It is said that such a presentation, or at least a preview of the building, will be visible on GTC, where NVIDIA can put its AMD and BIG Navi on again. And it looks like the war on graphics cards will have a new chapter in 2020, especially following the recent leaks that have occurred.
NVIDIA looks like it is currently working on three different chips that have already been completed and will be doing some final refinement, where it follows the standard naming line: GA102, GA103 and GA104.
The NVIDIA RTX 3080 and RTX 3070 will be the first to arrive on the market
For the first time in many years, NVIDIA will once again have a mid-chip chip between standard 4 series and series. GA103 looks like it will also add some graphics card builder grade and almost chip out of the GA102 flawless.
In terms of leaks, NVIDIA will affect its design in its leverage, increasing the number of SM's to the same range compared to Thwala. Therefore, detailed chips have the following characteristics:
- NVIDIA GA103: will come with 60 SM and a complete calculation of 3840 Shaders with an amazing bus for 320 bytes, which can give us a specific configuration of 10 or 20 GB of GDDR6.
- NVIDIA GA104: will be a little brother with a chip that focuses on the middle and high / high range. It will come with 48 SM
According to this, NVIDIA maintains i 64 Shaders per SM, which means following the path marked by Tured. Your lithographic process will be a 7 nm, but the basis it will provide remains a mystery.
The GA102 can come later to compete with advanced AMD models
It looks like NVIDIA won't be offering its top chip from scratch, since it can focus on beating AMD with BIG Navi and keeping it GA102 months later and as new reds. Therefore, we are able to see a RTX 3080 Ti and RTX Titan Ampere At the end of the year
This would prove that Huang has his own GA103 and GA104 may have more chances than AMD might present RDNA 2.0 and as a result the RTX 3080 and RTX 3070 are faster than its competitors.
Also that the new RTX 3080 would have 8 SM less than the current RTX 2080 Ti / RTX Titan with a lower memory bus. The developments in architecture and frequency They can compensate for this data and present it as a higher-value card than Turing-based card, with much lower usage and "lower" price, leaving AMD in a difficult position to compete even at price.