Good day,
So i'm in a bit of a bind. I've got a dissertation that is quite reliant on Deep Learning. Because of this i've taken the opportunity to have my own machine for Deep Learning. I've already purchased a second hand server (a HP Z420, 64 gigs of Ram, 1tb ssd, 4tb (2x2) hdd, E5-1650v2 CPU) that has the sufficent specifications for the sort of data structures i'll be dealing with. But I need to select a GPU sufficent for what I will be doing in my dissertation.
I've selected out of a cohort of GPUs the RTX 2060 and newly released RTX 3050. Those two have been chosen as they are the least expensive relative to the rest of the GPUs relevant for Deep Learning while being powerful enough to suit my computational needs. Thing is I can't decide between the two. While the RTX 3050 has the most CUDA cores, it only has 80 tensor cores while the RTX 2060 has just under 2k core cores (while the RTX 3050 has 2.5k-ish cuba cores) but has over 100 Tensor cores. Additionally a base RTX 3050 comes with 8GB while a base RTX 2060 only comes with 6GB, and the 8GB version is marketed at a higher price (450-ish for the 6GB version, 530-ish for the 8GB).
My research upon my issue resulted in little results. So appealing to this forum: Which one should I purchase?
Thanks in advance.