The early specifications of NVIDIA’s Ampere refresh, namely the GeForce RTX 3080 Super and the RTX 3080 Ti have surfaced. Unlike the previous Super lineup don’t expect the same levels of gains as the GA102 die is already close to full utilization with the RTX 3080.
|Graphics Card Name||GPU||FP32 Cores||Memory||Memory Bus||Bandwidth|
|NVIDIA GeForce RTX 3080 Super||GA102-xxx||~9,600||12GB GDDR6X||384-bit||936GB/s|
|NVIDIA GeForce RTX 3090||GA102-300||10,496||24GB GDDR6X||384-bit||936GB/s|
|NVIDIA GeForce RTX 3080 Ti||GA102-250||10,496||20GB GDDR6X||320-bit||760GB/s|
|NVIDIA GeForce RTX 3080||GA102-200||8,704||10GB GDDR6X||320-bit||760GB/s|
|NVIDIA GeForce RTX 3070 Ti||GA104-400||7,424||8GB GDDR6X||320-bit||760GB/s|
|NVIDIA GeForce RTX 3070||GA104-300||5,888||8GB GDDR6||256-bit||448GB/s|
|NVIDIA GeForce RTX 3060 Ti||GA104-200||4,864||8GB GDDR6||256-bit/192-bit||448GB/s|
|NVIDIA GeForce RTX 3060||GA106-300||3,840||12GB GDDR6||192-bit|
|NVIDIA GeForce RTX 3050 Ti||GA106||3,584||6GB GDDR6||160-bit|
|NVIDIA GeForce RTX 3050||GA107-300||2,304||4GB/6GB GDDR6||128/160-bit|
While the RTX 3080 and 3090 feature 68 and 82 SMs, respectively, the RTX 3080 Ti is supposed to get the same SM count as the latter but with a reduced 320-bit bus. The RTX 3080 Super, on the other hand, is reportedly getting a wider bus of 384-bit but a lower shader count of around 9,600 FP32 cores or 73-75 SMs.
In terms of memory, the RTX 3080 Ti will get a slightly cut down buffer compared to the 3090 while the RTX 3080 Super will pack 2GB more than the RTX 3080. Both cards are almost certainly getting Micron’s GDDR6X memory, possibly with an even higher operating clock.