Nanoraptor
Stygian Abyss Pie Factory@NanoRaptor okay but in an ideal world, installing a gpu into a laptop would be as easy as slotting something into a pcmcia slot
Ian Douglas Scott@psilocervine @NanoRaptor I suppose a big part of the challenge would be standardizing a card that integrates with the laptop's thermal system to disipate the heat from the GPU.
EmpathicQubit
DrYak@empathicqubit @ids1024 @psilocervine @NanoRaptor in the current world, @frameworkcomputer and their large swappable back modules (for discrete GPUs, etc.,) are about the only attempt I know of at swapping GPU including the thermals (by having everything self contained).
Before that MXM standardized the GPU cards, but was very limiting thermal wise (only compatible between a handful of the standard's sub-versions).
@dryak @ids1024 @psilocervine @NanoRaptor @frameworkcomputer oh i think i had an mxm card in an old asus laptop
@empathicqubit (Wikipedia has a nice table of which sub-version of MXM can fit in which slot, in case you'd be interested in upgrading that old ASUS with some more beafy second hand MXM off ebay/marketplace/etc.)
Kevin Karhan 
@dryak @empathicqubit problem with #MXM is that one needs to know the precise electrical/thermal/mechanical dimensions and demands, because MXM is a very loosely defined standard.
That being said it should be feasible for a hobbyist to put some low end 2D chip into a CardBus card
Kevin Karhan :verified: (@[email protected])
@[email protected] well, *"#shitpost"* aside there are #CardBus cards that do #PCI and PCI -> PCIe - chips like the #PEX8112 [are common](https://github.com/OS-1337/tiny486/blob/main/documentation/external/PEX8112.pdf) so it should be possible to just go CardBus -> #ExpressCard -> #PCIe. - In fact, if I were as good as @[email protected] when it comes to PCBs I'd shoved an #SM750 [into such a card](https://github.com/OS-1337/tiny486/blob/main/documentation/external/SM750.pdf) with like 2x #DVI over #MiniHDMI as output to avoid a #dongle or mechanically straining thicc connector. So the only absurd thing is the pictured miniaturization, which will likely take 10-25 years to reach that small spacial & thermal envelope. - I should really stop taking #Shitposts by #NanoRaptor seriously...
@kkarhan @empathicqubit (...or go the "PCI->PCIe bridge" route you mention (and which was used by GPU cards manufacturer around the transition from AGP & PCI to PCIe era) but break it out to a larger external module.)
Though yes, managing to fit the chipset within a PC Card form factor as you mention would have some interesting.
Also given how widely 3Dfx Voodoo chips were deployed (including on iMac Mezzanine) I am kind of surprised that 3Dfx CardBus weren't a thing back then.