Hubbry Logo
Mobile PCI Express ModuleMobile PCI Express ModuleMain
Open search
Mobile PCI Express Module
Community hub
Mobile PCI Express Module
logo
8 pages, 0 posts
0 subscribers
Be the first to start a discussion here.
Be the first to start a discussion here.
Mobile PCI Express Module
Mobile PCI Express Module
Mobile PCI Express Module
View on Wikipedia
from Wikipedia
MXM slot
Mobile PCI Express Module
Reference scheme of placing a basic elements (GPU and video RAM) on a first generation MXM-II cards for 35 mm GPU
Reference design of a first generation MXM-II card for 35 mm GPU
No. of devices1
StyleSerial
External interfaceno

Mobile PCI Express Module (MXM) is an interconnect standard for GPUs (MXM Graphics Modules) in laptops using PCI Express created by MXM-SIG. The goal was to create a non-proprietary, industry standard socket, so one could easily upgrade the graphics processor in a laptop, without having to buy a whole new system or relying on proprietary vendor upgrades.

A rare photo of a MXM-IV graphics card (GeForce Go 7950 GTX)
A GTX 780M GPU with MXM3 socket

Generations

[edit]

Smaller graphics modules can be inserted into larger slots, but type I and II heatsinks will not fit type III and above or vice versa.

Dell's Alienware m5700 platform uses a heatsink that will fit Type I, II, & III cards without modification.

MXM 3.1 was released in March 2012 and added PCIe 3.0 support.[1][2]

First generation modules are not compatible with second generation (MXM 3) modules and vice versa. First generation modules I to IV are fully backwards compatible.

Some MXM cards have different mounting screw hole configurations, always check the mounting holes of the MXM card and verify that they match those of the card you plan to upgrade to.

1st generation
MXM
Type		 Width Length Connector Module
Compatibility		 Thermal
Compatibility		 Max.
Power		 Max.
GPU size [3]
MXM-I 70 mm 68 mm Standard 230 I, II I 18 W 35 × 35 mm
MXM-II 73 mm 78 mm I, II 35 W
MXM-III 82 mm 100 mm I, II, III III 45 W 40 x 40 mm
HE

232

I, II, III, III (HE) III, III (HE) 75 W
MXM-IV* 117 mm I, II, III, III (HE), IV III, III (HE), IV

*Deprecated/abandoned, became/replaced by MXM-III (HE)

2nd generation (MXM 3)
MXM
Type		 Width Length Module
Compatibility		 Thermal
Compatibility		 Max. Slot
Power		 GPU
memory bus
MXM-A 82 mm 70 mm A A 55 W 64-bit or 128-bit
MXM-B 105 mm A, B A, B 200 W* 256-bit
MXM-B+ Not Public Not Public Not Public Not Public Not Public

*Although the slot can deliver 200 watts, it will run dangerously hot, separate power cable always used at high wattages.

Specification

[edit]

MXM is no longer freely supplied by Nvidia but it is controlled by the MXM-SIG controlled by Nvidia. Only corporate clients are granted access to the standard. The MXM 2.1 specification is widely available.[citation needed]

List of MXM cards

[edit]

First generation MXM cards

[edit]
Vendor Name Released MXM

Type

GPU Architecture Core

config

TFLOPS

(FP32)

TDP Dimensions
AMD Radeon HD 6530M 2010-11 MXM-II Capilano TeraScale 2 400:20:8 0.36 26 W 73 x 78 mm
Radeon HD 6550M 0.48
Lexington 0.52
Radeon HD 6570M Capilano 30 W
ATI Mobility Radeon X1800 2006-03 MXM-III M58 Ultra-Threaded SE 12:8:12:12 82 x 100 mm
Mobility Radeon X1900 2007-01 M68 36:8:12:12
Mobility Radeon HD 2600 2007-05 MXM-II M76 TeraScale 120:8:4 0.12 73 x 78 mm
Mobility Radeon HD 2600 XT 0.16
Radeon E2400 2007-06 RV610 40:4:4 0.05 25 W
Mobility Radeon HD 2700 2007-07 M76 120:8:4 0.16 35 W
Mobility Radeon HD 3670 2008-01 M86 30 W
Mobility Radeon HD 3650 0.12
Mobility Radeon HD 3470 M82 40:4:4 0.05
Mobility Radeon HD 3450 0.04
Mobility Radeon HD 3430 2008-07 12 W
Mobility Radeon HD 3410 0.03 7 W
Mobility Radeon HD 4570 2009-01 M92 80:8:4 0.11
Mobility Radeon HD 4530 0.08
Radeon E4690 MXM 2009-06 M96 320:32:8 0.38 30 W
Mobility Radeon HD 4550 2010 M93 80:8:4 0.09
Nvidia GeForce Go 6600 NPB 128M 2005-09 MXM-I NV43 Curie 8:3:8:4 70 x 68 mm
GeForce 8400M GS 2007-05 G86S Tesla 16:8:4 0.03 11 W
GeForce 9300M GS 2008-06 G98S 8:4:4 0.02 13 W
Quadro NVS 150M 2008-08 10 W
Quadro NVS 160M 12 W
Quadro FX Go 540 2004-08 MXM-II NV43 Curie 4:3:8:4 25 W 73 x 78 mm
GeForce Go 6600 2005-09 8:3:8:4
Quadro FX 1500M 2006-04 MXM-III G71 20:7:20:16 45 W 82 x 100 mm
GeForce Go 7900 GS MXM-II 20 W 73 x 78 mm
GeForce 8600M GS 2006-05 G86 Tesla 16:8:4 0.03 20 W
GeForce 8600M GT G84 32:16:8 0.06
GeForce 9600M GS 2008-06 G96C 0.07
GeForce 9600M GT 0.08 23 W
GeForce 9700M GT 2008-07 G96 0.1 45 W
GeForce 9650M GT 2008-08 G96C 0.09 23 W OEM Custom
Quadro FX 370M G98S 8:4:4 0.02 20 W 73 x 78 mm
Quadro FX 770M G96 32:16:8 0.08 35 W
Quadro FX 1700M 2008-10 0.1 45 W
GeForce GT 220M 2009-06 G96C 0.09 14 W
GeForce GT 320M GT216 Tesla 2.0 0.05 23 W
GeForce GTS 350M 2010-01 GT215 96:32:8 0.21 28 W OEM Custom
GeForce GTS 360M 0.25 38 W
GeForce Go 6800 2004-11 MXM-III NV41 Curie 12:5:12:8 45 W 82 x 100 mm
Quadro FX Go 1400 2005-02 8:5:8:8
Quadro FX 2500M 2005-09 G71 24:8:24:16 45 W
Quadro NVS 120M 2006-06 MXM-II G72B 4:3:4:2 10 W
GeForce Go 7950 GTX 2006-10 MXM-III G71 24:8:24:16 0.14 45 W
Quadro FX 3500M 2007-03
Quadro FX 1600M 2007-06 MXM-III (HE) G84 Tesla 32:16:8 0.08 50 W
GeForce 8700M GT 29 W
Quadro NVS 320M 0.07 20 W
GeForce 8800M GTS 2007-11 G92 64:32:16 0.16 50 W
GeForce 8800M GTX 96:48:16 0.24 65 W
Quadro FX 3600M 2008-02 64:32:16 0.16 70 W
96:48:16
Quadro FX 2700M 2008-08 G94 48:24:16 0.13 65 W
GeForce 9800M GTS 64:32:16 0.19 75 W
Quadro FX 3700M G92 128:64:16 0.35
GeForce 9800M GT 2008-07 96:48:16 0.24 65 W
GeForce 9800M GTX 112:56:16 0.28 75 W
GeForce GTX 280M 2009-03 128:64:16 0.38
GeForce Go 7950 GTX 2006-10 MXM-IV G71 Curie 24:8:24:16 0.14 45 W 82 x 117 mm

