So I have finally exhausted all the methods/suggestions that I found/was told about on the forums, wiki, and irc. The situation is this:
I have a new Asus eeePC 1015BX which has an AMD C-60 APU (that is, CPU+GPU) where the GPU is a Radeon HD 6290 (and, in case it matters, the audio is Radeon HD 6250/6310) which, depending on the cpu govenor and load, runs at anywhere from ~80 to ~90C! (confirmed with acpi -t, sensors and my fingers) This is very hot, and worrying indeed (emergency shutoff is a mere ten degrees away). I have put this down to the GPU running at full speed constantly (though I am not certain) I can confirm that the GPU clockspeed is changing when I use dynpm by checking "/sys/kernel/.../radeon_pm_info". I would really like to use this netbook, but I don't for fear of meltdown
As for what I have already tried/determined:
-From a cold boot, by the time I log into the tty, the temperature is already in the high seventies and climbs quickly from there (and never drops below ~78C, even on virtually zero CPU and GPU load).
-Every kernel avaliable in ARM (from 3.0.x through to the current 3.6.8 or whatever it is at the time of writing) as well as the LTS and even a 3.7.rc1 (or some such) and none of that seems to have any effect (there were Sandy Bridge folk reporting a change with that, but I am on AMD and it seems the causes are unrelated) [this was a painful process]
-Turning the GPU semi-off by installing acpi_call and running test_off.sh to check if there were any switches I could change. There are none.
-Turning on power management (of the most aggressive sort) using the echo dynpm > /sys/class/drm/card0/... trick as outlined in the wiki.
-The catalyst fglrx module and the ati-dri one (with and without kms).
-Plain, old reinstalling from scratch (which seems futile given that the live image on the install medium suffers this same problem too).
-All the relevant eee-related stuff in the AUR doesn't work on my particular model, and so has no effect.
Note that cpu scaling seems to have an effect on the temperature (albeit a small one) and so I don't think that that is the issue, and also that this is not a hybrid machine - there is no other GPU. Any feedback whatsoever would be greatly valued. Oh, and if anyone needs more info on the machine or whatever the case may be, I'm happy to provide - I just gave what I thought was necessary to understand the problem.
Last edited by Hiato (2012-11-30 11:38:39)
So, after exploring (nearly) every possibility at my disposal, I found that using kernels 3.5.7 (or thereabout) and 3.6.8 + proprietry catalyst driver + c6 mode enabled in bios + cool and quiet disabled in bios + X server running, I could get the temperature down to ~69C if I let the netbook idle for a [long] while. This is still unacceptably high, but the lowest I have found thus far. For some unapparent reason, the fan shuts off almost entirely at this temperature and so I suspect that there is a fan speed issue on this netbook - if it continued at its highest setting, nay, any setting at all, I have no doubt that the netbook would cool down to a healty 30-40C, but this is not the case (under any configuration).
Interestingly, the open source drivers (ati-dri and even ati-dri-git) produce a temperature about 10-12C higher than is achieved using the fglrx module - despite the fact that both scale the GPU properly. This is incredibly frustrating. So, I suppose a stop-gap solution would be direct control of the cpu fan and wedge it to maximum (though I have no way of telling if that would work). However, sensors does not report any fans and installing the linux-firmware module did not change the situation. Any thoughts?
I manage to solve the problem. First of all you should not be so afraid of temperature. The critical temperature is for weird reason by factory setting 98C.
So 85C is no problem, and when it comes to 98C, eee shuts down without any further chineese syndrome taking place ;-)
By the way, my problem was that the eee actually closed down during normal usage due to overheating. Annoying!!! What seems to solve the problem was jupiter. Installation on ubuntu is explained here.
http://askubuntu.com/questions/124511/h … u-temp-low
My hunch to why jupiter solved the problem was that jupiter, contrary to the default ubuntu installation, found the correct way to use the screen. I seems that the wrong screen settings involved overload of resourses and overheating.
Anyhow now I can run a video in gxine, AND a youtubevideo in firefox, while having libreoffice open. This should be a mean combination, but the fan was able to reduce temp. to under 80C.
However, as I tried a lot of other funny things, these might have contributed to the solution. Installing fgxlr, ati and sensor explained in the links below suggest this.
Right now the cpu-temp is 78C, and it works fine :-)
Hope this will help you. Now you know that there IS a solution.
I forgot to say that I never managed to regulate the fan rpm manually because digifan (from the packet digitools) was bugged in ubuntu, and because pwmconfig was bugged or ??? But, nevertheless, the installation of sensors may be important for the automatic fancontrol???
You should also know the fglrx packages that I have installed. The catalyst control center that you should be able to find in your programmenu after installation, may have been crucial as well for using the Radeon graphics card.
ii fglrx 2:8.960-0ubuntu1.1 Video driver for the AMD graphics accelerators
ii fglrx-amdcccle 2:8.960-0ubuntu1.1 Catalyst Control Center for the AMD graphics accelerators
rc fglrx-experimental-12 2:12.100-0ubuntu0.1 Experimental AMD binary Xorg driver and kernel module
One day more clever: Big surprise - it is the driver which is crucial. Neither jupiter nor sensors have any influence on the heating. The fglrx drivers and catalyst fix the problem. BUT, a new wellknown problem turns up: no rewake from suspend!!! But, the latest driver from Amd 13.3 has solved both issues.
http://support.amd.com/us/gpudownload/l … linux.aspx
Last edited by flemmingbjerke (2013-04-08 21:03:39)