Unsolved

1 Rookie

 • 

7 Posts

 • 

7 Points

11

June 26th, 2026 17:11

Alienware 17 R1 (P18E) Restomod

Major update on this project, since quite a lot has changed since my original post.

When I first posted, the tentative plan was an i7-4940MX, RTX 2070 MXM, upgraded cooling, and either the factory 1080p 120Hz display or some kind of experimental higher-resolution panel conversion. At the time, I was not sure whether the motherboard really had the display bandwidth, whether a modern RTX card could work properly, or whether the required eDP cable and backlight controls could realistically be adapted.

After finding the Compal LA-9331P motherboard schematic, speaking with several highly experienced Alienware/MXM modders, and finding active development work around modern RTX MXM cards, the situation now looks much more promising.

MOTHERBOARD FINDINGS

The Alienware 17 R1 uses the Compal LA-9331P platform. The schematic confirms that this is an unusually capable and modular motherboard:

  • Socketed Haswell rPGA947 CPU
  • Intel HM87 chipset
  • Four DDR3L SO-DIMM slots
  • MXM 3.x Type-B graphics slot connected through PEG
  • Multiple SATA devices, including mSATA and the two 2.5-inch bays
  • Separate LVDS and eDP internal-display arrangements
  • A four-lane internal eDP path through display mux/switching hardware
  • Display routing from either the Intel graphics or the MXM GPU, depending on configuration
  • Replaceable daughterboards and a remarkably serviceable chassis

Most importantly for the display conversion, the internal eDP route appears to be a full four-lane HBR2-capable path. Four-lane HBR2 provides enough usable bandwidth for a modern 2560×1440 165Hz panel using reduced-blanking timings.

This does not guarantee that every panel and GPU combination will work automatically, but the motherboard itself does not appear to be fundamentally bandwidth-limited to 1080p 120Hz. Compal seems to have provided far more capability than Dell ever used in the factory configurations.

LVDS, EDP, SG AND PEG

The factory arrangements are effectively:

  • 60Hz LVDS panel: SG/Optimus mode, with the Intel GPU driving the internal display and the NVIDIA GPU rendering through Optimus
  • 120Hz eDP panel: PEG/dedicated-GPU mode, with the MXM card driving the panel directly

An unlocked BIOS reportedly allows these graphics modes to be selected manually, but the normal factory behavior is determined largely by the display configuration.

This distinction is important because a modern RTX card may work under Optimus with the original LVDS panel even if its direct internal eDP output or backlight controls are incompatible.

THE MODERN RTX BREAKTHROUGH

A NotebookTalk member named SuperMG has developed a solderless “Backlight Mod V2” for modern MXM RTX cards.

Many newer industrial MXM cards produce the eDP video signal but do not provide the older laptop’s expected backlight-enable/PWM behavior on the appropriate MXM contacts. The mod consists of a flexible cable fitted directly over selected MXM edge contacts, plus a small external PWM controller.

The flex obtains 3.3V from MXM pins 278/280 and supplies the missing backlight-related contacts around pins 23–27. The PWM board allows brightness to be adjusted manually.

The instructions explicitly list the following as supported:

  • Alienware M17x R4 with 120Hz eDP
  • Alienware 17 R1 / M17x R5, 2013, with 120Hz eDP
  • Multiple Dell Precision, Clevo and HP ZBook models
  • GPUs from Turing onward, including ADLINK and X-VISION cards

SuperMG has personally tested RTX 3080 and RTX 4070 cards in the M17x R4 120Hz eDP configuration, as well as several other MXM systems. He states that the relevant M17x R4 and Alienware 17 R1 mux arrangements and MXM pinouts are the same.

The mod does not alter or carry the actual eDP video lanes. It only restores the missing backlight control. Video still travels normally from the GPU, through the MXM slot and motherboard eDP route, and then through the LCD harness.

RTX 4080 AND 4090 EDID ISSUE

X-VISION RTX 4050/4060/4070 cards reportedly work directly once the physical requirements are met.

The X-VISION/ZRT RTX 4080 and 4090 are more complicated. They can apparently show an image before the NVIDIA driver is installed, but the screen goes black after driver installation because of an eDP/EDID compatibility problem.

The documented workaround uses CRU to preserve and reinstall the panel’s correct PnP EDID information. The override is prepared using the old working GPU, exported, and then automatically reapplied using startup and shutdown scripts.

That means the 4080/4090 problem appears to be a known driver/EDID issue, rather than proof that the hardware cannot send an eDP signal.

GPU OPTIONS

I am now deciding between two very different routes.

Option 1: Quadro RTX 5000 16GB (couldn’t find an 82x105mm 20XX)

  • Around €489 for the industrial card I found
  • 16GB GDDR6
  • Considerably cheaper
  • Existing evidence of RTX 5000-class cards working in old Alienwares
  • More manageable cooling and power requirements
  • Supported by the Backlight Mod V2
  • Potentially the safer route

Option 2: X-VISION RTX 4080 or RTX 4090

  • Standard MXM-format modern Ada cards are being manufactured
  • RTX 4080 has 12GB
  • RTX 4090 has 16GB
  • X-VISION’s vBIOS reportedly works in laptops
  • The 4090 is substantially more expensive and can reach 150W
  • EDID override is required for direct internal eDP
  • Cooling, power throttling, and Haswell compatibility still need real testing

NotebookTalk member ssj92 owns an RTX 4090 for an Alienware 17 and is currently commissioning custom Alienware 17 CPU and RTX 4090 GPU heatsinks from Cicichen. He has not yet fully validated the complete 4090/AW17 configuration, so it should not yet be considered a proven drop-in upgrade.

