Originally posted by TechGeek
View Post
First off, it's nearly impossible (if not just impossible) for MXM GPUs to have the same TDP as their desktop counterparts. High-TDP GPU chips require high-power VRMs. To do that on the small space offered by an MXM card will require some really expensive and top-dollar parts to get really good efficiency. Otherwise the VRM could overheat. In contrast, a regular PCI-E card can have more space, and thus cheaper parts could be used. And in the case of really really high output power like RTX 4090 desktop chips, it just might be impossible to make a VRM that fits on an MXM module. So the limited space of an MXM card will certainly either a) drive up the costs of the card or b) limit the output power... or in many cases c) both.
Next... and let's imagine the VRM issues above do not exist for a moment... the question is, how are you going to get that kind of power to an MXM card? Add 6/8-pin PCI-E connectors? If so, then it's no longer MXM format - at least not something you can stick in a laptop again. Also, the MXM edge connector will surely not be able to handle the power required by anything higher than a 100-120W TDP GPU, if even that much (though such is the case with PCI-E too already.)
And lastly... let's not forget that laptop GPUs aren't always exactly the same as their full-blown desktop counterparts. Either they have parts of the core cut off or run at a lower frequency to use less power. But in the cases where they are, then the mobile GPU essentially has to be "creme of the crop" silicon - i.e. the chips that can run at the highest possible clocks with the lowest possible voltages. So what happens to all of the GPU chips of the same model that can still run at the same frequency as the "top" silicon parts, but with much higher TDP requirements (due to needing higher voltage)? They either would have to be nerfed down in performance in order to meet the same TDP spec or not used at all. The latter would be a major loss for any production, so clearly that's undesirable. And the former is a missed opportunity to sell the same silicon as a better-performing card.
To see what I mean, take for example a regular RTX 2060 and compare it to a mobile RTX 2060.
regular RTX 2060:
https://www.techpowerup.com/gpu-spec...rtx-2060.c3310
mobile RTX 2060:
https://www.techpowerup.com/gpu-spec...0-mobile.c3348
You can see above that while both GPUs use the same exact silicon die, the mobile version is running at a much lower frequency (960 MHz) vs. that of the desktop version (1365 MHz). Same goes for the boost frequencies. As a result, the TDP of the MXM card is also lower. But then that's a loss in performance - a whooping 30% in some cases.
So going to an MXM form factor doesn't bring any advantages to what we have already with PCI-E.
Leave a comment: