If we take into account the indications and signs leading towards a 20 series rather than 11 series Graphics cards, a ray-traced RTX name rather than the tradition GTX specified by the teaser released by Nvidia itself. The new graphics card will release (or at least announced) during the Gaming GeForce event on August 20. The time for the speculations is over its time for the facts that is why we try to answer some of the questions revolving around the RTX 2080 the next flagship Graphics card from Nvidia. These include how much performance boost will we get from the RTX 2080, how much will it cost? Let’s dive right into it!
It has been two years since Nvidia released new architecture for its Gaming oriented Graphics hardware. The architecture was called Pascal; it had 3x performance boost over Maxwell architecture, and due to best compatibility, alongside high-end Graphics cards like GTX 1080/1070 we also had Pascal architecture in entry-level Graphics cards such as GTX 1050s and GT 1030, it’s in most of the best graphics card of 2018 to buy now. Now the next gen architecture will be the Turing architecture which was officially unleashed by Nvidia’s CEO Jensen Huang during SIGGRAPH conference, and Nvidia claims it to have not only the generational advantages over Pascal architecture but also the revolutionary architecture in a decade.
Nvidia has been tightly lipped about their hardware, and most of the earlier rumors were all fake, especially the leaked benchmarks were wrong even the naming was wrong, who would have thought that Nvidia could change the traditional GTX branding.
Nvidia also recently trademarked both GeForce RTX and Quadro RTX brandings, while there are rumors that the GTX counterparts will remain simultaneously in the market, it’ll be a surprise if Nvidia releases a Graphics card with GTX branding down the road. The next-gen Graphics cards will start with the RTX 2080, and the cut down versions will be released later.
Coming back to the Turing architecture, and this is where Nvidia made sure to surprise the lot and refrain from the rumor mill. The architecture as reported earlier is more related to Volta architecture than the Pascal which it would be replacing, due to the addition of tensor cores in the design which were introduced in Volta.
The main features of Volta architecture were deep learning and AI and Nvidia was reluctant to build these features into the GeForce lineup at first that is why they said Volta is not designed for gaming purposes.
However, the times have changed now, and Nvidia is ready to try out its AI rendering into its gaming lineup especially after revealing its new Anti-aliasing algorithm called deep learning Anti-aliasing which would imply the use of deep learning abilities of tensor cores. Which also means that the rumors regarding tensor cores that Nvidia will refrain from the Tensor cores in gaming series this time too.
Regarding manufacturing process, the initial Turing GPU is built by the TSMC’s 12nm process, which is not a huge bump over the 14nm process there are no significant improvements on the die density, fabrication processes, component arrangement and power dissipation rather minor which may have a meaningful impact on the daily usage of the device. The starting devices will be built by TSMC’s process but there is a possibility that the later devices will be built by a better manufacturing process or the same, but fabrication process will be most likely held at both Samsung and TSMC as they did with the later Pascal devices such as GTX 1050s and GT 1030.
The surprising revelation this time around is the chip size; the first Turing GPU has a whopping 754mm2 which is almost equal to the die size of the chip used in Volta GV 100 GPU and also the number of transistors housed on the device are 18.6 billion in total.
For the sake of comparison, the biggest Pascal GPU had the 571mm2 size and 11.8 billion transistors, which shows the yields of the better manufacturing process. It also means that if the RTX 2080 will be using this GPU, then it would be the highest end consumer grade Graphics card. Or if Nvidia has plans for the RTX 2080Ti or Titan V than the smaller die would also be enough for high-end gaming Graphics cards.
Lastly, what would be the cost of the Graphics card? Things are not black and white here. We do know that Nvidia has done its homework to charge the extra premium out of the consumers. There are some whispers that the total cost of the RTX 2080 will be around 800 to 1000 dollars only if they decide to go with the smaller die. If they go all out with the bigger die, it would be no surprise to see the price of the 24GB GDDR6 RTX 2080 Graphics card to be in 1400 to 1500 dollars range considering the cost of the components. Not much time is left we’ll find out the exact details of the Graphics cards on 20th August 2018 at Cologne.