When I first got the machine, it quickly turned out that its fan unit was broken. This was indicated by a whirring noise in the fan, which turned on very often, and the machine crashing with a "critical temperature reached" message under Linux. Under Windows, it would simply be throttled to minimum CPU frequency. An email exchange with another TM 380 user convinced me this was not a Linux compatibility issue, but bad hardware.

Fortunately the problem was resolved by sending the machine back for repairs. Many thanks to both the reseller and Acer for good service - as it turned out they would not be able to repair it in time for an important trip I'm about to embark on, I was sent a replacement, which works correctly.

Second problem is related to the Xorg driver for the Intel 852GM graphics chip, blocking me from using a dualhead setup. If an external monitor is plugged in during boot, I get image there, otherwise on the laptop's own LCD only. Under Windows, dual-head works very well, so this is a limitation of the i810 Xorg driver.

Unlike last time, I admitted that I probably will need to use Windows some day, in particular since I have to install the Motorola phone management software somewhere, and allowed the machine to prepare its XP partition. Then, like last time, I took parted and minimized the partition to 7 gigs.

For Linux, I went to bleeding edge with Fedora Core 4 test2, primarily because I wanted Gnome 2.10 from the beginning. Installing had no major hurdles, but took an extremely long time (estimated four hours, but I didn't stick around to watch it), apparently because the installer kept slowing down long enough for the DVD drive to spin down, then spin it back up again, copy one package, wait a bit... and so forth.

After the first configuration reboot, and the first time Fedora was starting "properly", I did get some test-release related problems - namely, a bug in the xfs initscript prevented that from starting, and without a font server, X would not come up either. This was made more serious by the screen going blank, which I could not recover on any virtual console. A boot to single user mode to fix the initscript solved that, though, and the latest updates (towards FC4 proper) contained the fix.

The hardware consists of 1.6 GHz Pentium M 725 CPU, 1GB of RAM, 60GB HDD, external Firewire DVD+-RW drive, 12.1 XGA LCD and 802.11b/g wireless LAN. Of these, the basic functionality works straight away. Networking is configured automatically for the wired fast-Ethernet port, and the DVD drive required no configuration, although I have not yet tried burning anything.

WLAN setup is a matter of downloading the Intel PRO/Wireless firmware and installing it in /lib/hotplug (for Fedora Core, YMMW). GNOME's wireless tools are beginning to be usable, although setting the configuration profiles to support both a WEP network and an unsecured one still needs a lot of tweaking, none related to the hardware of the machine.

The display is a bit bitchy. The laptop's builtin XGA screen works fine, but I want to use an external SXGA LCD screen via the VGA port when on my work desk. I had scripts for autodetecting the external screen that worked fine on my old machine, although they require an X11 restart (Xorg does not grasp the possibility of screen number changing on the fly), but on this machine, it does not work. The only way I can get a signal on the external screen is to reboot with it connected, which is nasty.

I suppose it goes without saying that under these circumstances, a dualhead setup does not work either. I've configured everything as I understand they should be, but I get "SetDisplayDevices call failed", and apparently I'm not alone with this. i810 dualhead simply isn't as reliable as it is with Radeon or Nvidia. In any case, the following xorg.conf Device entry will at least permit use of virtual consoles:

Section "Device"

        Identifier  "i810"

        Driver      "i810"

        VendorName  "Acer"

        BoardName   "Intel 852"

        BusID       "PCI:0:2:0"

        Screen      0

        Option      "VBERestore" "yes"

        Option      "DevicePresence" "yes"

        Option      "CheckDevices" "yes"

        Option      "XvMCSurfaces" "6"

        VideoRam    65536

EndSection

Update: I have limited success with dualhead, see my later entry.

As I mentioned in the beginning, the first unit I received had a problem measuring and controlling its temperature, with the ACPI system temperature monitor fluctuating wildly between 47 and 85 degrees, the system fan spinning up and down with the slightest change in system load (typing something in OpenOffice is sufficient to create "load"), even while the CPUFreq ondemand governor keeping the CPU speed at 1-1.4 GHz (max 1.6). 85 degrees being the critical temperature, of course, causing the kernel to perform a panic shutdown.

After trying CPUSpeed and finding that it had difficulties parsing the 2.6.11 ACPI files, I developed another workaround in the form of the following tiny shell script which I use together with the kernel ondemand governor to throttle the CPU if the machine's temperature rises. A function like this would probably be a good addition to the ondemand governor itself, but should be unnecessary with good hardware.

#!/bin/sh

cd /sys/devices/system/cpu/cpu0/cpufreq

while :

do

   read nop temp nop < /proc/acpi/thermal_zone/THRS/temperature

   [ "$temp" -gt 75 ] &&

     cat cpuinfo_min_freq >scaling_max_freq ||

     cat cpuinfo_max_freq >scaling_max_freq

   sleep 1

done

Comment by oa on Fri, 12 Aug 2005 16:34:27:
filed bug https://bugs.freedesktop.org/show_bug.cgi?id=4057 regarding the multihead problems