BosaikNet

I have three Dell Latitude C610s and a broken C640 (when powered up, it only stays on for about 30 seconds, decreasing over successive boots). The C640 and the better of the C610s has a SXGA+ LCD, the other two C610s have XGA LCDs. One of the XGA panels is broken, so I used it for practice. I use the better C610 for some 16/32-bit software, a task that will be relegated to the lesser of the working C610s.

Some time ago, I bought a cheap SXGA+ LCD to repair my Latitude D600. The motherboard promptly blew out on that laptop, so I was left with a nearly unused SXGA+ panel. This laptop, panel and all, sat in my closet for about 10 years, and I was careful to avoid compressing the laptop too much. Additionally, somewhere else around 10 years ago, I swapped the C640 and C610 SXGA+ panels to use the brighter C640 panel on the C610 after the C640 died.

Yesterday, I stumbled across an article about the broad compatibility of laptop display panels, which gave me the idea that the D600's panel is close enough to a C-series panel (all C600, C610, C640 panels are interchangeable) that I could benefit from the much higher brightness and color accuracy of this aftermarket panel. For reference, the XGA panels are 20-pin with actual pins, and SXGA+ panels are 30-pin with a flat one-sided connector.

The swap was fairly straightforward. I disassembled all four laptops (the two working C610s, the C640, and the D600) down to their panels. The better C610 was to receive the D600's panel, and the lesser C610 was to receive the better C610's SXGA+ panel and get an upgrade. I ended up breaking the C610's SXGA+ ribbon cable, as the plastic for the inverter connector had crept somewhat and near-fused the two connectors together -- the connector separated from the flex board with ease. I purchased a new old-stock SXGA+ display cable on eBay for about $15, which will be installed alongside the old SXGA+ panel in the lesser C610. In the meantime, I pulled the display cable out of the C640 as I needed one immediately. Since the inverters have specialty connectors, the inverter follows the display cable. The inverters attach to the LCD by different means between panels, even between the same model of laptop, so some creativity may be needed to safely seat the inverter. The D600's inverter was held in by now-removed double-sided tape and a single screw; some bending to the bracket and patience with a plastic spudger allowed the D600 inverter to come out without issue, and the existing bracket fit the C610 inverter minus the screw.

Note that the finish on the back of the panels is somewhat fragile. Lift up on the ribbon cable, and use a sharp razor blade to cut the tape securing the ribbon cable to the back of the LCD. Don't let the razor blade touch the diffuser!

The rest of the panel swap process was straightforward and largely without issue. The first test was with the panel plugged into the partially disassembled laptop -- the keyboard and display shell are not needed for testing, and the panel was plugged directly into the motherboard with no inverter connection. A bright flashlight revealed a POST display in the center of the screen, so the panel and laptop were properly communicating. Following the C640's disassembly and display cable swap, I was able to verify backlight was working, and then reassemble the display assembly. The screw holes in the LCD were a hair closer to each other than the holes in the hinges, so I opened the 2.2mm-or-so hinge hole up with a 3/32" drill to give me about 0.02" of bonus tolerance. The hinge material is a fairly hard stainless steel, so if you need to drill the hinge be aware of its mechanical properties.

While I was inside, I also swapped out the hinges for a nearly unused set I had purchased in 2014. The old hinges were cracked and held in place by a single screw on each side, and the force of a closing laptop magnified by a short lever could crack the laptop's frame; replacement was necessary.

Everything is working well upon initial testing. The display is very bright, about 200 nits, and dims to a comfortable level. Colors are saturated, unlike the C610/C640 panels. Some light bleed is present at the bottom of the panel, but for any computer this age I could care less. All of my panels exhibit it anyways.

I have ordered a SSD conversion kit (IDE to SATA bridge wired to a M.2 B-key slot) with matching 128GB SSD. The better C610 receives this kit, and I will be installing a dual-boot of Windows XP Integral Edition and stable Debian onto the laptop. It will also be receiving an AR9220-based dual-band 802.11n Wi-Fi card for WPA2 support; ath9k drivers also supply a great deal of functionality within Linux. I was unable to find any dual-band 802.11n card in mini-PCI with any Bluetooth, so Bluetooth is sadly not going to happen on this laptop. Cardbus is an option but I am not aware of any reasonably modern Cardbus Bluetooth adapters. A USB Cardbus adapter with a tiny BT5.3 adapter plugged in may be a realistic option here. Once these upgrades are complete, this forum thread will be updated.

Note that the LTN141P4-L04 had a lengthy production run and, despite being discontinued, is still widely available for a low price (under $75 new, $30 used).

