Here is another new update on what’s likely to be NVIDIA’s next flagship GPU, the RTX 4090. And according to Moore’s Law Is Dead YouTube channel, the RTX 4090 GPU with the AD102 chip may be the first consumer GPU to employ GDDR7 memory, which should enhance both performance and energy efficiency.
Specifications for Nvidia GeForce RTX 4090 (AD102)
The RTX 4090 GPU will utilize GDDR7 video memory, according to the source. The GPU is believed to contain 18432 CUDA cores, which is the most CUDA cores we’ve seen on a graphics card. Furthermore, the GPU is reported to provide an 80 to 100 percent rasterization boost over the RTX 3090.
Even in terms of ray-tracing speed, the RTX 4090 should be at least twice as fast as the RTX 3090. Finally, the flagship RTX 4090 graphics card is rumored to have an on-air TDP of roughly 450-600W, making it a power-hungry GPU. The same AD102 processor is also believed to be utilized on the RTX 4080, which will feature somewhat fewer CUDA cores.
Nvidia GeForce RTX 4070, RTX 4060 Ti (AD103)
The RTX 4070 and RTX 4060 Ti are rumored to employ the AD103 processor and contain 10752 CUDA cores. It’s said to have 256-bit GDDR memory. However, there is no word on whether it will use GDDR7 or the current generation GDDR6x.
In terms of performance, the RTX 4070 is expected to be 10 to 30% quicker than the current RTX 3090, which is a significant improvement given the RTX 4070 will be a mid-tier GPU.
Finally, ray-tracing performance is reported to be at least twice as excellent as that of the most powerful Ampere graphics card.
Nvidia GeForce RTX 4060 (AD104)
The NVIDIA GeForce RTX 4060 will have 7680 CUDA cores and 192-bit GDDR video memory. In terms of performance, the RTX 4060 is touted to be as powerful as the RTX 3090 Ti, NVIDIA’s most powerful graphics card at the time, and the same holds true for ray-tracing capability.
Nvidia GeForce RTX 4050 Ti (AD106)
The NVIDIA RTX 4050 Ti is expected to employ the AD106 chip and have 4608 CUDA cores. This GPU is touted to equal the performance of the RTX 2080 Ti and RTX 2080 Super and to have 128-bit GDDR video memory. Again, no word on whether this GPU will employ the high-end GDDR7 video memory.
Nvidia GeForce RTX 4050 or Below (AD107)
The AD107 is supposed to be a chip for the entry-level RTX 4000 series of GPUs and will most likely be found in the RTX 4050. This GPU will have 3072 CUDA cores and is expected to perform similarly to the TX 3060. Even in terms of affordability, the RTX 4050 maybe NVIDIA’s most cheap next-generation GPU.