AMD Athlon 64 3000+, 1024MB RAM DDR, Hitachi DK23FB 40GB HDD, TEAC DV-W22E DVD-R/-RW, 15" XGA, 4xUSB (2.1), Intel 537 Modem, VIA Rhine-II 10/100 LAN, Intel PRO/Wireless 2100 3B LAN, VIA IEEE1394 (FireWire), 15" XGA Display, ATI Mobility Radeon 9600 NP, VIA 8235 Sound, ENE Technology Flashcard Reader.
Xeron have upgraded the WLAN in this machine: now they ship with the 802.11g chipset from Intel: ProWireless 2200 BG.
You may reach the Insyde BIOS with the F2 key.
This machine was installed with Debian in 2003-12 and with Novell/SuSE in 2004-04. Users of SuSE or recent kernels should read the Debian part first and then pursue to the SuSE part where all the new developments are mentioned.
Testing (Sarge) was installed via a Knoppix CD. Afterwards, kernels 2.4.23 (+ ACPI, SWSUSP, CPUFREQ, PPSCSI), 2.6.0 32-bit (+ ACPI, SWSUSP, X8664-1) and 2.6.0 64-bit (+ X8664-1) have been set up.
The following statements are valid for the kernel 2.6.0 32-bit if not stated otherwise.
Setting up the 64-bit system is a little tricky, checkout this document on alioth.debian.org(http://alioth.debian.org/docman/view.php/1314/21/debian-amd64-howto.html) on how to install a 64-bit system with Debian. Only a basic system is available in 64-bit (fortunately, nearly all 32-bit applications run in parallel). But you are free to re-compile the packages with e.g. apt-build and contribute. It is also important to note, that in 2003-12, you should debootstrap Woody and not Testing/Unstable. This is, because the 32-bit packages from Testing/Unstable are mostly newer than the 64-bit packages from alioth and you would have problems "downgrading".
After doing all that, how do you verify that you are running in 64-bit?
/proc/cpuinfohas some more entries:
TLB size : 1088 4k pages clflush size : 64 address sizes : 40 bits physical, 48 bits virtual power management : ts fid vid ttp
fileshows a different output on a binary you just compiled with
ELF 64-bit LSB executable, AMD x86-64, ....
But is it faster? Dunno, had no time to run a proper benchmark yet. Of course, the binaries get bigger. A simple "Hello world" takes about 1.5x more space in 64-bit than in 32-bit. The main aspect of using 64-bit now is to have the possibility to address larger memory areas, speed optimisations for PC platforms will follow in the time to come.
As far as I know, users of Novell/SuSE, RedHat and Mandriva (Mandrake) have it easier at the moment, because these distributions already ship in a version for AMD64.
APM is not tried.
ACPI seems to run fine,
ac_adapter, battery, alarm, embedded_controller, button seem supported,
processor only partly and
fan, power_resource and thermal_zone not at all. IRQ-routing looks sensible.
acpid receives a cpu event every two seconds, which seems strange.
The battery lifetime adds up to 2.25 hours in more or less idle mode.
SWSUSP 2.0-rc3 runs more stable with 2.4 than with 2.6, but runs with both kernels, submit
hibernate and freeze. When trying to sleep (S1 or S3), either via /proc/acpi/sleep or via /sys/power/state, SWSUSP tries to freeze the kernel, but does not succeed and returns to the command line. Well, SWSUSP should not intefere here, anyway, shouldn't it?
Sound works out-of-the-box with the driver snd-via82xx (ALSA).
radeonfb does not want to initialise the display, so the
vesafb seems to be the better choice if you want 1024x768 at 16-bit colour depth. Note that this observation is made with 2.4.23, for 2.6.0
radeonfb should be improved a lot. But the vesa driver also works better in 2.6 than in 2.4, as you have no garbage on your screen from time to time. So vesa will stay for a while.
The version shipped with Knoppix runs fine with driver
radeon, nevertheless a newer version from the DRI-trunk was installed. Unfortunately, ATI does not hand out the specs for their newer chips (>=9500), yet, so there is no 3D-acceleration at the moment. It is quite likely that they will, as they did before for other chips. The question is when they will do it.
No suprises here, take
via_rhine and feel happy. As for the WLAN: the commercial closed-source
driverloader from LinuxAnt works well. Recently, the ndiswrapper(http://ndiswrapper.sf.net) was announced to support Intel's chip, too.
ndiswrapper is as open-source as it can be, but this still does not relieve Intel from releasing a real Linux driver, ask them here(http://supportmail.intel.com/scripts-emf/welcome.asp) and sign the petition(http://www.petitiononline.com/xanthan/petition.html). :)
dvdrecord does not succeed in writing a DVD, since it is based on cdrecord 1.11a15 and its support for ATAPI-burning is very limited. And kernel 2.6.0 strongly discourages the use of
ide-scsi, you should use
ide-cd instead. Burning works fine with
cdrecord-proDVD which you can get for free as a private person for non-commercial use by the author, check the cdrecord page(http://cdrecord.berlios.de/old/private/cdrecord.html).
It is a Synaptics touchpad, but the little joystick does not work. Using the scrolling feature of the pad is not easy, because of the pad's round, rectangular structure, the vertical scrolling zone at the right border is quite small.
There are no extra keys, but some Fn+Fx keys which work for brightness, display switch and the bell.
When an external monitor is plugged in, the external display is distorted, which can be solved by using the switch button at least once. Afterwards, it works perfect.
