The current generation of graphics cards in the form of AMD’s Radeon RX 6000 and Nvidia’s RTX 3000 is hardly available anymore, when the first rumors about the next-generation specifications are making the rounds on the internet.
Specifically, rough key data of a GPU based on Nvidia’s Ada Lovelace architecture emerged. The basic design should be much more luxurious than that of the current generation and thus also offer more space for the arithmetic units.
There is Lovelace
Nvidia traditionally names its architectures after famous mathematicians. Ampere is named after André-Marie Ampère, Turing after Alan Mathison Turing. Ada Lovelace was also a British mathematician who some historians also refer to as the first real programmer. The Ada programming language is also named after her.
Hardware guides at GameStar Plus: Waiting for Nvidia’s graphics cards, probably called RTX 4000, is taking too long and you prefer to plan with the current generation instead? At GameStar-Plus you can find out which hardware you need:
Nvidia Ada Lovelace with up to 18,432 cores?
The latest information comes from the leaker kopite7kimi (known in hardware circles)@ kopite7kimi), which, for example, predicted the specifications of Nvidia’s current amp architecture very accurately in mid-2019.
Link to Twitter content
The colleagues from 3DCenter took the data and derived more detailed specifications from it. The information 12*6 is therefore as GPC * TPC to understand, i.e. raster engines (GPC) times texture processing clusters (TPC). The GA102 chip, which is built into the RTX 3090 and the RTX 3080, is based on one 7*6-Structure.
This is how the AD102 chip would look compared to its predecessors:
Lovelace (RTX 4000)*
Ampere (RTX 3000)
Turing (GTX 1600/RTX 2000)
Pascal (GTX 1000)
* Information not confirmed
If the specifications are reasonably correct, we see a massive increase in the number of cores and the theoretical computing power with Nvidia Ada Lovelace. The latter, in turn, is based to a not inconsiderable degree on the clock rate, for which 3DCenter assumes a very conservative 1.8 GHz.
Criticism in the comments
In the comments, however, criticism of the high core number is already spreading. Twitter user Barry (@ConsumerBro) writes about:
“Another doubling of the cores !? Thought they would fix Ampere’s scaling problem with Lovelace. That alone will increase performance by 50% +. Half of the Ampere cores aren’t used for gaming!”
User WildCracks (@wild_cracks) contradicts:
“This is more of a tweaking thing for the game developers. They could fix it by adding dedicated INT cores again, but I think Lovelace is more of a refinement of Amps. Hopper will likely be the architecture overhaul.”
What about Nvidia Hopper?
The first rumors about Hopper architecture arose over a year ago. This was to make use of the so-called MCM (multi-chip module) design, also known as chiplet design, for the first time.
So we may see Nvidia Hopper a little later and get an intermediate step with Lovelace. It was similar with Ampere and Turing.
Enjoy rumors with caution: Nonetheless, all of this is mere speculation for the time being. For a long time, experts assumed that the chips for the RTX 3000 were mainly manufactured by Samsung using a 7 nanometer process, in the end it was an 8 nanometer process from Samsung. In addition, hardly anyone had a quasi-doubling of the CUDA cores on their screen.
In the GameStar-Talk you can find out how strong the AMD Radeon RX 6800 XT is compared to the Nvidia Geforce RTX 3080:
AMD versus Nvidia – That’s how strong the Radeon RX 6800 XT is compared to the RTX 3080