Second generation MXM cards

[edit]
Vendor Name Released MXM
Type 		GPU Architecture Core config[a] TFLOPS
(FP32) 		TDP Dimensions
AMD FirePro M5800 2010-03 Type-A Madison XT Terascale 2 400:20:8 0.52 26 W 80 × 70 mm
FirePro M5950 2011-01 Whistler XT 480:24:8 0.69 35 W
FirePro M4000[4] 2012-06 Chelsea XT GL GCN 1 512:32:16 0.69 33 W
FirePro M5100 2013-10 Venus XT 640:40:16 1.0 45 W
FirePro M6100 Type-B Saturn XT GL GCN 2 768:48:16 1.7 100 W 82 × 105 mm
FirePro W5130M 2014-08 Type-A Tropo LE GCN 1 512:32:16 1.0 35 W 80 × 70 mm
FirePro W5170M Tropo XT 640:40:16 1.1 45 W
FirePro W6150M 2015-11 Type-B Saturn XT GL GCN 2 768:48:16 2.0 100 W 82 × 105 mm
FirePro S4000X 2014-08 Type-A Venus XT GCN 1 640:40:16 1.0 45 W 80 × 70 mm
FirePro S7100X 2016-05 Type-B Amethyst XT GCN 3 2048:128:32 3.0 100 W 82 × 105 mm
Radeon Pro WX 4130 Mobile 2017-03 Type-A Baffin LE GCN 4 640:40:16 1.4 45 W 82 × 70 mm
Radeon Pro WX 4150 Mobile[5] Baffin PRO 896:56:16 1.9
Radeon Pro WX 4170 Mobile Type-B Baffin XT 1024:64:16 2.5 60 W 82 × 105 mm
Radeon Embedded E6465 2015-10 Type-A Caicos TeraScale 2 60:8:4 0.2 28 W 80 × 70 mm
Radeon Embedded E8860 Cape Verde XT GCN 1 640:40:16 0.8 37 W
Radeon Embedded E8870[6] Type-B Bonaire Pro GCN 2 768:48:16 1.4 75 W 82 × 105 mm
Radeon Embedded E8950 Tonga XT GCN 3 2048:128:32 3.0 95 W
Radeon Embedded E9172[7] 2017-10 Type-A Polaris 12 (Lexa) GCN 4 512:32:16 1.2 35 W 82 × 70 mm
Radeon Embedded E9174 1.2 50 W
Radeon Embedded E9260[8] 2016-09 Baffin PRO 896:56:16 2.2 50 W
Radeon Embedded E9550[9] Type-B Ellesmere XT 2304:144:32 5.8 95 W 82 × 105 mm
Radeon RX 550 512SP 2017-10 Type-A Baffin 512:32:16 1.1 50 W 82 x 70 mm
Radeon RX 580 2048SP 2018-10 Type-B Polaris 20 2048:144:32 5.3 120 W 82 x 105 mm
Radeon RX 580 2017-04 2304:144:32 6.2
Radeon RX 5500 XT 2019-12 Navi 14 RDNA 1 1408:88:32 5.2 105 W
Radeon RX 6600 2021-10 Navi 23 RDNA 2 1792:112:64 8.9 132 W
Radeon RX 6600 XT 2021-07 2048:128:64 10.6 160 W
Intel MXM-AXe (Arc A350E) 2022-03 Type-A DG2-128 Generation 12.7 768:48:24 3.4 25 W 82 x 70 mm
MXM-AXe (Arc A370E ) 1024:64:32 4.2 35 W
Arc A350M 768:48:24 3.4 25 W
Arc A370M 1024:64:32 4.2 35 W
Arc A380M 2023-01 4.1
Arc A530M 2023-08 Type-B DG2-256 1536:96:48 4.0 95 W 82 × 105 mm
Arc A570M 2048:128:64 5.3
Arc A730M 2022 DG2-512 3072:192:96 12.6 120 W
Arc A770M 4096:256:128 13.5 150 W
Nvidia Tesla M6 2015-11 Type-B GM204 Maxwell 2.0 1536:96:48 3.0 100 W
Quadro K610 Mobile 2013-07 Type-A GK208 Kepler 192:16:8 0.4 30 W 80 × 70 mm
Quadro K1100 Mobile GK107 384:32:16 0.5 45 W
Quadro K2100 Mobile GK106 576:48:16 0.7 55 W
Quadro K3100 Mobile Type-B GK104 768:64:32 1.1 75 W 82 × 105 mm
Quadro K5100 Mobile 1536:128:32 2.4 100 W
Quadro M520 Mobile[10] 2017-01 Type-A GM108 Maxwell 384:16:8 0.8 25 W 80 × 70 mm
Quadro M620 Mobile[11] GM107 512:32:16 1.0 30 W
Quadro M1000M Mobile 2015-08 1.0 40 W
Quadro M1200M Mobile[12] 2017-01 640:40:16 1.4 45 W
Quadro M2000M Mobile 2015-12 1.4 55 W
Quadro M2200M Mobile[13] 2017-01 GM206 Maxwell 2.0 1024:64:32 2.1 55 W
Quadro M3000M Mobile 2015-08 Type-B GM204 2.1 75 W 82 × 105 mm
Quadro M4000M Mobile 1280:80:64 2.5 100 W
Quadro M5000M Mobile 1536:96:64 3.0 100 W
Quadro P1000 Mobile 2017-02 Type-A GP107 Pascal 512:32:16 1.6 40 W 80 × 70 mm
Quadro P2000 Mobile 2019-02 GP106 1152:72:32 3.0 75 W
Quadro P3000 Mobile[14] 2017-01 Type-B GP104 1280:80:32 3.1 75 W 82 × 105 mm
Quadro P3200 Mobile 2018-02 1792:112:64 5.5 75 W
Quadro P4000 Mobile[15] 2017-01 4.4 100 W
Quadro P4200 Mobile 2018-02 2304:144:64 7.6 100 W
Quadro P5000 Mobile[16] 2017-01 2048:128:64 6.2 100 W
Quadro P5200 Mobile 2018-02 2560:160:64 8.9 100 W
Quadro T1000 Mobile 2019-05 Type-A TU117 Turing 896:56:32 2.6 50 W 82 × 70 mm
GeForce GTX 965M[17] 2016-01 Type-A GM204 Maxwell 2.0 1024:64:32 2.4 75 W 80 × 70 mm
GeForce GTX 965M 2014-10 Type-B 2.4 75 W 82 × 105 mm
GeForce GTX 970M 1280:80:48 2.7 75 W
GeForce GTX 980M 1536:96:64 3.2 100 W
GeForce GTX 980MX 2016-06 1664:104:64 3.9 148 W 83 × 115 mm
GeForce GTX 980 2015-09 2048:128:64 4.7 200 W 102.6 × 115 mm
GeForce GTX 1050 Mobile 2017-01 Type-A GP107 Pascal 640:60:16 1.9 75 W 82 × 70 mm
GeForce GTX 1050 Ti Mobile 768:48:2 2.4 75 W 82 × 70 mm
