CheckMag | Give us real SFF, Nvidia, you cowards.

I could be measured. I could lay my thoughts down in a succinct way. I could say that the guideline dimensions of 304 x 151 x 50 mm, a form factor longer and deeper than what we used to call "full size", is a pretty absurd stretching of the word "small".

Or I could say: calling 2.3 litres in volume small form factor is pure sin.

It's more than twice the volume of the first reference GTX TITAN, itself a height of decadence for its time, or the RTX 2080 Ti that kicked off the current era of power and pricing excess. It's more volume than some entire systems; bigger than six out of Notebookcheck's top ten Mini PCs, and all of the other four have discrete GPUs of their own. You could nab a Legion Go (say, on Amazon) for that space - even Lenovo's giant amongst gaming handhelds is smaller.

We've done small graphics cards before. We've even done small high-end graphics cards before. The Vega 56 Nano sat around the same wattage of heat to cool as the RTX 4070 does today, except it got by with a half-length, two-slot cooler. Galax's GTX 1070 Katana did its job with just a single PCIe slot in width - and though it only had 150 watts to deal with, its newer Lovelace counterpart often sits below nominal TDP, not far above the 1070's power draw. Don't get me wrong, neither of those cards were particularly fantastic in terms of noise and performance. But they make these new guidelines seem almost laughable; it simply feels like board partners are conceding defeat to their own laziness, and needing Nvidia to cover their retreat from design challenges that they exist to solve.

Of course, Nvidia is the one pitching this idea, and it's all the crazier for it to do so. GeForce's advantage in efficiency in comparison to its Radeon competition has been entrenched for well over a decade - the GTX 1060 and RTX 2080 used a third less power than the RX 580 and Radeon VII they were pitted against - and though TDP on many modern high-end RTX GPUs has shot up, it's been shown that they could comfortably run at far low power with no ill effects. So the decision to disregard Nvidia's own competitive advantage - to not even try and offer products in compact form factors that AMD (or Intel, for that matter) simply can't, is just baffling.

Look. I know there are case partners already signed up. I know designs made to accommodate large GPUs are getting big now (pardon the pun), that at Computex we saw mainstream brands getting into sandwich-style cases that were previously dominated by boutique designs shopped around on forums. I know that it's possible to fit triple-slot foot-long GPUs into a sub-10L FormD T1, because that's the build I'm writing this on.

But I also know that the sacrifices made to maximise GPU space made it a right pain to put together. And I know new builders would happily save the expense of SFX power supplies, or the hassle of trying to stop cables from taking up every inch of internal volume, if OEMs just set aside the time and resource to design something truly compact. But instead, we get Nvidia moving the goalposts so that anything short of an RTX 3090 or 4090 counts as "small".

Whatever. I undervolt the damn thing to 750 mV anyways.