We recently had a chance to talk with representatives from NVIDIA regarding a new initiative they've spearheaded, with the goal of bringing a consistent graphics interface to PCI Express equipped notebooks.  Historically, the graphics processors used in laptops / notebooks have been impossible, or prohibitively expensive, to upgrade.  A few companies, like Alienware and Toshiba, have recently introduced upgradeable notebooks, but they each use different standards.  Alienware's parts can't be used in Toshiba's notebooks, and vice-versa.  The video cards used in desktop systems, however, can be easily upgraded because there is a consistent interface standard (i.e. AGP) and they usually offer performance that is an order of magnitude greater than anything concurrently available in the mobile market.  NVIDIA plans to eliminate this shortcoming with their new MXM, or Mobile PCI eXpress Module.

The timing of this initiative could not be better, with notebook manufacturers already preparing for the transition to PCI Express.  The move away from AGP to PCI Express means every notebook manufacturer will have to redesign their graphics interfaces anyway, so why not settle on a standard in the process?  With this in mind, NVIDIA worked with numerous notebook manufacturers, like Quanta, Wistron, AOpen and FIC (among others), to define the MXM specifications and components.  The proposed benefits of MXM are...

•Faster time to market, with the latest graphics technology
•One notebook design, sold into many notebook products
•One notebook design that can use any graphics solution, from any vendor
•Notebooks whose graphics technology can be upgraded through the channel or, potentially by consumers

MXM also had to offer solid compatibility, lower costs, smaller sizes and a high level of signal integrity.  These are some lofty goals, but if MXM is widely adopted and these goals are met, it could mean the mobile graphics market is headed in a totally new direction.

MXM Components The Connectors, The Board and its I/O Capabilities...

  
CLICK ANY IMAGE FOR AN ENLARGED VIEW

Motherboards equipped with an MXM slot will have a connector like the one pictured above to accommodate different graphics modules.  Its footprint is very small, at 72mm long and 11mm wide, but its height will vary between 5mm, 2.7mm, and 1.5mm depending on the vendor.  The MXM slot has 230 pins, and will accommodate full 16-lane PCI Express signals, as well as all display I/O signals.  Having the display I/O signals running through the MXM slot means notebook manufactures can simply route the display signals to whichever connectors are installed on the motherboard (VGA, DVI, HD, S-Video, Etc.), which could make multi-display capable laptops more common.

The modules themselves come in three "flavors" - MXM-I, MXM-II, and MXM-III.  The proposed design is installed by sliding the module into the slot at a 20 - 30 degree angle, rotating it into its final position (in contact with the cooling solution) and locking it into place with four screws.  Also note that the stand-off locations for the screws will be the same for all MXM graphics modules, which would allow one motherboard to support all three MXM graphics module types.  Each of these modules will use the same slot, and each is backwards compatible with the other, but they have different physical attributes and can accommodate different components...

As you can see in the chart above, each module varies in physical size, max power consumption, memory configuration and maximum GPU size.  MXM-I is designed for entry-level notebooks, MXM-II for medium form factors with somewhat higher performance, and MXM-III is designed for high-performance systems.  MXM-I and MXM-II slots were also designed with the same thermal considerations in mind, but MXM-III slots will be equipped with beefier cooling solutions.  At a maximum of 35W for an MXM-III slot, we obviously won't be seeing the power-hungry GeForce 6800 Ultra populating an MXM module anytime soon, but NVIDIA did tell us that this initiative will facilitate the introduction of their GeForce Go 6 Series of mobile GPUs derived from the NV40 architecture.  I'm sure we'll have some more information regarding mobile NV40 derivatives as it becomes available...

To help keep pace with their desktop counterparts, notebook ODMs (Original Design Manufacturers) need this type of standard to help accelerate the design cycle.  There has typically been a 3 - 6 month lag between a product's introduction and the time it was available to consumers.  This is the case because proprietary modules were created for particular notebook designs, but not for all notebooks.  So a proprietary module for a Dell Inspiron 8600, for example, may not work with an Inspiron 5150M, and each needed to be designed, produced, tested and brought to market individually.

Multiple Designs for Multiple Laptops...

A Dell Designed Graphics Module, one populated with an NVIDIA GPU, the other ATi.

An advantage to having a standard like MXM is that hardware manufacturers like NVIDIA will be able to release reference designs for MXM modules using their latest graphics hardware. This in turn will give notebook manufacturers the ability to plug any MXM reference module into their systems if an MXM connector has been properly implemented on the system's motherboard.  This is the same model used in the desktop space, where any AGP graphics card will work in any AGP slot.

Proposed Thermal Solutions

A component of the MXM standard dictates that an MXM can automatically detect power supply and thermal limitations of the notebook platform and can automatically throttle its clock speeds to stay within the power and thermal limits.  If the notebook's power supply and MXM thermal solution are equal too or exceed the MXM's requirements, the module will run at its full speed.  The MXM system information data contains the necessary information for MXM throttling.  This data can be stored in either the BIOS or in the MXM system information ROM installed on the notebook's motherboard.

The MXM platform has also been designed to work in conjunction with motherboard chipsets that have an Integrated Graphics Processors, or IGP.  A notebook with an IGP that is also designed to support MXM graphics modules will work with or without a MXM graphics card installed.  If no MXM graphics module is present, however, the installation of a loop through card is required to complete circuit paths.

Whether or not the MXM standard gets widely adopted remains to be seen, but considering this is an open standard and NVIDIA already has multiple launch partners on board including some of largest Taiwanese ODMs, there is a good chance we'll see MXM equipped notebooks available relatively soon.  NVIDIA tells us notebooks powered by MXM graphics cards will start appearing later in the second half of this year, coinciding with the move to PCI Express on the desktop.  The big question is weather or not the other players in mobile graphics, namely ATi, S3, VIA and Neomagic, will produce MXM graphics modules.  We haven't been able to get any official comment, but it may not matter what these other companies think of MXM.  There's nothing to stop any of the Taiwanese notebook ODMs from designing their own MXM cards using GPUs from these other companies.  The process shouldn't be much different than what they've already been doing for years.  We're going to be talking with ATi and S3 shortly, and will be sure to fill you in as we find out more.