GeForce GTX 1060 Mobile "P872L" 2017-05 Type-B GP106 1280:80:48 4.4 100 W 100 × 124 mm
GeForce GTX 1060 Mobile 4.3 78 W 82 × 105 mm
GeForce GTX 1070 Mobile "P872L" GP104 2048:128:64 6.7 150 W 100 × 124 mm
GeForce GTX 1070 Mobile 120 W 82 × 105 mm
GeForce GTX 1080 Mobile "P872L" 2560:160:64 9.1 190 W 100 × 124 mm
GeForce GTX 1080 Mobile "P870L" (MSI GT73VR) 200 W
GeForce GTX 1080 Mobile "V1.0" (MSI GT83) 150 W 94 × 105 mm
GeForce RTX 2060 Mobile (Clevo) 2019-01 Type-B TU106 Turing 1920:120:48 4.6 80 W 100 × 124 mm
GeForce RTX 2070 Mobile (Clevo) 2304:144:64 6.6 115 W
GeForce RTX 2080 Mobile (Clevo) TU104 2944:184:64 9.4 150 W
Quadro RTX 3000 Mobile 2019-05 TU106 1920:120:64 5.3 80 W 82 × 105 mm
Quadro RTX 4000 Mobile TU104 2560:160:64 8.0 110 W
Quadro RTX 5000 Mobile 3072:192:64 10.9 110 W
GeForce RTX 3050 Mobile 2021-05 Type-B GA107 Ampere 2048:64:32 5.5 85 W 82 x 105 mm
GeForce RTX 3050 Ti Mobile 2560:80:32 5.3
GeForce RTX 3050 8GB 2022-01 GA106 9.1 90 W
GeForce RTX 3060 Mobile 2021-01 3840:120:48 10.9 135 W
GeForce RTX 3060 12GB 3584:112:48 12.7 90 W
GeForce RTX 3070 Mobile GA104 5120:160:80 16.0 150 W
GeForce RTX 3070 Ti Mobile 2022-01 5888:184:96 16.6
GeForce RTX 3080 Mobile 2021-01 6144:192:96 19.0
Quadro RTX A3000 2021-04 4096:128:64 11.8 95 W
GeForce RTX 4050 Mobile 2023-01 Type-A AD107 Ada Lovelace 2560:80:48 9.0 100 W 82 x 70 mm
GeForce RTX 4060 Mobile 3072:96:48 11.6 110 W
GeFroce RTX 4070 Mobile AD106 4608:144:48 15.6
GeForce RTX 4080 Mobile Type-B AD104 7424:232:80 24.7 130 W 82 x 105 mm
GeForce RTX 4090 Mobile AD103 9728:304:112 33 135 W

Other uses

[edit]
VIA QSM-8Q90 Qseven computer-on-module using a MXM-2 connector
Congatec SMARC computer-on-module using MXM-3 connector

The Qseven computer-on-module form factor uses a MXM-II connector, while the SMARC computer-on-module form factor uses a MXM 3 connector. Both implementations are not in any way compatible with the MXM standard.[citation needed]

Notes

[edit]

References

[edit]
[edit]
Revisions and contributorsEdit on WikipediaRead on Wikipedia

Mobile PCI Express Module

View on Grokipedia
from Grokipedia
The Mobile PCI Express Module (MXM) is a standardized electromechanical interface and form factor for graphics processing units (GPUs) designed for integration into mobile computing devices, such as laptops, workstations, and embedded systems, utilizing the (PCIe) protocol for high-speed data transfer, developed by the MXM (MXM-SIG). It specifies electrical, mechanical, and requirements to support discrete graphics adapters in environments with constrained size, power, and heat dissipation, typically enabling up to 16 PCIe lanes, multiple display outputs, and power delivery from 7-20V at up to 10A. Developed as a non-proprietary standard to facilitate GPU upgrades and repairs without replacing the entire system, MXM was introduced in 2004 to address in mobile graphics solutions. The specification has evolved through multiple generations, with early versions (1.x) focusing on lower-power modules around 18-75W and dimensions up to 82x117mm, while second-generation (2.x) designs like MXM-A (55W, 82x70mm) and MXM-B (up to 200W, 82x105mm) introduced broader compatibility within their series but not across generations. Later iterations, such as version 3.1 released in , support PCIe generations up to Gen3 (8 GT/s), DDR3/GDDR5 memory configurations, and up to six outputs compliant with VESA standards, alongside form factors Type A and Type B for varying sizes and thermal profiles. MXM modules find applications in high-performance scenarios including gaming laptops, mobile workstations for and , and industrial embedded systems for AI acceleration, , and autonomous , where they provide desktop-like capabilities with features like up to 40 TFLOPS FP32 performance and 24GB GDDR memory in modern implementations. Despite initial promise for upgradability, adoption has shifted toward industrial and specialized markets due to evolving designs favoring soldered components, though the standard persists in ruggedized and environments with support for recent GPU architectures like NVIDIA's , , and Blackwell, with recent initiatives like Framework's modular s aiming to restore upgradability.

