The difference is pretty small, comparing GDDR6 with upto 20 Gbps (for the RX 7900 XT/XTX) and 21 Gbps for GDDR6X (e.g. the RTX 4090).
Also 20Gbps GDDR6 is relatively recently released compared to the development timelines of 3000 and 4000 series GPUs - but for something like the 4090 it would likely be a bottleneck while it gets more complicated on the 4080, etc. lower down the stack you get a bit of a hit and miss story between bandwidth requirement vs GDDR6 and the cost and power/thermals of getting the required bandwidth.
Last edited: