They were MUCH slower .. "Efficiency" = Performance per Watt
It’s worth noting that efficiency is usually very much related to the power class. Low power class are extremely efficient at idle but often poor under full load.
It begs the question having moved in power class (it’s no longer really what was a 50 power class card in power). The performance scaled up to 70% better in the more ideal benchmarks, but it took 73% more power to do that.
Since the performance and power scaled almost linearly in the ideal conditions, it seems to imply that efficiency is about the same as previous 50 class cards, and the RTX 3050 doesn’t represent actual improvements to 50 class if all it did was move up the power scale for a linear improvement. It’s more akin to comparing a GTX 1660 Super (125W) to the GTX 1650 (75W) and GTX 1050 (75W).
Yeah it’s faster but it seems like a hollow bump after doubling the price…and you also bumped the power requirements up where there were essentially no power requirements before so you could install the 50 class cards in anything with an open PCIe slot for an upgrade. Now you can’t do that.
It would be interesting to see actual efficiency comparisons to the related cards because I seem to remember the GTX 1650 GDDR6 being a bit more efficient than the GTX 1650 Super. If the RTX 3050 went slightly backwards in efficiency (it appears so, 3050 130-135W is bigger than 1660S 125W) then I see no redeeming reason to buy it. There at least needs to be improved efficiency that is in line with the cost increase above the 1660S, because I don’t value DLSS much at all over the other qualities I can get out of an expensive GPU.
Relabel it the RTX 3050 Super and give me the RTX 3050 I’d rather have. If the RTX 2050 appears, it's looking like it will be gimped in yet another way that ensures you buy Intel's new GPUs. Seeing how Alder Lake returned to undercut AMD on price, that might be a wise decision.