Overview

Definition and Purpose

The Mobile PCI Express Module (MXM) is an interconnect standard for graphics processing units (GPUs) in laptops and other systems, specifying the electrical, mechanical, and thermal interfaces to enable modular discrete graphics solutions based on technology. The primary purpose of MXM is to facilitate the independent upgrade or replacement of GPUs without necessitating changes to the motherboard or other core system components, thereby enhancing repairability and extending the usability of mobile devices in performance-demanding environments. This standard addresses the constraints of space, power, and thermal management in compact systems like notebooks and mobile workstations, where soldered GPUs traditionally limit flexibility. By promoting a standardized socket and interface, MXM supports among GPUs from different vendors, allowing for easier integration and maintenance in high-performance applications such as gaming and workloads. Its initial scope focused on discrete GPUs optimized for graphics-intensive tasks, including and , while ensuring thermomechanical compatibility to simplify system design.

History and Development

The Mobile PCI Express Module (MXM) was introduced by in May 2004 as an open interconnect standard aimed at enabling interchangeable graphics solutions in laptops, addressing the limitations of proprietary and soldered GPU integrations that hindered upgrades and customization in . To promote adoption, formed the MXM Special Interest Group (MXM-SIG), initially comprising GPU vendors like ATI (now part of ) and original design manufacturers (ODMs) such as , , , and Inventec, with later participation from system integrators including , HP, and . This collaborative effort sought to standardize the physical, electrical, and software interfaces for mobile GPUs based on , fostering broader industry compatibility. The MXM 1.0 specification was released in 2005, marking the first formal milestone and directly targeting the challenges of non-upgradable soldered GPUs prevalent in early 2000s laptops by defining a modular slot for easier replacement and future-proofing. Subsequent developments built on this foundation: the MXM 2.0 specification arrived in 2007, expanding support for additional module types to accommodate varying performance needs and thermal envelopes in diverse laptop designs. In 2009, MXM 3.0 introduced higher power delivery capabilities, enabling more demanding graphics workloads while maintaining where feasible. The MXM 3.1 update, released in March 2012, further enhanced the standard by incorporating 3.0 compatibility for improved bandwidth. Initially positioned as freely available to encourage widespread implementation, the openness of MXM shifted in the when , as the controlling entity of MXM-SIG, restricted access to the full specifications to corporate members only, requiring nondisclosure agreements and limiting public dissemination to curb potential misuse and protect . This change contributed to moderated adoption beyond enterprise and OEM channels, though the standard persisted in niche applications. As of 2025, the MXM-SIG remains operational but maintains a low public profile, with no official release of an MXM 4.0 specification; instead, later MXM modules have incorporated support for PCI Express 4.0 and beyond through extensions and compatible hardware implementations in embedded and industrial systems.

Technical Specifications

Form Factors and Connectors

The Mobile PCI Express Module (MXM) standard defines several form factors to accommodate varying performance needs in mobile graphics solutions, with physical dimensions tailored to balance size, power, and thermal requirements in compact systems. The first generation introduced Type I, II, III, and IV modules, each with distinct footprints: Type I measures 70 mm × 68 mm (width × length), suitable for lower-power applications with limited GPU area; Type II is 73 mm × 78 mm, enabling higher power delivery and expanded capabilities; Type III is 82 mm × 100 mm, supporting more complex designs with increased memory and processing density; Type IV, now deprecated, is 82 mm × 117 mm for even higher-end configurations. Second-generation form factors include Type A (82 mm × 70 mm) and Type B (82 mm × 105 mm). The connector design for MXM modules employs a 0.5 pitch edge-card interface, facilitating reliable insertion into the host system. Early generations, such as MXM 1.x and 2.x, utilize up to 230 contacts (with Type III extending to 232 in some cases), providing sufficient lanes for PCIe signaling and auxiliary functions like display outputs. In MXM 3.0 and later, the connector evolves to support up to 314 contacts while maintaining the core 0.5 pitch, allowing denser integration of PCIe lanes without altering the fundamental module specification. This design includes retention mechanisms, such as mounting holes and backing plates, to ensure stability and resistance to shock and in mobile environments, with tolerances limiting displacement to 0.615 under load. Mechanical features of MXM modules emphasize and thermal management, with variations in heatsink compatibility to suit different system integrations. Configurations support both shared and dedicated heatsinks, requiring a minimum 0.5 mm clearance from components to accommodate thermal solutions like spreader plates and heat exchangers. is a key aspect, as smaller form factors (e.g., Type I) can insert into slots designed for larger types (e.g., Type II or III), though adapter heatsinks may be necessary to align thermal interfaces and prevent incompatibility. The MXM 3.0 and 4.0 connectors, standardized by , enhance this by enabling denser layouts for additional PCIe lanes while preserving the overall mechanical envelope across generations.

