Nvidia's RTX 5080 Super Rumored to Feature 24GB GDDR7 Memory and Over 400W Power Consumption

New details have emerged suggesting that Nvidia could be planning to release an RTX 5080 Super with 24GB of GDDR7 memory, along with a Total Graphics Power (TGP) exceeding 400W. The information comes from a well-respected hardware leaker known as Kopite, who has a strong track record with Nvidia's product lines. Pricing and availability details, however, remain undisclosed. GDDR7 memory modules are now available in densities of 16Gb (2GB) and 24Gb (3GB), with even higher capacities expected in the future. By using the denser 24Gb modules, the RTX 5080 Super will achieve a 50% increase in VRAM capacity compared to the standard RTX 5080, despite maintaining the same 256-bit bus width. This upgrade is particularly notable given that the rest of the Blackwell lineup, including the RTX 5080, uses standard 16Gb modules similar to GDDR6X. The RTX 5080 Super is rumored to use the full-fledged GB203 die, identical to the RTX 5080, featuring 10,752 CUDA cores and 84 Streaming Multiprocessors (SMs). The 256-bit interface accommodates eight memory modules, each populated with 24Gb (3GB) chips to reach 24GB of GDDR7 memory, operating at 32 Gbps. This configuration results in a memory bandwidth of 1 TB/s, which is approximately 6.6% faster than the stock RTX 5080's 960 GB/s. For context, here’s a comparison with other Blackwell series GPUs: RTX 5090: Uses the GB202 die with 21,760 CUDA cores and 170/192 SMs. It features a 512-bit bus width and 32GB of GDDR7 memory, delivering a bandwidth of 1,792 GB/s. RTX 5080 Super: Employs the GB203 die with 10,752 CUDA cores and 84 SMs. It has a 256-bit bus width, 24GB of GDDR7 memory, and a bandwidth of 1,024 GB/s. RTX 5080: Uses the same GB203 die as the Super variant, but with 16GB of GDDR7 memory and a bandwidth of 960 GB/s. RTX 5070 Ti: Features the GB203 die with 8,960 CUDA cores and 70/84 SMs. It has a 256-bit bus width, 16GB of GDDR7 memory, and provides 896 GB/s of bandwidth. RTX 5070: Utilizes the GB205 die with 14,448 CUDA cores and 48/50 SMs. It has a 192-bit bus width and 12GB of GDDR7 memory, offering 672 GB/s of bandwidth. RTX 5060 Ti: Equipped with the GB206 die, it has 4,352 CUDA cores and 34/36 SMs. It uses a 128-bit bus width and 8GB of GDDR7 memory, with a bandwidth of 448 GB/s. RTX 5060: Also based on the GB206 die, it has 3,840 CUDA cores and 30/36 SMs. It shares the same 128-bit bus width, 8GB of GDDR7 memory, and a bandwidth of 448 GB/s. It is unlikely that Nvidia will introduce a Super/Ti model for the top-tier RTX 5090, as this could cannibalize sales of its professional GPUs. The most affordable workstation card, the RTX Pro 4000 Blackwell, is expected to retail for around $1,500 and is equipped with 24GB of memory. This suggests that 24GB will be the maximum VRAM capacity for consumer Super models, as higher capacities are primarily targeted at AI/ML developers and enthusiasts. Pricing for the RTX 5080 Super is estimated to fall between $1,000 and $1,500, assuming manufacturers adhere to the official MSRP. A more optimistic scenario would see Nvidia reducing the base MSRP, similar to what happened with the RTX 4080 Super. However, given the current market trends where MSRPs often serve as mere suggestions, this remains a hopeful projection. Nvidia has faced criticism for restricting access to preview drivers during reviews and for limiting pre-release testing to very specific benchmarks, particularly those focused on Multi Frame Generation. This approach has raised concerns among the hardware community and may influence consumer perceptions when the RTX 5080 Super is eventually released. To stay informed about the latest developments in the world of graphics cards, including any updates on the RTX 5080 Super, follow Tom's Hardware on Google News and make sure to click the Follow button.