Guruplug overheating?

[unaiur]

First generation copper PHYs used about 5W per port. Current generation, between 1 and 1.5W. The most modern PHYs which are able to detect cable length and adjust TX power, between 0.3W and 1.5W depending on cable length and quality.

But the problem isn't the heat generation but the heat dissipation. It seems that GP+ has very bad thermal design, the heat accumulates and the temperature raises. I've an Alix board in an aluminium chassis in which, while consuming 12W, you can touch every component in the board with the hand and they are barely warm (about 30 Celsius degrees).

And about PHYs in 100Mbps mode, it can be 0.1W or less. Remember that power consumtion doesn't scales linearly with the frequency.

[mola]

Result: When idle, the GPP is at around 57°C and under stress goes up to 64°C.  

..right - small update as I went for a bigger passive solution. Filled cavities in original heat sink with copper, grinded the sink so that it is even, glued a pretty massive copper shim to it (after adding M3 threaded holes) and screwed the assembly into an old hard disk cooler. PSU and JTAG module fitted as well.

Result: It doesn't get warmer than 38°C anymore (hottest point I measured inside). Outside max is 35°C. And this was measured after 1 hour massive stress using:

guruplug:~# stress --cpu 8 --io 4 --vm 2 --vm-bytes 128M --timeout 3600s

(NOTE - don't try that with an unmodded GuruPlug or you'll almost certainly kill it. I have never seen such a load of >16 before...). Rest unchanged, i.e. 1Gbit eth0, eSata drive, microSD, 1 USB device, no BT/WLan.

Pics below - I guess I'll leave it like this (even though I will of course look at GlobalScale's official solution). Case is open at the back though because I want to access the SD card and am too lazy to do the back properly. In particular since this thing will sit in the basement and replace a server in my 19" cabinet.