Introduction

One of the key reasons why the desktop PC has been so successful over the years is its unrivaled flexibility. With industry heavyweights such as Intel and Microsoft (among others) driving new technologies and interfaces, the PC has been able to keep up with the latest innovations without batting an eye. USB/FireWire, and Serial ATA have all seamlessly made their way to the PC, and soon, PCI Express will as well. This is because the entire industry has rallied around these technologies, creating an open standard that’s accepted by all. As a result, when ATI or NVIDIA releases a new graphics chip, it can be easily dropped into existing PC designs with little or no modifications.

Life isn’t quite so easy on the mobile side however. There is no reference specification that everyone adheres too. When a new mobile graphics product is introduced, it may or may not be a drop-in replacement for its predecessor. Manufacturers have different ways of dealing with the space, power, and thermal constraints that exist in their mobile lineup, resulting in proprietary designs that can’t be easily upgraded. But what if a standard interface was available that all manufacturers could adhere to? This is the premise behind NVIDIA’s new Mobile PCI Express Module, dubbed MXM. NVIDIA has worked with its mobile partners in producing a new standard interface for its mobile PCI Express graphics products.

One of the key sticking points for MXM was the connector. Manufacturers wanted a connector with a small motherboard footprint, yet at the same time provided high signal integrity, reliability, and was inexpensive to implement. NVIDIA has defined three MXM specifications, each with varying footprints and designed to appeal to different markets.

On the high-end is MXM-III. MXM-III is the largest, at 82mm wide and 100mm long supporting up to eight memory modules (for a total of 256MB) and supporting 256-bit wide memory interfaces like those found on GeForce FX 5900 and GeForce 6800 series GPUs. MXM-III will be used in desktop replacement notebooks. For the mainstream market, NVIDIA and its partners have prepped MXM-II. The module is a little smaller at 73x78mm, but still supports up to 256MB of memory. Since this is a mainstream offering, MXM-II is limited to 64 and 128-bit memory interfaces.

Finally, for the thin and light segment is MXM-I. Up to four modules are supported in MXM-I limiting maximum memory size at 128MB, but still boasts the flexibility of 64 and 128-bit memory interface support. And at 70x68mm, MXM-I is considerably smaller than MXM-III.

Despite their different sizes, it’s important to note that all MXM types have identical mounting holes, so manufacturers can implement a smaller MXM board type if they wish. NVIDIA has also laid out thermal guidelines and reference cooling designs for each MXM type, in fact, MXM-I and MXM-II reference module designs have the same thermal cooling solution.

The 230-pin MXM connector supports all display I/O signals (VGA, DVI (2 output support), S-Video, component HD video output, and 24-bit LVDS display panel, as well as a pass-through connection for integrated graphics. If an integrated graphics solution is already present in the system, the MXM module can override the integrated graphics -- just like plugging an AGP graphics card in a desktop PC with an integrated graphics solution. The connector also supports the maximum 16-lane PCI Express specification.

Of course, the idea of a notebook with a graphics module isn’t new, ATI and NVIDIA have been doing it for years with original design manufacturers (ODMs). In fact, Alienware has provided a user-upgradeable graphics module for a little over a year now. The difference with MXM is that the module isn’t vendor-specific (i.e. tailored for one manufacturer), any OEM or ODM that implements NVIDIA’s MXM is compatible with others. So in theory you could transfer your graphics module from one MXM system to another MXM system, regardless of manufacturer. Or, if you wish to upgrade, you could replace your current MXM module with a faster one once it’s released (assuming the notebook design you choose provides a chassis that supports a design that can be upgraded by the user in the field, most likely the Alienwares and VoodooPCs of the mobile computing world will be more likely to provide this capability than one of the more conservative OEMs).

Summing it up

NVIDIA’s MXM module is designed to bring some of the flexibility found on the desktop PC to the mobile graphics market. It’s important to note that MXM isn’t an open industry standard however. ATI wasn’t consulted in the design of MXM, although in our discussion with them, they indicated that they would be willing to work with an ODM that wishes to implement a RADEON MOBILITY VPU into an existing MXM design. ATI has their own standard that they’ll be rolling out with their PCI Express products once they’re launched.

So when will MXM debut? NVIDIA and its partners have just finalized the spec so we won’t see retail products until the second half of this year, most likely beginning this fall. So far NVIDIA has signed on Aopen, Alienware, Arima, Clevo, FIC, Quanta, Tatung, Uniwill, and Wistron with more to follow (keep in mind that part of this holdup is on the system chipset, which hasn’t been officially announced by Intel or anyone else in the mobile market).

According to NVIDIA, MXM will be “the key delivery vehicle for our next generation GeForce 6 Go Series GPUs”, indicating its importance going forward for the company. And with MXM providing a consistent interface and reference design across all PCI Express notebooks that supports multiple segments and can be adapted for use with their upcoming PCI Express GPU lineup we see no reason why they shouldn’t. MXM is definitely a positive step for the mobile industry and should lower development costs for manufacturers as well as providing reduced time to market. The end result is a win for the consumer and the ODMs and OEMs.