Part numbers!

• C610 XGA LCD: Dell 03H471 / 3H471 Rev. A00, LG-Philips LP141XB (C1) XGA (C4). 20-pin.
• C640 SXGA+ LCD: Dell 03N383 / 3N383 Rev. A00, Quanta QD141F1LH01 (LK01). 30-pin. Dated March 2003. Panel rated 3.3V @ 1240mA, inverter 9-21VDC @ 595mA (probably 16V from battery system).
• C610 SXGA+ LCD: Dell 0695EM / 695EM Rev. A02, Sharp LQ141F1LH02. 30-pin. No date, likely early 2002 or very late 2001.
• D600 SXGA+ LCD: Dell 0HF239 / HF239 Rev. A00, Samsung LTN141P4-L04. Additionally labeled LTN141P4/U2. Dated 18th week (April) 2009. 30-pin.
• C610/C640 SXGA+ cable: Dell 059MNK / 59MNK, alternately labeled DD0TM7LC559. They come from different vendors, with Foxconn being the original supplier of my replacement cable.

Thermal pads arrived today, so today's work is thermal management. I'll start with a mechanical description of the thermal assembly.

The original thermal pads are white silicone pads with an aluminum-adhesive backing applied against the metal heat spreader they sit on (Dell part number 5H627 / 05H627). Both pads are 15mm x 15mm and are 1mm thick. The GPU thermal bridge is a H-shaped piece of aluminum extrusion attached to the plate by thermal epoxy. The contact surfaces (the sides of the H) are 16mm long (along extusion) by 18mm height (height of the H shape). The H-shape's width, equal to the separation between the GPU thermal pad top and the bottom of the metal plate, is 0.250 inches -- I measure 0.252 but adhesives add height. The center bar of the H shape is 6mm across and the plate features that make up the thermal contacts are 1.5mm tall. If this piece were to fall off somehow, I expect 0.25" bar stock to work as a replacement.

Still working on the metal heat spreader plate, I removed both thermal pads using a razor blade, then prepped the thermal contact surfaces by cleaning them with ethyl acetate (acetone would work too) and cotton swabs to remove adhesive residues. I then cut my 1mm thermal pad sheet to size, removed one layer of plastic backing, and applied the pads to the metal plate. Before continuing, I began work on the CPU (and spilled much of my 5mL vial of ethyl acetate on the work table).

The CPU heatsink should be the same across the C6x0 models -- its Del part number is 9D422 / 09D422, and it also has Auras part number 920711003-0725. Inside is a metal plate, inside of which is a short thermal pipe, that sits on the CPU at one end and goes to a heatsink on the other end. The heat pipe does not attach directly to the CPU but rather through the cast aluminum heatsink assembly, so the poor idle thermals make sense. A devoted machinist could probably make a better heatsink. Anyways, the heatsink was cleaned with ethyl acetate to remove previous thermal paste, and the CPU cleaned with 99% isopropyl alcohol for the same effect with less severity. All surfaces (heatsink thermal contact, CPU top, and the tops of the i830 and the Radeon M6 chips) were wiped down with clean isopropyl alcohol to remove any smeared paste residues. I removed the remaining plastic film from the heat spreader thermal pads, then reattached the heat spreader plate; I did this first as it helps locate the CPU heatsink. I then applied thermal paste (from my final gram of Arctic Silver 5), replaced the CPU heatsink, and reassembled the laptop. As I did this on the lesser C610, I left the hinge covers off as I need to replace the increasingly scarce hinges at some point. I repeated the process on the better C610, and will report back on thermals after the Linux adventures.

While working on the better C610, I decided to drop the C640's Radeon 7500 in the C610 just to see what would happen. The system attempts to boot, but because BIOS doesn't know how to talk to the card, no video shows up.

Linux attempts have been an adventure into changing views on obsolescence in the face of new platforms becoming the norm. In particular, SSE2 support and the graphics chipset have been limiting factors.

I started with Debian as it is the operating system I am most familiar and comfortable with. The current release of Debian, Trixie, drops official kernel support for i686. This is something I would call an omen. The previous release, Bookworm, retains support for i686, but cannot be simply upgraded to Trixie due to the lack of a prebuilt kernel. I went with Bookworm just to see how well the system ran.

The graphical installer, as well as the task-desktop pseudo-package that is automatically installed, requires SSE2 to work properly, which wasn't disclosed to me upon downloading. GRUB also failed to install, but it correctly repartitioned the hard drive and didn't break the Windows bootloader. A reinstallation using the 'advanced' text installer worked perfectly well, and I built it without a desktop environment until I could pin down what packages are breaking on a non-SSE2 system.

