NVIDIA always wants to bake rather big rolls, preferably only the really big ones. But without crumbs in the mass market it does not go now once, because also with AMD (and soon also Intel) one recognized for a long time that cake dough alone does not make so correctly fat, but above all the Streusel on it. With the RTX 2050 NVIDIA now closes a little bit the open bottom hatch, because there was nothing below a RTX 3050 and 3060 in the portfolio so far.
If you ask the engineers at the board partners and the colleague from Moore’s Law Is Dead, with whom I chat from time to time, the RTX 2050 is the smallest Ampere-Sibling and not Turing, as the name might imply. So now comes a GeForce RTX 2050 as an RTX crumb monster with DLSS creaminess and a denied amp fill. This also has its charm somehow, because in it there is not only a supposed self-denial of an entry-level chip, but also clever calculation and probably even a hint of attempted humiliation.
If you analyze the situation and the possibilities of the new card, you put the GeForce RTX 2050 very deliberately against Navi24 and especially Intel’s ARC 128. So you definitely need them, even if you would prefer not to bake them that way. And while the MX550 uses a full TU117 (and thus is still based on Turing), the RTX 2050 probably relies on a kind of “1650 Max-Q replacement” from Ampere leftovers. NVIDIA will certainly not produce a special GA108, because they obviously have already finished the production of smaller chips.
So an even more slimmed down variant of the RTX 3xxx will have to do. And now we come to the calculus mentioned above. DLSS 2.x for the entry-level gamer? That makes perfect sense if AAA titles would suddenly be playable on such a small chip due to the graphics booster. At least in full HD and not quite as gorgeous. But they could well run, which you would not expect from Navi24 and also Intel’s ARC 128 so first rather. DLSS for the footy and the targeted fix for the drug shot after. DLSS can be quite addictive when the FPS suddenly just bubble up.
And what was that about humiliation? That they’re trying not to water down their lavishly pushed 3000 series with a crumb card like this is certainly kind of understandable. But it is actually also a nasty humiliation of the competitors, if you tell the not so well informed customer that even the last generation is completely sufficient to be able to easily compete with the current competitors in this performance class. Somehow you really get the feeling that NVIDIA doesn’t really like this card. But my subjective feeling is one thing, Jensen’s wallet is another. Because it will certainly continue to fill up even with such a construct.
Update from 12/23/22 – 10:30 am.
Further inquiries at AIC have revealed that NVIDIA is currently not (yet) planning this card as a real desktop card. If and when this idea will become reality also depends on the competitors and their schedule. Thus, this card does not have a real priority for the time being.