Electrical and Interface Standards

The Mobile PCI Express Module (MXM) interface is fundamentally based on the (PCIe) standard, utilizing up to x16 lanes for high-bandwidth data transfer between the graphics module and the host system. Early implementations supported PCIe 1.0 with data rates of 2.5 GT/s, while the MXM 3.1 specification officially accommodates PCIe 3.0 at up to 8 GT/s per lane, enabling aggregate throughputs suitable for demanding mobile graphics workloads. Modern MXM modules, particularly those from industrial vendors, have extended compatibility to unofficial PCIe 4.0 and beyond, leveraging the physical interface's while requiring host system support for higher speeds. Power delivery in MXM is managed through dedicated rails to balance performance with the constraints of mobile environments. The primary PWR_SRC rail operates at 7-20 V with a maximum current of 10 A, supporting up to 200 W for high-end configurations in second-generation and later modules, though practical implementations often cap at 120-150 W to mitigate risks. First-generation high-end (HE) types relied on 12 V rails limited to approximately 75 W, as exemplified by modules like the NVIDIA FX 3600M. Auxiliary rails at 5 V (up to 2.5 A) and 3.3 V (up to 2.0 A) provide power for VRAM, fans, and peripheral components, ensuring stable operation without external connectors. Signaling in MXM employs differential pairs for PCIe transmit (TX) and receive (RX) lanes, with of 100 Ω differential (85 Ω recommended for PCIe Gen 3 and higher) to maintain over short mobile traces. Sideband signals, including PWR_GOOD for power validation, WAKE# for low-power state transitions, and thermal overtemperature indicators like TH_OVERT#, facilitate module detection, clocking via REFCLK, and dynamic configuration. These signals ensure reliable and compliant with PCIe protocols. Standards compliance for MXM is governed by the Mobile PCI Express Module (MXM-SIG) specifications, such as Revision 3.1 (), which reference the Base Specification Revision 3.0 for core electrical and mechanical requirements. The design incorporates (ESD) protection for hot-plug signals and measures, including limits up to 12 GHz and decision feedback equalization (DFE) at receiver points, tailored to the compact thermal envelopes of and embedded systems. This ensures and robustness in vibration-prone mobile applications.

Generations

First Generation

The first generation of Mobile PCI Express Modules (MXM), spanning releases from MXM 1.0 in 2004 to MXM 2.1 around 2009, marked the initial implementation of a standardized interconnect for interchangeable graphics processing units in laptops, building on 's announcement of the specification in 2004. Developed collaboratively by and major manufacturers, this era introduced MXM as an to facilitate modular GPU designs using the x16 interface, thereby shortening design cycles and enabling configure-to-order configurations across price points. This generation encompassed several form factor types optimized for varying laptop segments: Type I for basic, entry-level systems with a power envelope of up to 18 W; Type II as the standard variant supporting up to 35 W for mainstream applications; Type III (HE) for high-end setups at up to 75 W; and Type IV (deprecated, non-standard) with limited adoption. The focus was on seamless integration with NVIDIA's professional series and consumer Go GPUs, such as the Go 7 series launched in 2006, which leveraged MXM to deliver 9.0-compliant rendering with features like transparency antialiasing for enhanced visual quality. Key innovations centered on the modular GPU , which allowed OEMs like Quanta and Asustek to standardize graphics interfaces, reduce time-to-market, and support scalable performance through technologies like NVIDIA SLI for multi-GPU configurations. By providing a consistent electrical and mechanical framework, MXM enabled easier field upgrades compared to soldered solutions prevalent at the time. Despite these advances, the first generation faced constraints inherent to early mobile hardware, including a relatively low power envelope that confined implementations to mid-range GPUs incapable of matching desktop-level performance. Native support was limited to PCI Express 1.x, restricting bandwidth and precluding compatibility with emerging PCIe 2.0 systems without modifications. Adoption was further hampered by ecosystem limitations, primarily suiting early laptop platforms from select vendors. By 2010, Type IV variants were deprecated in favor of enhanced second-generation designs offering greater power and interface capabilities.

Second Generation

The second generation of Mobile PCI Express Modules, beginning with MXM 3.0 around 2009, introduced significant advancements in form factors, power delivery, and interface capabilities to support higher-performance graphics in compact mobile systems. MXM 3.0 defined two primary module types: Type A, measuring 82 mm × 70 mm with a power budget of up to 55 W, and Type B, measuring 82 mm × 105 mm with support up to 200 W, enabling more robust GPU implementations while maintaining compatibility within the ecosystem. In 2012, MXM 3.1 extended these specifications by incorporating PCIe 3.0 support at 8 GT/s across up to 16 lanes, doubling the bandwidth potential over prior generations and facilitating faster data transfer for demanding applications. Key enhancements included a theoretical power budget increase to 200 W via the PWR_SRC rail (7-20 V at up to 10 A), though practical implementations typically capped at around 150 W to align with thermal constraints in designs; this allowed for improved thermal interfaces, such as refined spreader plate designs with maximum temperatures of 90°C for GPU and memory components, supporting denser integration in thinner . Additionally, MXM 3.1 ensured by allowing Type A modules to function in Type B slots through shared electrical and mechanical interfaces. These developments shifted focus toward enabling desktop-class GPU in mobile form factors, exemplified by NVIDIA's GTX 10-series implementations like the GTX 1060 and GTX 1080 MXM variants, which delivered up to 6-8 TFLOPS of FP32 compute while fitting within the power envelope. Support for up to 16 GB of GDDR5/GDDR6 further enhanced capabilities for graphics-intensive tasks, with memory buses scalable to 256-bit widths. As the foundational standard for modern MXM deployments, the second generation remains relevant in the 2020s through extensions like PCIe 4.0 in industrial implementations, such as ADLINK's EGX-MXM series, which leverage MXM 3.1 Type A/B form factors for AI and edge computing with up to 16 GB GDDR6 and 115 W TGP.

Adoption and Compatibility

Use in Laptops and Systems

The Mobile PCI Express Module (MXM) found its primary adoption in high-end laptops between 2005 and 2015, particularly in gaming and professional workstations where upgradeability was valued. Manufacturers like Dell integrated MXM slots into models such as the Precision M series (e.g., M6500, M6700, and M6800) for CAD, 3D rendering, and engineering tasks, while Alienware utilized the standard in gaming rigs like the m17x and m18x series to support powerful discrete graphics. This era saw peak implementation through partnerships between NVIDIA, AMD, and OEMs, enabling standardized GPU swaps that extended device longevity and reduced development costs for vendors. System integration of MXM required dedicated slots on motherboards, typically positioned to accommodate the module's form factor alongside cooling solutions, which limited its use to thicker designs. It became prevalent in professional systems for compute-intensive applications like and , but was uncommon in ultrabooks or slim consumer notebooks due to spatial and constraints. First- through third-generation MXM modules, supporting evolving PCIe standards, were commonly deployed in these configurations to balance performance and power efficiency. Market trends shifted post-2015, with MXM adoption declining sharply as laptop designs prioritized thinness and portability, leading OEMs to favor soldered GPUs for better integration and heat management. NVIDIA's introduction of Max-Q technologies in 2017 further accelerated this trend by optimizing for low-power, non-upgradable architectures, while phased out MXM reference designs for newer mobile chips. By the late 2010s, MXM was largely confined to niche high-end segments, reflecting broader industry moves toward integrated or fixed graphics solutions.

