There could be two distinct variants of the Nvidia GeForce RTX 3050 8GB in the near future: those fitted with a GA106 GPU and those with a GA107 GPU. The differences between the two should be minimal, but there's the possibility that the GA107 model will be slightly more efficient than the existing GA106 one.
Nvidia released the GeForce RTX 3050 as a discrete graphics card on January 27, and it offers a moderate list of specs for the money: 20 SMs, 2,560 CUDA Cores, 80 Tensor Cores, 20 RT Cores, and a boost clock speed of 1,777MHz. Besides its 8GB of GDDR6 memory, that's all courtesy of the GA106 GPU at its core.
That's the same GPU as the RTX 3060, albeit with far slimmer specs than its bigger, mid-range sibling.
Igor's Lab now reports that there will be an RTX 3050 graphics card with the GA107 GPU, one and the same with that used by the mobile RTX 3050 and RTX 3050 Ti.
The differences in the two versions of the RTX 3050 are expected to be minimal: core counts, clock speeds, and performance should all match, spec-for-spec.
Igor's Lab posits there could be a slight reduction in power consumption for the GA107 model: 115W total board power (TBP) to the standard 130W seen today. This switch-up could have a knock-on effect for overclocking or undervolting—the GA107 chip should take to undervolting a little better as a result of the more efficient design overall.
It appears that Nvidia has been rather smart with the specs this time in order to open the door to this potential switcheroo.
As its foundry partner Samsung produces GA106 chips, it will be amassing stock of those that aren't capable of hitting the required specifications of the RTX 3060. So it makes sense to use these as the basis for an even more budget card. But
Read more on pcgamer.com