Dell Inspiron 6000d
Components: Graphics

James F. Carter <jimc@math.ucla.edu>, 2005-03-30 (updated 2005-12-10)

Component Details: Graphics

Graphics

ATI Radeon Mobility X300 (M22) 5460 PCI Express. (The model number may be reported as M300.) There is 64 MB of video RAM with the chip (128 MB is an ordering option). Motherboard features include a VGA port for a projector, and S-video out to a TV. The Radeon driver of Xorg-X11 (v6.9.0) recognizes the chip; Xorg-X11-6.9.0 can do more acceleration than previous versions, but still can't do direct rendering (3D acceleration).

Thanks to Markus Weiss for pointing me at the ATI proprietary driver hosted (at that time) on SuSE's site. However, since 8.16.20 you need to use ATI's own installer to create RPM packages. See the HOWTO. When you navigate on ATI's site, beware that it uses Macromedia Flash heavily and is Windows-centric. Firefox can handle the site.

I installed this driver following their instructions. As of 2006-06-19 the current version is 8.25.18. I had no end of trouble with their installer script, and I ended up transferring to their Linux download page and directly installing the RPM for SuSE 10.1. Salient points include:

• Being proprietary, the driver it will taint your kernel. Ignore the warning.

• Per the FAQ, the driver (older version 8.6.20) needs the OpenGL library provided by ATI. Add /usr/lib/fglrx/lib at the beginning of /etc/ld.so.conf and run ldconfig, to force ATI's library to be used. 8.25.18 does not come with a special library; revert this hack from your /etc/ld.so.conf.

• Unfortunately, software suspend and fglrx-8.12.10 (old) have a bad interaction. When the system resumes and tries to reinitialize the graphics driver, it hangs. Starting with v8.19.10 and kernel 2.6.11, suspend to disc became useable (but see below). Starting with v8.25.18 and kernel 2.6.16, suspend to RAM finally became useable.

• Starting in Fall 2005, fglrx-8.19.10 (or possibly earlier versions) overcame the compilation issues with the 2.6.11 and later kernels, and also the software suspend problem was fixed. Now it is completely reliable, with one minor glitch: if the X-server is terminated while 3D graphic operations are in progress, the video RAM will be corrupted and the machine will lock up solid, not even responding to the magic SysRq key. You see this when you are trying to shut down the machine. Careful tuning of timeouts (so the SIGTERM happens when the screen is static) minimizes the problem. If you use a journal filesystem (ext3 or Reiser) you will thank yourself; also be sure to do "sync" as the last thing before beginning the shutdown process.

  As of version 8.19.10, suspend to disc never failed for me. As of version 8.25.18, I did a formal test of the new suspend to RAM capability, and 10 successive suspensions worked properly. However, one suspension in normal use failed (screen blank), but I can't be 100% sure if the Radeon driver is to blame. I did find that if you flip to VT 1 and then suspend to RAM, when you resume the screen will be blank and you can't flip back (or maybe you can but the screen will still be blank). But the rest of the machine will be fully operational. If you find yourself in this situation, try Ctrl-Alt-Delete and see if it will do a normal shutdown; if not, try the magic SysRq key to shut down.

Here are some speed comparisons with other systems. Glxgears is often used as a speed benchmark; however, I think the old Mesa demo fire (from SuSE 8.2, or compile it yourself) puts a heavier load on the 3D system for a given CPU usage. Fire was run with fog and help text turned off. Quantities shown are in frames per second. A line is shown for comparison in which the X.org Radeon driver is used, with software 3D only. Reported units are frames per second.

Hostname	Video Card	Ware	glxgears	fire
Xena	ATI Radeon Mobility X300 (PCIE)	Soft (6.8.x)	216	0.64
Xena	ATI Radeon Mobility X300 (PCIE)	Soft (6.9.0)	744	1.60
Xena	ATI Radeon Mobility X300 (PCIE)	Hard (8.12.10)	832	116
Xena	ATI Radeon Mobility X300 (PCIE)	Hard (8.19.10)	1211	158
Orion	ATI Radeon Mobility M6 LY	Hard	663	78
Baobei	nVidia GeForce2 MX 400	Hard	1578	181
Fafnir	nVidia GeForce MX440 AGP 8X	Hard	1864	240

The X300 performs reasonably, but can be outrun by a good desktop graphics card. Reviews indicate that the X300 is excellent for normal graphics, e.g. photo editing or CAD-CAM, but it is a bit underpowered for the serious gamer. The ATI Radeon X800 (and a correspondingly more powerful CPU in a desktop machine) gets about twice the frame rate as the X300 in a number of current games.

I have a report that the Inspiron 6000 with Intel graphics and BIOS A08 (likely starting with A06) has a conflict with the X-Windows driver so that under X the screen is completely black. It was fine with BIOS A05. I haven't heard that he fixed the problem, and I don't have any more details about his configuration. BIOS A08 did not poison my machine with the ATI Radeon M300 graphics chip.

