr 16, 2015 9:57 am
by dciliske
I have not heard of any issue with this before. What's the length of your cable ethernet cable?

Also, what else do you have inside this enclosure? By the sound of it you may have an internal network there. Is there a switch?

Re: opinion about heat problems

Posted: Thu Apr 16, 2015 9:59 am
by rnixon
Also, how are you measuring the temperature? The rating are usually for the junction temperature of the IC, not the ambient. So if the ambient is 60 or higher, the junction may be over 85.

Re: opinion about heat problems

Posted: Thu Apr 16, 2015 11:00 am
by chourizo
The enclosures are all metal so I might be able to use them to drive the heat. Yes, there is a switch inside every enclosure.

The network cables are short, all of them. Less than a foot from the switch. All the components are good quality.

I am measuring the ambient temperature inside the enclosure. It is industrial range, so the working ambient temperature is -40 to 85C or that's what I thought.

Any experience reducing the clock frequency to reduce the heat?

Thanks for the comments.

Re: opinion about heat problems

Posted: Thu Apr 16, 2015 1:05 pm
by dciliske
Ok, the reason why I asked about a switch is that your latency went up and you said everything was affected. This sounds like the issue is the switch, and not the Nano.

As for heat sinking the module, perhaps for the time being, consider getting something like this and sticking it on the enclosure wall. (Note: I have never actually done this, but I did just order one to test with). Also, depending on your chassis layout, while you may be measuring 60 C, you may have hotspots elsewhere where the ambient is much higher.

No, I don't think anyone has ever underclocked a Nano to reduce power usage. While I think all the drivers are written such that you could dynamically change the clock speed, you would need to ensure that the DDR bus still meets the minimum timing specs for DDR RAM.

Re: opinion about heat problems

Posted: Thu Apr 16, 2015 8:48 pm
by chourizo
Thanks for your comments.

We will continue with more testing, but so far the problem seems to be quite repeatable and consistent. We will try heatsinking and hopefully the network problems go away.

I have to say that the problem only shows up when the enclosures are hot (50-60C) AND we have a busy network (normal condition of our system). If we remove most of the network traffic everything looks normal and the lag goes away. When it's cold (maybe the first hour after turning the system on) there is no problem.