From what we know to date, NVIDIA is waiting for CES 2025, which will take place during the first days of January to officially launch the long-awaited new generation of graphics cards, the RTX 50. But in the meantime, and as is normal at this point, the leaks and rumors are happening one after the other, and on this occasion new data has appeared about the next NVIDIA RTX 5080the second most powerful in the family, but even so everything indicates that it would be a real beast… and not just for gaming.
The NVIDIA GeForce RTX 5090 and RTX 5080 GPUs will be the top-of-the-range graphics cards of the new Blackwell generation, and if it happens as in previous generations, they will be the first to see the light. We already “know” a lot about them through leaks and rumors, but new information ensures that the RTX 5080 will have a faster VRAM solution… in fact, the fastest on the market to date.
More memory and faster
Apparently, the GeForce RTX 5080 that will use the GB203 GPU will come equipped with 16GB GDDR7 memory at an effective speed of 32 Gbps (this is 4 Gbps more than the 28 Gbps that the RTX 5090 will theoretically have), under a 256-bit bus and that, in total, would offer up to 1 TB/s of memory bandwidth .
If we compare it with the current RTX 4080, it has a bandwidth of 736 GB/s since it has GDDR6X VRAM at 23 Gbps. This memory is already crazy, but GDDR7 looks like it’s going to change the rules of the game again with a bandwidth that’s hard to believe for a 256-bit bus only.
It must be taken into account, and it is curious, that apparently only the RTX 5080 will have this type of memory, at least at 32 Gbps; Other models in the RTX 50 family will use GDDR7 as well, but as far as we know so far, none at this speed.
Everything we know about the RTX 5080
According to all the leaks and rumors we have had to date, the NVIDIA GeForce RTX 5080 will be based on the PG144/147-SKU45 PCB and will incorporate the NVIDIA Blackwell GB203-400-A1 chip. This graphics card will use everything the chip provides, that is, it will not be a “capped” model, and that means that it would have 84 SM and a total of 10,752 cores, 51% less than the RTX 5090 will have.
We also know that this graphics card will have a TBP (Total Board Power, that is, a general consumption of the entire device, not just the GPU) of 400 watts, which represents a 25% increase in power wall, although we must Keep in mind that we can find quite a discrepancy between the figures given by the manufacturer and those we will obtain later in the real world.
That said, NVIDIA is expected to offer 16GB and 24GB models of this graphics card. This would be a very nice improvement, especially for gamers and professionals who want to use this extra memory for content creation, intensive AI tasks, or simply to ensure that high-resolution games have plenty of VRAM to load textures and maps.