He expects SG/Optimus mode to work and is actively investigating the Haswell power-throttling issue, including a reversible method involving one MXM contact and a more invasive resistor method.

COOLING

I already have:

  • A five-pipe GPU heatsink for the 9XX and 10XX-series MXMs
  • A four-pipe CPU heatsink with a vapor chamber
  • A higher-output GPU fan (~11 CFM)
  • New keyboard
  • New touchpad
  • New battery
  • Replacement Ethernet/USB daughterboard (LS-9339P)
  • BE200 Wi-Fi 7 card plus an M.2 to mini-PCIe adapter

Cicichen is also currently making Alienware 17 heatsinks specifically for modern RTX 4090 PCBs for ssj92, which is very encouraging for the mechanical side of the project.

The final GPU heatsink must match the exact PCB layout, including the GPU die, VRAM, MOSFETs and inductors. The Backlight Mod V2 flex must also have sufficient clearance so that the heatsink does not crush or cut it.

DISPLAY PLAN

The target is now a modern 17.3-inch 2560×1440 165Hz BOE panel, probably from the NE173QHM-NY family.

These panels generally use:

  • Four-lane eDP
  • 40-pin connector
  • 3.3V panel logic
  • 2560×1440 at 165Hz
  • Modern slim construction
  • No factory side brackets

The original Alienware LVDS panel uses removable side mounting rails. My damaged original panel can therefore serve as the dimensional template. The likely mechanical solution is to reuse the original Alienware side rails with thin adapter strips/spacers or a custom cradle for the modern slim panel.

I do not intend to drill into the panel. The adapter would attach only to the metal perimeter/backplate.

EDP CABLE


The factory Alienware 17 R1 eDP harness is the Dell N392W / Compal DC02C004000.

I would prefer to modify this OEM harness rather than manufacture the entire assembly from scratch because it also carries the webcam, microphones and other lid wiring.

The likely approach is:

  • Preserve the motherboard-side connector
  • Preserve the webcam and microphone branches
  • Preserve the original shielded four-lane eDP pairs
  • Modify or replace only the panel-end branch
  • Route it to the modern panel’s connector position
  • Match the exact 40-pin panel pinout
  • Provide proper 3.3V panel logic power if the OEM harness outputs 5V
  • Preserve backlight power, enable, PWM, AUX, HPD and grounding correctly

The voltage at the panel end of a genuine N392W cable still needs to be measured. I do not intend to connect the BOE panel until the cable has been continuity-mapped and all voltages have been verified with the panel disconnected.

TESTING SEQUENCE

I am deliberately trying not to change everything at once.

The current plan is:

1. Reassemble the machine using the original GTX 880M, original LVDS cable and damaged but potentially usable 60Hz panel.
2. Confirm motherboard POST, BIOS access, HDMI output, CPU, RAM, storage and cooling.
3. Test the modern RTX card first in SG/Optimus mode using the original LVDS setup.
4. Confirm that the RTX card POSTs, appears in Windows, accepts a driver and runs under load.
5. Install the factory eDP harness and Backlight Mod V2.
6. Test direct PEG/eDP operation separately.
7. Build and test the modified N392W panel branch.
8. Start the new panel at 2560×1440 60Hz, then test 120Hz, 144Hz and finally 165Hz.
9. Apply the correct EDID override only if the chosen GPU requires it.

No system BIOS, EC or GPU vBIOS will be flashed blindly. Firmware modification will only be considered if a specific, reproducible failure proves it is necessary.

1 Rookie

 • 

7 Posts

 • 

7 Points

June 26th, 2026 17:11

The complete end goal is increasingly looking technically possible:

  • i7-4940MX
  • 32GB DDR3L-2133MHz (worst-case scenario it downclocks, though I think there are some ways to overclock it)
  • RTX 5000 16GB or RTX 4080/4090
  • 2560×1440 165Hz internal eDP display
  • Wi-Fi 7
  • SATA/mSATA SSDs
  • Upgraded custom cooling
  • Working webcam and microphones
  • New front bezel cover, palm rest, keyboard, trackpad+buttons
  • Windows 11

The CPU, DDR3 memory, and SATA storage will obviously remain the platform’s main historical bottlenecks. However, the motherboard, MXM interface and internal display path appear far more capable than I originally realized.

The project is still experimental, and a complete modern-RTX-plus-QHD165 configuration has not yet been demonstrated in my particular machine. Nevertheless, almost every obstacle now has either a tested solution on a closely related Alienware, an active developer working on it, or a reasonably straightforward engineering route.

I will continue updating this thread as the machine is reassembled and each stage is tested. This motherboard is overbuilt to the moon and back.

No Events found!

Top