Screen

Liquid crystal display, 1680x1050 pixels, by Samsung. Characteristics match the model LTN154P1. Here are some characteristics from Samsung's data sheet:

Resolution	1680 x 1050 dots, 3 pixel/dot
Format	WSXGA (wide, 8x5)
Size	331.4 x 207.1 mm (15.4 inch diag)
Dot pitch	0.197 mm/dot, 128 dot/inch
Number of colors	218, 6 bits/pixel
Color gamut	45%
Brightness	185 nit (candela/sq meter)
Contrast (white/black)	500:1
Response time	25 msec (at 25 C)
View angle (up-down-l-r)	50, 50, 65, 65 degrees
Mass	590 gram
Power	6 watts (lamp bright)

Each dot has a separate red pixel, a green pixel and a blue pixel. The color gamut refers to the range of colors that the panel can achieve, and the percentage compares this with the range which a theoretically perfect color source can produce. For brightness, typical sky brightness is 8000 nit. On a partly cloudy day with the sun directly on the screen, it can still be read, barely.

The ID string is H4700^B154P1. With the control-B this is bogus, and I ended up telling the installer, successfully, that it was a Dell 1600X Laptop Display Panel. As usual with flat panels, the refresh rate is 60 Hz. Xorg-X11 got good mode timings without any trouble.

The screen looks good. Its view angles are ±65 degrees horizontally and ±50 degress vertically, much better than the screen on the Inspiron 4100. These numbers come from the specs, but my experience with the screen bears them out. The screen on the Inspiron 4100 had quite a bit less view angle, and I appreciate now being able to tilt the screen to avoid glare, or to show something to a co-worker without having to worry about color rendition.

Dell also offers panels at 1280x800 pixels (WXGA) and 1920x1200 (WUXGA). The WXGA is old technology similar to what's on the Inspiron 4100, with a smaller viewing range which I found somewhat annoying; I think the WSXGA panel was well worth the extra US $100. As for the WUXGA, I don't really need the extreme resolution, and there have been various non-quantitative statements in web reviews that people (never the author of the piece) think the screen looks dirty. It would probably be a good idea to eyeball the WUXGA screen before buying one.

If you use screens with a variety of resolutions, you will thank yourself if you specify font sizes in points rather than pixels. The X Window System generally gets the sizes right if this is done.

A LCD panel can be cleaned with a lint-free cloth or paper dampened with plain water or, for stubborn globs, alcohol. While the body of the panel is glass, the front surface is a thin polarizing layer made of PVC, which is essential for function and is not rugged. Avoid touching the panel except for occasional gentle cleaning.

Touchpad

Alps Glidepoint with two buttons. Its resolution is 330 DPI and the active area is 73x43 mm, or 943x682 pixels (empirically). Some of the pad is covered by the upper deck, and likely the real dimensions are 1000x750 pixels. On Windows the pad is quite slow on the default settings, and you should use the Touchpad tab on the mouse configuration tool in the control panel to set it the way you like.

My distro, SuSE, has recently switched over from the XFree86 X-Windows distro to Xorg. I'm not clear on the politics here, but at the present time the X-servers have the same ancestry, and their configuration files are pretty much interchangeable. Below I refer to xorg.conf, but if you use XFree86.conf, you can use my file too.

On Linux I imported a xorg.conf file from the Inspiron 4100, which has a Synaptics pad with an active area of about 5500x2500 pixels, and the pad was totally unuseable. After a fair amount of work I got the pad into reasonable shape. See Peter Osterlund's driver page for the latest version, which as of 2005-03-28 is 0.14.1. There's no indication that the distro's version, 0.13.2, contributed to my troubles with the pad; also be aware that starting (?) with 0.14.0 the time unit in many of the parameters changed from per packet to per millisecond, where packets come out every 10 to 20 msec, so going from 0.13.x to 0.14.x you'll have to tweak the speed parameters again. The synclient utility from that package was helpful in experimenting with settings, but remember that the client and the X driver must match exactly in versions.

Here is my X-Windows configuration file (xorg.conf); see the first mouse section. I have fairly radical accelerated motion, so that a quick slice across the pad moves the cursor all the way across the screen, yet slow motion goes at 1/4 speed. Corner taps work. (You might want to adjust the edge boundaries so your finger fits reliably in each corner.) Drags in the left or bottom edges emulate the mouse's wheel (there are corresponding decorations on the pad). Edge coasting is turned on. I have just about the complete set of functions set up for this pad.

The remaining problem with the touchpad is that double clicks are weirded out. The issue is that the Alps pad has hardware tap detection, which involves a 100 msec timeout. If you want a double click, do it really slow, and configure applications to wait for it. Triple clicks are not possible until this is fixed. The SuSE kernel has a patch to detect and specially handle Alps pads, and it tries to turn off hardware tap detection, but this seems to be ineffective on some models including the one in the Inspiron 6000. The driver page has a link to an additional kernel patch ( http://marc.theaimsgroup.com/?l=linux-kernel&m=109104309904873&w=2), but it did not apply cleanly to the SuSE sources (2.6.8) with their set of patches. Here is a diff versus the actual SuSE original; however, before using it you should first check on the driver page to see if an even better kernel patch has been posted. This patch revives double clicking if the pad parameters are set correctly. (It is already merged into the official kernel 2.5.12-rc1.)

The touchpad is recessed 2 or 3 millimeters below the deck, and I have not had a problem with hitting it accidentally. Some other vendors' touchpads are flush with the deck, and in reviews the owners blame this placement for frequent accidental hits.