Yesterday's leak told us a lot about Nvidia's upcoming GeForce RTX 50 series of laptop graphics cards. First of all, the GeForce RTX 5090 and RTX 5080 will both have 16 GB of GDDR7 VRAM. So it makes sense that the only difference between the two is in the CUDA cores. Moore's Law is Dead has now revealed new information about Nvidia's penultimate Blackwell laptop GPU.
The GeForce RTX 5080 will be equipped with 7,680 CUDA cores. This figure is slightly higher than the 7,424 CUDA cores in the GeForce RTX 4080 laptop. Tom adds that the Blackwell GPU could offer a 40-60% performance improvement over the RTX 4080 (GeForce). Given the marginal increase in CUDA core count and the equally marginal change in manufacturing node from TSMC 4N to TSMC 4NP this figure seems a bit optimistic.
Additionally, the GeForce RTX 5080 will launch with a maximum TGP of 175 Watts across the entire display – the same as its last generation counterpart. This is unlikely to change anytime soon due to thermal constraints. While Blackwell's new architectural improvements and GDDR7 memory will result in a sizable performance gain, it might not be as high as shown above.
Interestingly, this information goes against Tom's previous leak which indicated that the portable version of the GeForce RTX 5080 would be equipped with 8,192 CUDA cores. This makes sense for a Ti-branded SKU, but there's no evidence of an RTX 5080 Ti so far. It also can't be the RTX 5090 (GeForce), because the difference in CUDA cores between the two is too minimal. This may be the full, uncut die that Nvidia plans to shrink for stable operation.
Tech
