The RTX 20, GTX 16 Super series have been laid out (except for the RTX 2080 Ti Super, which has no interest in coming out), and in the face of AMD’s constant fightback, NVIDIA’s next generation of products is not far off, with news that the new architecture core “Ampere” will be seen at the GTC conference in March (Ampere).
Today’s heavy exposure, learned of the GA103, GA104 two new core code, apparently affiliated with the Ampere family’s high-end sequence, is the first time in many years to see the end of the number 3 NVIDIA core code.
GA103:60 set of SM arrays, 3480 stream processors, memory width 320-bit, memory capacity 10/20GB GDDR6.
GA104:48 Set SM array, 3072 stream processors, memory width 256-bit, memory capacity 8/16GB GDDR6.
Two new core will be 7nm process, but the foundry is unknown, and considering that the specifications of the two are relatively close, do not rule out by TSMC, Samsung foundry may be, after all, TSMC 7nm capacity has been very tight, NVIDIA and Samsung relations are also very good.
Unsurprisingly, they will become RTX 3080, RTX 3070, specifications compared to the RTX 2080 Super, RTX 2070 Super has been greatly improved, the number of stream processors and memory capacity greatly increased, 20GB is already a professional card level, the memory space width has also increased.
As to whether the first time will continue to hide some of the strength, follow-up to the full version of Super, still do not confirm, or depends on the process rate and competitive situation.
In addition, the sequence naming of GA103 is the first time to see, coupled with the core size is not up to the level of TU102 RTX 2080 Ti, it will certainly have a GA102 on the upper side, corresponding to THE RTX 3080 Ti, Titan.