According to kopite7kimi, a reputed source on NVIDIA’s upcoming products, the next-gen NVIDIA RTX 40 series GPUs may feature as many as 18,432 FP32 cores and a peak compute rating of more than 66 TFLOPs. That means we’re looking at a 12 GPC (Graphics Processing Cores), nearly twice as much as the fastest Ampere GPU, the RTX 3090.
The GeForce RTX 3080 and 3090 feature 6 and 7 GPCs, respectively which yield 8,704 and 10,496 cores, respectively. Both GPUs are based on the GA102 die which features a total of 10,752 cores, but neither feature the fully enabled die.
Therefore, if the AD102 includes a total of 18,432 cores, we can expect around 16,000 cores on the RTX 4080 and 18,000 on the RTX 4090. Considering that we’re likely to see the AD102 fabbed on Samsung’s 5nm/6nm node, the increase in transistor density should absorb the otherwise expected increase in die size.
The next-gen RTX 40 series GPUs from NVIDIA will be reportedly named after Ada Lovelace, the British mathematician and writer, known for her work on Charles Babbage’s proposed mechanical general-purpose computer, the Analytical Engine.
The first MCM or chiplet based design, Hopper has supposedly been delayed by a generation and will follow Lovelace sometime in early 2025.