The current rumors are about the clock speeds of the Geforce RTX 4000 and how AMD with Navi 31 may not see a country against AD102.
After following the rumors about the graphics cards, you already have a good impression of the Geforce RTX 4000, the chefs are now adding one of the most important ingredients that determine the performance: the clock speed. The rumored 100 teraflops already suggests around 2.8 GHz for AD102 and according to @kopite7kimi one should be able to expect even higher clock speeds. For the current state of graphics chips, that would sport a jump of around 1GHz when switching generations.
To what extent this will be combined with the AD102 remains to be seen. However, the high energy consumption that everyone expected so far already gives an idea that a lot is going on despite improvements in architecture and production. The number of FP32 shaders alone has increased from 84 to 144 groups, at least for the top model. Also, about 1 GHz more clock and the “drunk cow” is done, so to speak. A Geforce RTX 4060 and DLSS could find completely new interest groups there.
Geforce RTX 4000: New specifications for AD102, AD103 and AD104 [Gerücht]
Nvidia should particularly benefit from the move to TSMC, where an adapted 5nm process is used. And that’s also one of the reasons why @greymon55 sees AMD a bit behind with the Radeon RX 7000. Especially with the AD102, Nvidia is said to have an edge over Navi 31. According to the latest status, it’s probably true that Nvidia is operating at the higher end of what is currently feasible for consumer graphics cards when it comes to the AD102. Currently, more than 600 watts are hardly marketable for gamers, because that also requires a correspondingly elaborate cooling solution. AMD had already lamentedthat you have to play along because the competitor is doing it too, and instead of the planned 12,288 FP32 shaders, now there is also a model with 16,384 shaders FP32 – presumably composed of two Navi 32.
The reason you see AMD drop in performance is the theory that it still finds early evaluations realistic, but that Nvidia has had more and more pieces of the puzzle driving expectations upwards. First almost twice as many shaders, then conversion to larger caches, now high clock speeds… According to rumors, AMD is within the realm of expectations and Nvidia is above. But since both camps are ultimately in the business of making money, the skins aren’t gone yet. Especially when quantities are sold, AMD is certainly involved and can also perform better than an AD103 or AD104. Saving the flagship with Radeons was an issue from the start and was more focused on the RTX 4080 and below. We’ll find out in the fall.
Introvert. Beer guru. Communicator. Travel fanatic. Web advocate. Certified alcohol geek. Tv buff. Subtly charming internet aficionado.