Alienware 17 R1 Restoration and Upgrade

[AtomSnakes343]

I am currently doing a fairly extensive restoration and modernization of an Alienware 17 R1 built around the Compal LA-9331P motherboard. The machine is now almost completely disassembled. So far I have acquired a replacement keyboard, trackpad, speakers, and battery, along with a five-heatpipe GPU heatsink intended for GTX 800/900/10-series MXM cards and a higher-CFM GPU fan. I still intend to replace several worn chassis parts, including the palmrest, front display bezel, and trackpad buttons, and I am also acquiring an upgraded CPU heatsink and possibly a better CPU fan.

 

The planned electronic upgrades are a faster i7-4940MX CPU, 32 GB of DDR3L, several multi-terrabyte SATA SSDs and an mSATA, a modern display, Wi-Fi 7 via a BE200 and adapter (already done), new antennae , and eventually a substantially newer MXM GPU. The LA-9331P has four DDR3L SO-DIMM slots but is officially a DDR3L-1600 platform. Faster DDR3L-1866/2133 modules should normally fall back to a supported JEDEC speed, although running them above 1600 would depend on BIOS and memory-controller support. For storage, the machine provides two normal SATA drive connections, an ODD/HDD bay and mSATA. There is unfortunately no native NVMe interface, so my current plan is to put Windows on a high-quality 2.5-inch SATA SSD and use the mSATA slot and additional SATA bay for secondary storage.

 

For the GPU, the most interesting card I have actually found for sale is an industrial Turing Quadro RTX 5000 MXM module with 16 GB of GDDR6. The seller has confirmed that it is an 82×105 mm MXM 3.0/3.1 Type-B card, and the listing photographs show an EGX-MXM-RTX5000-style PCB. It appears physically suitable for the system, but the remaining questions are the vBIOS, UEFI GOP support, power limit, subsystem/device IDs and compatibility with the Alienware BIOS. The heatsink would also need to be checked and modified for the card’s exact VRAM and VRM layout. Ampere RTX 3060/3070 MXM-B modules apparently exist through embedded-GPU suppliers, but most are inquiry-only industrial products and are often custom OEM boards with uncertain compatibility. I have also not yet found a practical retail source. At this point, the RTX 5000 looks like the best balance of performance, VRAM, physical compatibility and likelihood of actually working. A GTX 1070 or 1080 MXM remains the cheaper and more established fallback in the worst-case scenario. 
 

I am also not sure if Eurocom sells MXM GPUs, as I have no experience dealing with them. I am, however, aware people here have purchased from them before, so I am thinking of sending them a message and exploring that possibility. As for my exact choice, Turing is my go-to since it is a modern RTX standard that can bring the laptop at least partially into the 2020s but also not so far away from the Ranger to make me worry about further firmware issues or breaking the bank. I could theoretically go for Ampere or even Ada if I can find them in 82x105mm MXM form, as I have the cooling for them — but my main issue is firmware. The LA-9331P can electrically make any MXM card work — even the Ada monstrosities — but, as always, it is the BIOS, vBIOS, and EC behavior that gives me anxiety. 

 

The display side is especially interesting. My original 60Hz panel is damaged, so the machine currently has no usable internal screen. I would like to install a 17.3-inch 2560 × 1440 165 Hz eDP panel rather than another original 1080p unit. The LA-9331P schematic shows a four-lane internal eDP path and switching circuitry capable of accepting display signals from either the Haswell CPU or the MXM GPU. It includes signals such as DGPU_SELECT#, DP_MXM_CARD_SEL and PANEL_SW, so the hardware is more sophisticated than a simple fixed dGPU-only connection. However, the known factory-style 120 Hz eDP conversion normally disables the Intel graphics/Optimus configuration and uses direct MXM output. My current understanding is therefore that a modern QHD high-refresh panel should be technically plausible over the eDP path, but direct dGPU operation is the most likely out-of-the-box result.

 

The unanswered question is whether the existing display mux can be made user-selectable through BIOS/EC modification. In principle, an expert might be able to create an iGPU/Optimus mode and a direct-dGPU performance mode, but I have not found proof that this has actually been implemented on the Alienware 17 R1. I also would not assume that Intel HD 4600 can reliably drive 1440p at the full 165 Hz, even if the CPU eDP path is selected. A possible compromise would be 1440p at 60–120 Hz in an experimental Intel-driven mode and the full available refresh rate in direct-MXM mode. Before purchasing a panel, I still need to verify the exact 40-pin pinout, cable lane population, panel voltage and backlight requirements, connector position, and mechanical mounting. Modern slim panels will probably require custom brackets or spacers.

 

Losing Optimus sucks, as this means I lose flexibility with the entire computer as a whole. If the MXM starts acting up or I do not have one ready yet, there goes my display. However, I have found a temporary solution: for initial testing, I intend to use the laptop through HDMI with the internal display cable disconnected. The physical HDMI connector is dual-purpose and supports both HDMI input and HDMI output through separate switching circuitry. The schematic also shows that the HDMI-output path can be sourced from either the CPU graphics or MXM graphics, although actual POST behavior will still depend on the BIOS and installed GPU. The board has not been powered for roughly a year, the RTC/CMOS battery is currently absent, and it has been cleaned repeatedly with 99.5% isopropyl alcohol. Before testing, I will install a fresh correctly wired RTC battery, known-good RAM, the CPU, and complete cooling system, then attempt a minimal HDMI boot.

 

Cooling will be addressed before the final CPU and GPU upgrades. I have purchased a much larger custom copper multi-heatpipe CPU assembly from Cicichen, who has previously produced upgraded parts specifically for the Ranger. His CPU heatsink is made for the Alienware 17 R1/P18E mounting geometry, but adapts the later Alienware vapor-chamber design philosophy (from the R5, I believe) to the older R1 chassis. The listing photos show a large copper vapor chamber over the CPU, four tightly nested heatpipes, and a substantial all-copper fin stack. It is essentially the kind of cooler Dell might have fitted if the Ranger had been designed around a hotter CPU from the beginning. Unfortunately, he ran out of upgraded CPU fans and got rid of the tooling, which he said he deeply regrets, so I’ll just use the stock one. My final intention is to use Conductonaut, but I will probably first validate the heatsink contact and temperatures with PTM7950 paste before taking the additional electrical and mechanical risk.

 

Cicichen has also offered to manufacture a completely custom GPU heatsink once I select and receive the graphics card. His current five-pipe Ranger GPU heatsink I acquired is designed for older GTX layouts and will not directly fit an RTX board, but he can build me a new heatsink around photographs and dimensions of the exact PCB. That removes one of the biggest mechanical problems with fitting an RTX 5000, 30XX, or 40XX (provided they are 82x105mm).

 

My overall aim is not to min-max the cheapest performance per dollar/euro. It is to preserve and modernize a machine very dear to me with a socketed CPU, replaceable MXM graphics, four RAM slots, several storage interfaces, separate I/O daughterboards, and unusually flexible display routing, courtesy of the overbuilt and, IMHO, legendary Compal LA-9331P. A machine that is not like the soldered, planned obsolescence-obsessed, enshittified stuff we have today

 

I am completing the work in stages and will leave the expensive GPU purchase until the rest of the platform is assembled, stable, and tested. I would particularly appreciate any first-hand information about RTX MXM vendors and vBIOS compatibility, the LA-9331P eDP mux controls, retaining Intel graphics with an eDP panel, and modern 1440p panel installations in the Alienware 17 R1. Any info on potential RAM overclocking and CPU/GPU underclocking would also be appreciated, as I heard somewhere this is not possible on this machine for some reason.