Compatibility Issues and Solutions

One major compatibility challenge with Mobile PCI Express Modules (MXM) arises from generational differences in form factors and connectors. For instance, first-generation MXM cards, which adhere to earlier connector pinouts and physical dimensions, are often incompatible with MXM 3.0 or later slots due to changes in length, keying notches, and overall module height, preventing direct insertion without modification. Additionally, variations between MXM 3.0 and 3.1 specifications, such as support for versus LVDS interfaces, can lead to signal mismatches when attempting cross-generational upgrades. Thermal management poses another significant hurdle, primarily due to inconsistent heatsink designs across MXM implementations. Although the MXM standard defines Type A and Type B variants for standardized heatsink mounting and retention screw layouts, many manufacturers deviate from these by customizing heatsink shapes and mounting points to fit specific , resulting in thermal throttling or inadequate cooling when swapping modules. This lack of adherence exacerbates overheating in compact environments, where airflow is limited. Power delivery and BIOS limitations further complicate MXM deployment. Host systems may enforce power caps below the module's rated thermal design power (TDP), typically 60-100W, due to variations in power pin configurations or insufficient voltage regulation (7-20V range), potentially causing instability or reduced performance. firmware often fails to detect or fully support non-OEM modules, requiring updates to enable proper initialization, while vendor-locked video BIOS (VBIOS) on modules like cards restricts operation to specific host systems. Solutions to these issues include third-party adapters and community-driven modifications. Adapters, such as those bridging MXM to PCIe or NVMe interfaces, allow limited cross-generational or external use by rerouting signals and power, though they do not fully resolve pinout discrepancies. For and VBIOS challenges, users employ tools like NVCleanInstall for driver modifications or VBIOS flashing to unlock power limits and improve detection, often guided by MXM-SIG electromechanical specifications that outline slot design and power guidelines for better interoperability. The waning adoption of MXM can be attributed to manufacturers' shift toward soldered or integrated GPUs, prioritizing , thinner designs, and control over upgradability. Post-2010s, restricted access to MXM specifications and the rise of technologies, such as NVIDIA's Max-Q and Intel's tailored embedded solutions, further diminished the standard's openness, with no new consumer MXM cards beyond the RTX 30 series.

Notable MXM Cards

NVIDIA Implementations

has been a primary driver of the Mobile PCI Express Module (MXM) standard since its inception, developing numerous graphics cards that adhere to MXM specifications for enhanced upgradability in mobile workstations and high-performance laptops. The company's implementations span consumer and professional series, with the latter optimized for CAD, 3D modeling, and rendering applications through certified drivers that ensure stability and precision in professional workflows. 's MXM cards have historically commanded a dominant , reflecting their widespread adoption in upgradeable systems. In the first generation of MXM (circa 2004-2006), NVIDIA introduced several foundational cards compatible with MXM-I and early standards, focusing on balancing performance and power efficiency for emerging mobile platforms. The , launched in November 2004, was among the earliest high-end MXM implementations, featuring the NV41 GPU on a with 256 MB GDDR3 memory and a maximum power draw of 45 W, enabling 9.0c gaming in laptops like the XPS. Following in 2005, the Quadro FX 2500M targeted professional users with its G71-based architecture on 90 nm, offering 512 MB GDDR3, support for 2.0, and a 45 W TDP, making it suitable for CAD and visualization tasks in mobile workstations. By 2006, the lower-power Quadro NVS 120M emerged as an entry-level professional option on the G72M core (90 nm), with 128 MB DDR2 memory and just 10 W consumption, prioritizing multi-display support for business applications over raw graphics power. Transitioning to second-generation MXM (2009 onward), NVIDIA's offerings scaled up performance while adhering to MXM 3.0 (Type B) interfaces, incorporating advanced features like DirectX 11 and higher for demanding gaming and . The GeForce GTX 285M, released in 2009, utilized the GT215 GPU (55 nm) with 1 GB GDDR5 and a 75 TDP, delivering significant improvements in performance for laptops. In 2011, the GeForce GTX 675M advanced this lineage with the GF114 chip (40 nm), 2 GB GDDR5, and 100 power envelope, supporting 2.0 x16 and enabling smooth 1080p gaming in systems like workstation-class portables. By 2016, the GeForce GTX 1080 Mobile represented a peak in this era, based on the GP104 (16 nm) with 8 GB GDDR5X, up to 150 TGP via MXM 3.1, and Pascal enhancements for VR and 4K rendering. Entering the 2020s, NVIDIA's MXM implementations have embraced and architectures, with higher TGPs and PCIe 4.0 integration for AI-accelerated professional and gaming workloads. The RTX 3080 Mobile MXM, introduced in 2021, leverages the GA104 GPU (8 nm) with 16 GB GDDR6 and a maximum 165 W TGP, supporting ray tracing and DLSS for high-fidelity visuals in upgradeable laptops. Vendors like X-VSION have offered RTX 4090M and RTX 4080M MXM modules based on the AD102 and AD103 GPUs (4 nm), with up to 16 GB GDDR6 and 175 W TGP, supporting advanced ray tracing and AI features in industrial applications as of 2024. More recently, in 2024, the RTX 2000 Ada Generation MXM module arrived as a professional-focused card on the AD107 (4 nm), featuring 8 GB GDDR6, PCIe 4.0 x8 interface, and a 60 W TGP, optimized for embedded systems and CAD with support. These developments underscore NVIDIA's continued emphasis on /RTX successors for professional reliability, maintaining MXM's relevance in specialized .

