Yesterday's leak gave us several insights into Nvidia's upcoming GeForce RTX 50 series laptop graphics cards. To begin with, both the GeForce RTX 5090 and RTX 5080 are expected to feature 16 GB of GDDR7 VRAM. Hence, it appears that the only distinction between the two models will lie in their CUDA cores. Recently, Moore's Law is Dead has disclosed additional details about Nvidia's near-final Blackwell laptop GPU.
CUDA Cores and Performance
The GeForce RTX 5080 is set to come equipped with 7,680 CUDA cores, which is a slight increase compared to the 7,424 CUDA cores in the GeForce RTX 4080 laptop version. According to Tom, the Blackwell GPU could potentially provide a performance boost of 40-60% over the GeForce RTX 4080. However, considering the modest rise in CUDA core numbers and a tiny shift in manufacturing from TSMC 4N to TSMC 4NP, this projection might be overly optimistic.
Thermal Constraints and TGP
Moreover, the GeForce RTX 5080 will debut with a maximum TGP of 175 Watts, which is identical to its predecessor. This figure is unlikely to change in the near future due to thermal limitations. While the architectural enhancements introduced by Blackwell and the GDDR7 memory should lead to a noticeable performance increase, it may not reach the levels suggested earlier.
Conflicting Information
Interestingly, this new information contradicts Tom's previous leak, which indicated that the GeForce RTX 5080 laptop version would feature 8,192 CUDA cores. That might be reasonable for a Ti-branded model, but there’s currently no proof of an RTX 5080 Ti existing. It also can’t be the GeForce RTX 5090, as the CUDA core difference is too small. Perhaps it refers to the full, unaltered die that Nvidia intends to downsize for stable performance.
