The RTX 4090 could be significantly more powerful than the RTX 3090

The RTX 4090 could be significantly more powerful than the RTX 3090

Over the past few days, various previously secret information regarding NVIDIA's current and future plans has leaked. Yesterday, we talked about the leak of the DLSS 2.2 source code and some references to the code names of the next consumer and data center video cards, while today we bring you other very interesting data relating to the alleged specifications of the products based on the Ada Lovelace architecture, which will make up the RTX 4000 series.

J'ai fait un petit graphique

- La Frite David (@ davideneco25320) March 1, 2022

In fact, the leaker @ davideneco25320 published a message on Twitter in which the names of the different dies belonging to the RTX 40 family were reported accompanied by the respective number of SM (Streaming Multiprocessor) , giving the possibility to compare them with the current RTX 30 range, based on Ampere. It starts from the top model, distinguished by the AD102 die, which will probably equip a GeForce RTX 4090 or Titan, which will be equipped with 144 SM, a decisive leap forward compared to the "only" 84 of GA102. This value, if confirmed, should provide the new flagship with considerable additional computational power, even if other elements, such as frequencies, as well as additional performance benefits coming from the Ada architecture itself, will have to be taken into consideration. AD102 is followed by AD103 (for a potential RTX 4080) with 84 SM; AD104 with 60 SM; AD106 with 36 SM and finally AD107 with 24 SM.

if (jQuery ("# ​​crm_srl-th_hardware_d_mh2_1"). is (": visible")) {console.log ("Edinet ADV adding zone: tag crm_srl-th_hardware_d_mh2_1 slot id: th_hardware_d_mh2"); } Photo Credit: NVIDIA Currently, it's still a bit early to talk about NVIDIA's upcoming gaming / consumer video cards. Most likely, the RTX 4000 line will be the first to adopt the PCI Express 5.0 interface, as well as next-generation Tensor and RT cores. It will also be interesting to see if the Californian company will make major changes to the fundamental blocks of GPUs, as happened with Ampere compared to Turing. Finally, we could move on to the use of GDDR6 + or GDDR7 memories, which will offer substantial improvements over the current GDDR6X.

Obviously, for the moment these data should be taken as mere rumors, pending a future official presentation directly from NVIDIA.

Powered by Blogger.