There is a driver for 2.4 kernels, check Intel's site(http://downloadfinder.intel.com/scripts-df/Filter_Results.asp?strOSs=All&strTypes=DRV&ProductID=1230&OSFullName=All+Operating+Systems) for download. That driver does not compile with 2.6 and thus is not yet tried.
The serial driver via IrTTY works, FastIrDA - FIR is not tried.
There seems no driver available from ENE Technology. They still did not answer an email about Linux support that was sent to them some months ago.
Well, it is an Athlon 64 with lots of heat dissipation. The machine weighs about 5 Kg. So the fan is nearly always running (despite cpufreqd), even the power supply has a little fan built-in. The noise is not screechy or irregular, but steady and always audible. This machine is not made for watching DVDs in a silent environment, that is for sure! But the computer is really fast, boot messages fly by and compiling is incredibly quick.
00:00.0 Host bridge: VIA Technologies, Inc.: Unknown device 3188 (rev 01) 00:01.0 PCI bridge: VIA Technologies, Inc.: Unknown device b188 00:09.0 CardBus bridge: ENE Technology Inc: Unknown device 1411 (rev 01) 00:09.1 FLASH memory: ENE Technology Inc: Unknown device 0510 00:0a.0 Network controller: Intel Corp. PRO/Wireless LAN 2100 3B Mini PCI Adapter (rev 04) 00:0b.0 FireWire (IEEE 1394): VIA Technologies, Inc. IEEE 1394 Host Controller (rev 80) 00:10.0 USB Controller: VIA Technologies, Inc. USB (rev 80) 00:10.1 USB Controller: VIA Technologies, Inc. USB (rev 80) 00:10.2 USB Controller: VIA Technologies, Inc. USB (rev 80) 00:10.3 USB Controller: VIA Technologies, Inc. USB 2.0 (rev 82) 00:11.0 ISA bridge: VIA Technologies, Inc. VT8235 ISA Bridge 00:11.1 IDE interface: VIA [...] VT82C586A/B/VT82C686/A/B/VT8233/A/C/VT8235 PIPC Bus Master IDE (rev 06) 00:11.5 Multimedia audio controller: VIA Technologies, Inc. VT8233/A/8235 AC97 Audio Controller (rev 50) 00:11.6 Communication controller: VIA Technologies, Inc. Intel 537 [AC97 Modem] (rev 80) 00:12.0 Ethernet controller: VIA Technologies, Inc. VT6102 [Rhine-II] (rev 74) 00:18.0 Host bridge: Advanced Micro Devices [AMD] K8 NorthBridge 00:18.1 Host bridge: Advanced Micro Devices [AMD] K8 NorthBridge 00:18.2 Host bridge: Advanced Micro Devices [AMD] K8 NorthBridge 00:18.3 Host bridge: Advanced Micro Devices [AMD] K8 NorthBridge 01:00.0 VGA compatible controller: ATI Technologies Inc: Unknown device 4e50
In April 2004, another machine was installed with Novell/SuSE 9.0 for AMD64. It already ships with all binaries for the 64-bit platform. So installation is easy (if the double-sided DVD had not been bewitched: booting was possible, selecting packages not; after a net-install of the base packages, the DVD could be attached as installation device and then it could find the packages). The machine runs with kernel 2.6.5 + patch for X86_64-2.6.5-rc3.
Alas, running in pure 64-bit mode has its disadvantages. If you need all of the hardware, you should install a second system with 32-bit.
S1 wakes up at once, without really sleeping. S3 crashes the kernel in
Support for SWSUSP in 64-bit is really bad. PMDISK and SWSUSP2 do not compile,
SWSUSP does not seem to have any effect when issueing
echo -n "disk" > /sys/power/state.
This should not be a problem nowadays, but the kernel says:
powernow-k8: fid transition failed, fid, currfid a all the time when cpufreqd is run and at module startup it says:
BIOS error: currfid/vid do not match PST, ignoring.
/proc/cpuinfo shows only 800 MHz, though you can modify this value by echoing a proper value into
/proc/sys/cpu/0/speed.Unfortunately, the machine falls back to its old frequency only a few seconds afterwards. Having full speed at 2 GHz right from the start is a good thing, so the following was put into
modprobe powernow-k8; cat /proc/sys/cpu/0/speed-max > /proc/sys/cpu/0/speed; rmmod powernow-k8
Also, a patch from muru.com/linux/amd64(http://www.muru.com/linux/amd64/patches/patch-2.6.4-powernow-k8-buggy-bios) was tried but to no avail.
SuSE seems to have packaged a special 64-bit version (there is a directory
/usr/X11R6/lib64/) which unfortunately does not run with the
synaptics driver. Although it has to be admitted, that the kernel has some problems with the touchpad as well, since it complains about
synaptics reset failed twice.
Well, the open source driver from Intel does not support the new chip for 802.11g, even the support for the old 802.11b chip is limited. Ndiswrapper and Driverloader might be able to support it, but cannot do that at the moment. This is because they use the fine _32-bit_ drivers from MS Windows which, of course, do not run (easily) in a _64-bit_ kernel. See the X64-64 mailing list(http://www.x86-64.org/lists/discuss/msg04761.html) for a brief discussion. So the solution is to wait for 64-bit drivers for Windows which should happen this year or run a 32-bit kernel. Ndiswrapper does not even compile due to some assembler code used in it.
This is similar to WLAN. The shipped driver from smartlink ships with some binary which cannot run in a 64-bit kernel without proper support. Also, the driver does not compile due to some symbols missing.