AMD and Other Implementations

has produced several notable MXM-compatible graphics modules, primarily targeting gaming and professional workstations. The HD 5870, launched in 2010, was available in a 100W Type II MXM configuration, featuring 800 shader cores, 1 GB on a 128-bit bus, and support for DirectX 11, making it suitable for high-performance mobile gaming setups. In 2012, introduced the FirePro M4000 as a 50W MXM 3.0 module with GCN , 512 stream processors, 1 GB , and certified drivers for professional applications like CAD and . More recently, in 2019, the RX 5500M appeared in 60W MXM variants based on the 7 nm Navi 14 , offering 1,408 stream processors, 4 GB , and up to 4.6 TFLOPS of for mid-range gaming. Intel entered the discrete MXM space with its Arc A-series in 2023, leveraging the Alchemist generation and Xe-HPG architecture for modular . The Arc A380E MXM module, for instance, provides 75W operation in select configurations, 8 Xe-cores, 128 XMX engines for AI acceleration, 6 GB GDDR6 , and PCIe 4.0 x8 interface, emphasizing low-power akin to integrated solutions but with enhanced AI and capabilities for embedded systems. Other vendors have offered limited MXM implementations, often for niche or legacy applications. S3 Graphics developed early modules like the GammaChrome series in 2005, featuring MXM II form factors with under 12W TDP, 128-bit cores, and basic 3D acceleration for low-power mobile devices. Overall, non-NVIDIA MXM models are predominantly from and emerging from , with sporadic custom embedded options from lesser-known suppliers. 's offerings have historically emphasized value-oriented gaming performance, while 's focus on low-power AI and embedded use cases highlights their complementary roles in the MXM ecosystem.

Other Applications

Embedded and Industrial Uses

The Mobile PCI Express Module (MXM) finds significant application in embedded systems through standards like Qseven and SMARC, where its connector is repurposed for general high-speed I/O rather than full graphics acceleration. Qseven modules, designed for compact (COM) solutions supporting both ARM and x86 processors, utilize the MXM-II connector with 230 pins in a 0.5 mm pitch to enable cost-effective, high-speed PCI Express integration in mobile and low-power embedded designs. Similarly, SMARC (Smart Mobility ARChitecture) modules employ the MXM 3.0 connector with 314 pins to support ultra-low-power ARM and x86 SoCs in space-constrained IoT applications, providing versatile I/O for sensor processing and edge connectivity while prioritizing energy efficiency over intensive graphics workloads. In industrial settings, MXM modules enhance performance in specialized domains such as medical imaging, defense, and AI edge computing. For medical imaging, embedded MXM GPUs accelerate AI-driven analysis in devices like retinal tomography systems and portable ultrasound units, delivering real-time image processing with NVIDIA architectures for diagnostic precision. In defense applications, rugged MXM-based boards like Mercury Systems' GSC6202 6U OpenVPX GPGPU coprocessor integrate NVIDIA Quadro GPUs to support real-time sensor fusion, electronic warfare, and AI inference in harsh military environments, offering up to 12.8 TFLOPS and 32 GB GDDR5 memory across dual GPUs. For AI edge devices, solutions like Advantech's SKY-MXM-2000A module, released in 2024, incorporate the NVIDIA RTX 2000 Ada Generation GPU with 3072 CUDA cores and 8 GB GDDR6, enabling compact, high-performance inference in industrial automation and robotics. MXM's advantages in these contexts stem from its and ruggedization potential, allowing upgrades in fixed industrial installations without full system replacement and ensuring reliability in vibration-prone environments through specialized carrier cards that withstand shock, extended temperatures, and mechanical stress. This modularity supports long-term deployment in embedded and industrial systems, where compatibility facilitates seamless integration with diverse host processors.

Recent Developments and Future Outlook

In 2023, ADLINK introduced the MXM-AXe module, an MXM 3.1 Type A graphics solution based on Intel Arc A-series mobile GPUs (Alchemist architecture), featuring up to 8 Xe cores, 8 ray-tracing units, 128 execution units, 4 GB GDDR6 memory, and PCIe 4.0 x8 interface at a 35 W total graphics power (TGP), targeted for AI and embedded applications. This module marked one of the first discrete graphics options in the MXM form factor leveraging Intel's discrete GPU technology, enhancing AI inference capabilities in compact systems. By 2025, Cincoze expanded MXM integration in industrial computing with the GM-1100 series embedded GPU computers, supporting 14th-generation processors alongside NVIDIA MXM 3.1 GPU modules for high- edge applications. These systems deliver up to three times the computing of prior generations, with features like 2.5 GbE LAN, 20 Gbps USB 3.2, and robust cooling for demanding environments, earning recognition including the 2025 Red Dot Design Award and Vision Systems Design Innovators Award. Revival efforts for upgradable graphics in consumer laptops gained momentum in 2025 through Framework's Laptop 16, which introduced swappable GPU modules supporting NVIDIA GeForce RTX 5070 Laptop GPUs at up to 100 W TGP with 8 GB GDDR7 memory, emphasizing modularity and user repairability. This design aligns with growing e-waste regulations and right-to-repair initiatives, allowing seamless upgrades from to options without full system replacement. prototypes have explored adapting standard MXM modules to the Framework platform, further bridging legacy MXM compatibility with modern modular hardware. Looking ahead, MXM's role in edge AI is expanding, as seen in modules like the ADLINK EGX-MXM-P5000, which provides 2048 CUDA cores, 16 GB GDDR5 memory, and 6.4 TFLOPS peak FP32 performance for local AI processing in bandwidth-constrained environments. While no official MXM 4.0 specification has emerged to support PCIe 5.0 or 6.0, ongoing PCIe advancements could enable future iterations for higher-bandwidth applications, though challenges persist in balancing MXM's modularity costs against integrated soldered GPUs in mainstream designs.

