NVidia’s next-gen Ampere line of graphics cards will be announced soon. The public is already familiar with the name RTX 3080 Ti. However, recent news dropped revealing that the GeForce RTX 3080 Ti could be 40% faster than the predecessor RTX 2080 Ti.
These leaks contain some impressive specs for the public to feast their eyes upon. While the HPC (High-Performance Computing) has been confirmed to utilize the TSMC’s (Taiwan Semiconductor Manufacturing Company) node, the consumer-grade GPUs will be based on Samsung’s 10nm node.
The leaked specification slide showed 5 variants of the new Ampere (full-fat) lineup in descending order. On the top, we have the GA102 with a Bus Width of 384-bit. This graphics card packs 5376 Cuda Cores and a total memory of 12gb. Right under this falls the GA103 which has a lower 3840 Cude Core count. This has a bus-width of 320-bit and has a total 10gb ram.
Then the last 3 are the GA104, 106 and 107 each having lower specs in comparison with the first two. Notice that a G105 is not in existence.
All of the GPUs above will use Samsung’s 10nm (8LPP) process node and will support RTX (ray-tracing). It’s good news that even the lower-end GPUs in the next lineup will support ray tracing.
Moving on forward, this leaves us with assumptions that the GeForce RTX 3080 Ti packing 3840 CUDA cores could perform 40% better than the RTX 2080 Ti. This can be a secure assumption if we highlight the difference between previous versions of the lineup like the GeForce GTX 1080 Ti (3584 Cuda cores) and 1080 Ti (4352 Cuda cores). With that said, the GA102 will most likely be a successor to the TITAN RTX with it’s whopping 5376 Cuda cores.
The first generation of the RTX Nvidia cards used the Turing architecture. Ampere promises to further optimize and offer more ray-tracing hardware support using NVidia TensorRT. The Ampere lineup will also take advantage of PCIe Gen 4.0 which enables higher bandwidth speed.
The RTX 3080 Ti will be a huge leap in comparison to its older brother RTX 2080 ti by having approximately 700+ Cuda cores. NVidia may further push the clock speeds thanks to the new advanced process node in their possession now.
The RTX 3080 Ti will feature a 320-bit bus-width which enables it to support 10gb to 20gb of GDDR6 memory. Although 20gb would be a sweet spot in terms of winning the audience over, doing so may raise higher costs in terms of production as GDDR6 shortage rumors lurk because of increased demands in the console market.
The following news does hype up the Ampere line in a fair manner. NVidia never fails to offer amazing performance and reliability over the years compared to its rival AMD (Advanced Micro Devices). However, we should keep in mind that these speculations are based on rumors only. For now, we can only wait for an official reveal for the next lineup of NVidia graphics products.