I chose the LXQt desktop environment due to it being lightweight and modern. The lxqt package installs a bunch of other recommended packages alongside it, many of which require SSE2. This was unacceptable, and I was willing to accept a limited desktop experience provided I could build and/or install packages further down the road. There's a package called lxqt-core, which installs just the core user interface without such things as themes or applications; it must be installed using the --no-install-recommends switch to avoid grabbing additional libraries. After installing this (and sddm), I was left with a sluggish but functioning GUI.

The 'sluggish' part is new. Having used Linux on these back in the day, I knew that I should have a better experience than what I was getting. It didn't take a huge amount of research to find out that the rendering engine under the hood is Mesa, and Mesa had dropped support for the ATI R100 (Mobility Radeon 7000, among others) and R200 (Mobility Radeon 7500, among others). The R100 chipset is used in the C610 and the R200 in the C640, so both laptop models have a fundamental hardware limitation despite the C640 having SSE2.

A branch of the Mesa subsystem called Mesa-Amber exists to maintain support for these chipsets by (to my limited understanding) implementing an older version of Mesa running in a separate namespace to handle legacy interactions. In theory, Mesa-Amber is installed in parallel with Mesa, and Mesa-Amber will provide the R100/R200 drivers to Mesa. Mesa-Amber is apparently not offered as a package in all Linux distros, and Debian notably does not provide such a package. They do, however, provide a source package that can be built using debuild. Perhaps due to DRI2 support being yanked from Mesa, or some other legacy code bit, I was unable to get Mesa-Amber to coexist with Mesa. APT began freaking out when I removed Mesa and installed Mesa-Amber, and it yanked my entire GUI system as it believed there was no good copy of Mesa installed (which, due to the separate namespace Amber lives in, is somewhat expected). After giving up on Mesa-Amber after a few more rebuilds with the same behavior, I instead downloaded, compiled, and installed Mesa 21.3.8-1, then froze the packages to prevent an upgrade into Mesa 22 or above. This didn't work either, because apparently the R100 driver wasn't built at all for some reason I do not have the knowledge to answer. It was a different error message in glxinfo -B, which gave me false hope for a few more hours. Being stuck with an old Mesa version also wasn't good for installing package versions with later dependencies, and it began to feel like I needed to start compiling the whole OS from scratch. And if the answer is compiling the whole OS from scratch, there are better distros for that.

There were some rules for picking a distro: it had to be reasonably current, it had to support i686 without SSE2, and it had to provide a working Mesa-Amber package. I then had to look at other distros:

• Slackware: It didn't appeal to me, as it requires technical knowledge of Linux (and indulges in this fact). Documentation is too scarce for me to feel comfortable tinkering with it.
• Arch32: A community distribution, appears to be largely unmaintained as of late.
• Manjaro: it meets all requirements, and is a choice. Not confident in its weak PC performance.
• Chimera: doesn't support i686, and probably would require SSE2 if it did.
• Adelie: claims legacy hardware support as a pillar of the distro, and is a choice. Fairly new, lean, and lacking in available packages.
• Gentoo: difficult to set up, yet among the most popular.

I selected Gentoo as my first choice since it is relatively popular, and having it compiled against hardware has a performance benefit. I might regret it later on, but we'll see -- reviewers are very clear it is only for expert users, especially on old hardware. My particular choice of Gentoo is systemd-based since I am familiar with systemd already. My second choice is Manjaro, and my third choice is Adelie. I may try those on the lesser C610 just to see how they run, provided I don't give up on Gentoo.

Since I had Debian running, I did some performance testing on the new thermals. CPU temperature stress-test peaks at 66.5C about seven minutes in, and drops to 61C when the CPU fan kicks into low. I was unable to get the CPU fan to kick into high, and I was not willing to suffocate the fan to find out. For reference, fan goes into high speed at 70C. GPU temperatures are quite low, about 40C during the CPU stress test. New thermal pads helped tremendously here.

I added an extra pad over the graphics SGRAM to simulate the C640's additional thermal pad on the heat spreader (which in Dell's manuals is called an EMI shield bracket) to cool this area. Two pads, a 2mm and a 1mm, were cut to 0.7 x 0.5 inches with the 2mm pad on the heat spreader plate and the 1mm directly to it to give an air-free contact.

I added some photos I took along the way to this Imgur album, and will link individual images throughout this post. Should Imgur die, I can post them to the forums directly now that we have disk space.