Nvidia’s RTX 3060 Ti could be coming in a new version built around a different GPU, if what we’re hearing from the rumor mill is correct.
The existing RTX 3060 Ti uses the GA104 chip, but the purportedly incoming new version of the desktop graphics card will switch to the GA103, a beefier GPU, as per a tweet from HXL.
AIDA64 6.60.5918 betahttps://t.co/Z3dIq3JV7B pic.twitter.com/c6wEOUmk3yJanuary 28, 2022
As you can see above, the Twitter leaker spotted the 3060 Ti with GA103 GPU as a new entry in the release notes for the latest beta of benchmarking tool AIDA64.
Take all this with the usual helping of condiments, naturally, and consider that maybe someone at AIDA64 has made a mistake here. However, that doesn’t seem likely, as we have heard this rumor before, with the GA103 listed for the RTX 3060 Ti in an Nvidia Studio driver. (Further note that the GA103 has been spotted as the engine of an RTX 3080 Ti laptop card).
It’s also important to make it clear that while the GA103 is a higher-end GPU than the existing GA104, it will be cut-down so the spec of the theoretical new model of the RTX 3060 Ti will be the same (4,864 CUDA cores). Obviously you couldn’t have a 3060 Ti with a beefier spec than the models already on sale (or sold).
Analysis: If the new variant is happening, expect some marginal improvements
If this is happening, then the question that likely immediately pops into your mind is why is Nvidia doing this? We don’t know, of course, but can guess that it’s necessary for allocation reasons in whatever production runs are going on with Team Green’s GPUs right now.
While there will be no outward differences between these two RTX 3060 Ti models, as Tom’s Hardware (opens in new tab), which flagged this up, points out, there could be small advantages for the new 3060 Ti in theory. It might run a little cooler due to that GPU, and therefore be capable of being clocked up to higher speeds by those enthusiasts who are interested in overclocking; but any variations should be pretty marginal.
It’s never ideal when a graphics card SKU ends up with two different versions like this, as it leads to a transition period where the new model comes onto shelves as the old one still has stock going out there (albeit there isn’t much inventory of anything around right now), and buyers could be left wondering which variant they’re getting.
Nvidia would certainly need to make sure any slight differences were positive in favor of the new graphics card, and not drawbacks for it, for obvious reasons.