Rumors have begun circulating about the new Nvidia GPU
Nvidia’s new generation of GPU Ampere will be announced soon. In public is already known by the name RTX 3080 Ti. However, recent rumors say that the GeForce RTX 3080 Ti can be 40% faster than the RTX 2080 Ti.
These leaks have some fantastic specifications. While HPC (High-Performance Computing) is confirmed to use the TMSC node, the consumer GPU will be based on Samsung’s 10nm node.
Specifications leaked shows 5 variants of Ampere in descending order. At the top, we have the GA102 with a 384-bit bus width. This graphics card has 5376 CUDA Cores and 12GB VRAM. Below it is the GA103 with 3840 CUDA Cores, 320-bit bus width and 10GB VRAM. The last three are GA104, 106 and 107, with specifications lower than the first two. Note that GA105 does not yet exist.
All GPUs will use Samsung’s 10nm 8LPP node and support RTX (ray-tracing). The good news is that the weak GPUs will already have this feature.
This makes us calculate that the GeForce RTX 3080 Ti with 5376 CUDA Cores will perform 40% better than the RTX 2080 Ti. This is a good guess if you point out the difference between earlier versions like the GTX 1080 Ti with 3584 CUDA Cores and the 2080 Ti with 4352 CUDA Cores.
The first generation of NVIDIA RTX GPUs used the Turing architecture. Ampere promises better optimization and more ray-tracing support with Nvidia TensorRT. Ampere will also utilize PCIe 4.0, which allows higher bandwidth for the same bandwidth.
The RTX 3080 Ti can be on a whole other level with the RTX 2080 Ti seeing that it has 700 CUDA Cores more. Nvidia may also be pushing the clock speed even further because of the new node.