References

  1. http://gyctech.cn/uploadpic/2021070115414714aq.pdf
  2. https://www.bvm.co.uk/faq/mobile-pci-express-modules-mxm-modules/
  3. https://www.aetina.com/products-list.php?t=23
  4. https://www.assured-systems.com/faq/what-is-mxm/
  5. https://www.techbuzz.ai/articles/framework-delivers-3-minute-laptop-gpu-swaps-with-rtx-5070-upgrade
  6. https://www.aetina.com/products-detail.php?i=694
  7. https://www.pcworld.com/article/431809/these-upgrade-kits-swap-out-your-laptops-old-graphics-with-a-new-nvidia-gpu.html
  8. https://ieee-hpec.org/2012/index_htm_files/17_ID25.pdf
  9. https://d18rn0p25nwr6d.cloudfront.net/CIK-0001045810/ac23c11f-a5bf-40d7-be57-051de2280872.pdf
  10. https://www.igorslab.de/en/mxm-mobile-pci-express-module-the-death-of-an-open-standards-for-graphics-modules-in-notebook-and-the-behind-behind/
  11. https://hothardware.com/reviews/nvidias-mxm-graphics-module
  12. https://hackaday.com/2024/04/18/mxm-powerful-misused-hackable/
  13. https://www.amphenol-cs.com/product-series/mxm-3-0.html
  14. https://www.adlinktech.com/en/mxm-gpu-modules
  15. https://www.notebookcheck.net/NVIDIA-Quadro-FX-3600M.8834.0.html
  16. https://www.theregister.com/2004/05/17/nvidia_mxm/
  17. https://http.download.nvidia.com/ndemand/product_overview/NV_PO_GeForce_Go_7_MAY06_HR.pdf
  18. https://lekensteyn.nl/files/docs/mxm-2.1-software-spec.pdf
  19. https://graphicscardhub.com/mxm-graphics-card-gpu-list/
  20. https://lekensteyn.nl/files/docs/mxm-graphics-module-software-specification-version-3.0-revision-1.1.pdf
  21. https://www.ebay.com/b/NVIDIA-GeForce-GTX-1060-DDR5-6-GB-Memory-Computer-Graphics-Cards/27386/bn_110679032
  22. https://videocardz.com/newz/socketed-cpu-laptops-and-high-end-mxm-graphics-are-fading-into-history
  23. https://www.bvm.co.uk/faq/mxm-gpu-modules-compact-edge-ai-computing/
  24. https://globalamericaninc.com/are-mxm-graphics-worth-it-in-2022/
  25. https://www.notebookcheck.net/Mobile-Graphics-Cards-for-Professional-Users-AMD-FirePro-NVIDA-Quadro.21333.0.html
  26. https://www.techpowerup.com/gpu-specs/?f=manufacturer_NVIDIA&f=slot_mxm
  27. https://www.techpowerup.com/gpu-specs/geforce-go-6800.c1315
  28. https://www.techpowerup.com/gpu-specs/quadro-fx-2500m.c1444
  29. https://www.techpowerup.com/gpu-specs/quadro-nvs-120m.c1470
  30. https://www.notebookcheck.net/NVIDIA-GeForce-GTX-285M.23822.0.html
  31. https://www.techpowerup.com/gpu-specs/geforce-gtx-675m.c343
  32. https://www.techpowerup.com/gpu-specs/geforce-gtx-1080-mobile.c3683
  33. https://laptopmedia.com/video-card/nvidia-geforce-rtx-3080-laptop-165w/
  34. https://www.x-vsion.com/product/mxm-embedded-module-rtx-4090m-gpu-beyond-fast/
  35. https://www.x-vsion.com/product/mxm-embedded-module-rtx-4080m-gpu-beyond-fast/
  36. https://www.aetina.com/products-detail.php?i=664
  37. https://www.nvidia.com/en-us/products/workstations/rtx-2000/
  38. https://www.techpowerup.com/gpu-specs/mobility-radeon-hd-5870.c338
  39. https://www.notebookcheck.net/AMD-FirePro-M4000.77893.0.html
  40. https://www.techpowerup.com/gpu-specs/firepro-m4000.c1677
  41. https://www.techpowerup.com/gpu-specs/radeon-rx-5500m.c3460
  42. https://www.x-vsion.com/product/mxm-amd-radeon-rx-5500-xt-4gb-gddr6/
  43. https://www.aetina.com/products-detail.php?i=684
  44. https://www.hexus.net/tech/news/laptop/1039-cebit-2005-s3-gammachrome-mxm-ii/
  45. https://www.directindustry.com/industrial-manufacturer/mxm-graphics-card-262472.html
  46. https://www2.advantech.com/embedded-boards-design-in-services/com/product-Qseven.aspx
  47. https://www.adlinktech.com/en/computer_on_modules_smarc
  48. https://www.advantech.com/en-us/resources/news/advantech-partners-with-miis-to-upgrade-retinal-tomography-with-nvidia-ai-technology
  49. https://www.mrcy.com/products/compute/rugged-edge-servers/boards/gpgpu-video-graphics/gsc6202-6u-openvpx-gpgpu-es-board
  50. https://www.advantech.com/en-us/resources/news/advantech-releases-sky-mxm-2000a-powered-by-high-performance-nvidia-ada-lovelace-architecture
  51. https://ieee-hpec.org/2012/index_htm_files/Straznicky.pdf
  52. https://www.adlinktech.com/en/news/mxm-axe-graphic-module-intel-arc
  53. https://www.techpowerup.com/305655/adlink-puts-intel-arc-a-series-gpus-on-mxm-form-factor
  54. https://www.eejournal.com/industry_news/adlink-launches-mxm-axe-a-first-discrete-graphics-mxm-module-powered-by-intel-arc-gpu1/
  55. https://www.cincoze.com/en/bulletin_info.php?id=227
  56. https://www.cincoze.com/en/goods_info.php?id=546
  57. https://www.businesswire.com/news/home/20250527954879/en/Double-Glory-Cincoze-MXM-GPU-Computer-Wins-Red-Dot-Design-and-Vision-Systems-Design-Innovators-Awards
  58. https://www.embedded.com/cincoze-brings-ai-to-the-edge-with-mxm-gpu-computers-for-industrial-and-defense-sectors
  59. https://frame.work/products/laptop16-graphics-module-nvidia-geforce-rtx-5070
  60. https://spectrum.ieee.org/upgrade-laptop-gpu-framework-nvidia
  61. https://www.gamesradar.com/hardware/laptops/frameworks-modular-gaming-laptop-finally-gets-an-nvidia-gpu-upgrade-kit-and-the-whole-thing-is-starting-to-make-sense/
  62. https://community.frame.work/t/framework-16-to-mxm-gpu-v0-1-prototype-design/74018
  63. https://www.adlinktech.com/products/embedded_graphics/embedded_mxm_modules/egx-mxm-p5000?lang=en
  64. https://cdn-assets.inwink.com/d5ab79d8-fb20-4b79-81c6-882153a86c6c/402034df-03b5-4200-b415-4fb444f7a14e1
  65. https://www.intel.com/content/www/us/en/gaming/resources/what-is-pcie-4-and-why-does-it-matter.html
Add your contribution
Related Hubs
User Avatar
No comments yet.