Nvidia appears to be working on a graphics card to bridge the performance and price divide between its RTX 5080 and RTX 5090 models. Known as the RTX 5090 SE, this rumored GPU is expected to feature 32GB of GDDR7 memory, aligning it with the high-end RTX 5090’s capacity, and target both gaming and AI workloads.
The RTX 5090 SE reportedly uses the GB202 GPU chip, which also powers the flagship RTX 5090 and professional variants like the RTX Pro 6000. However, it will come with fewer CUDA cores—around 14,080—which places it roughly 31% above the RTX 5080 but still trailing the full RTX 5090 by about a third. This suggests a tailored balance between raw power and efficiency.
While matching the RTX 5090’s memory size at 32GB, the SE model is expected to have a narrower 384-bit memory bus, resulting in slightly reduced memory bandwidth compared to the flagship. Power consumption is reportedly set at approximately 500 watts, lower than the RTX 5090’s 575 watts, though the card may still offer headroom for overclocking or higher power draws, given existing RTX 5090 variants exceed 1,000 watts.
Pricing projections from various sources place the RTX 5090 SE around $1,500, aiming to fill a gap between the original $1,000 launch price of the RTX 5080 and the $2,000 base price of the RTX 5090. However, current market prices have drifted significantly, with RTX 5080 cards commonly selling for between $1,400 and $2,000 and RTX 5090 units often exceeding $4,000.
If launched, the RTX 5090 SE could offer enthusiasts and AI professionals a powerful alternative that balances memory capacity and CUDA core count while avoiding the premium price of Nvidia’s flagship GPU. This move may also allow Nvidia to maintain dominance in the highest-performance segment by providing a more accessible intermediate option.