Nvidia RTX 2080 and RTX 2080 Ti Review

GPU Core

RTX 2080Ti

nVidia say that the TU102 GPU that runs the RTX 2080Ti is their largest GPU in a non-industry application, and judging from the block diagram below that certainly is the case. There are just under eight hundred more CUDA cores on the latest Turing GPU when compared to its GTX 1080Ti predecessor. Obviously these kind of images make for nice wallpaper, but they are a bit beyond our sphere of interest. It’s all very well seeing how something has been designed, but once you’ve seen one Graphics Processing Cluster you’ve kinda seen them all and we wouldn’t begin to pretend we have anything to contribute, given our syllabus didn’t include “how best to squeeze billions of transistors onto a piece of silicon 101”. Marvel at it from an aesthetic viewpoint, and let’s move on.

  

RTX 2080

The basis of the TU104 that forms the heart of the RTX 2080 is almost identical to that of the TU102 we saw above. As is always the case with lower – although that’s extremely relative – performing cards it’s a slightly cut down version of its big brother. The data path and general order of the GPU is extremely similar, and only in the number of Graphics Processing Clusters and the amount of NVLinks available can you spot the difference.