Could the RTX 4090 almost double the core count of the RTX 3090?

We independently review everything we recommend. When you buy through our links, we may earn a commission.

According to rumors, one of the biggest leaps in a single generation has occurred.
New leaks from Twitter user La Frite David offer further details regarding the suspected RTX 4090, including how Nvidia’s next-generation Ada (or Lovelace) compares to the RTX 3090, the company’s current most powerful GPU.
According to the released material, the 4090 designation refers to a number of different models. The AD102 is most likely to be the flagship model, with the AD103 being the next most powerful variant. The AD104-106 will most likely be the mid-range choice, while the AD107 will likely be the entry-level option.

The flagship AD102 GPU model seems to have 144 SMs in a single chip, compared to Ampere’s GA102’s 84 SMs, a 71 percent increase and one of the largest in a generation. The AD102 might have up to 18,432 CUDA cores, a whopping 75 percent more than Nvidia’s upcoming RTX 3090 Ti, which has only 10,752 CUDA cores.
The 16-pin power connections, which offer a maximum power output of 600W and considerably improve performance, would almost certainly increase power usage. It might also be the first PCIe 5.0 graphics card, with 4th generation Tensor cores and 3rd generation RT cores, as opposed to Ampere’s 3rd and 2nd generation, respectively.
In terms of memory, it’s quite probable that Nvidia will equip the AD102 GPU with either GDDR6+ or GDDR7, given both are currently on Samsung’s roadmap.
Of course, none of this is confirmed, so take these reports with a grain of salt until Nvidia makes an official announcement.

We will be happy to hear your thoughts

Leave a reply