While there were some rumors earlier, even since Nvidia teased the Super series, latest specifications come from a rather credible source, Igor Wallossek from Tom's Hardware Germany. While it was pointless to talk about clocks, Igor revealed on his new Igor's Lab Youtube channel that the entire RTX series will get a decent bump in CUDA core number as well as come with faster 16Gbps GDDR6 memory, with RTX 2060 also getting more memory.
According to released information, Nvidia will actually use different GPU variants for the Super series, so the RTX 2080 Super will use TU104-450 GPU and come with 3,072 CUDA cores and 8GB of GDDR6 16Gbps memory. The RTX 2070 Super will use TU104-410 and get additional 256 CUDA cores, for a total of 2,560 but won't get faster memory, sticking to 8GB of 14Gbps GDDR6 memory.
The RTX 2060 Super will get the biggest update as this will be based on a TU106-410 GPU, a cut-down version of the TU106-400 on the standard RTX 2070, leaving it with 2,176 CUDA cores. It also gets additional 2GB of GDDR6 memory, ending up with 8GB of GDDR6 14Gbps memory on a full 256-bit memory interface.
Unfortunately, pricing is still unknown, but the word on the street is that Nvidia might launch the new Super series at the current MSRP of the RTX series while giving the standard RTX series a much-needed price cut. This will put a lot of pressure on AMD's upcoming RTX 5700 series, especially if Nvidia manage to get it all together before AMD's cards become available.