The NVIDIA GEForce RTX 50 SUPER lineup of graphics cards are set for a Holiday 2025 release, based on a new rumor.
The chip maker’s next big refresh for its Blackwell-based graphics cards is reportedly arriving sooner than expected. The NVIDIA GeForce RTX 50 SUPER series, which includes the RTX 5070 SUPER, RTX 5070 Ti SUPER, and RTX 5080 SUPER, is now rumored to launch during holiday 2025, with availability expected in Q4.
The major highlight of the SUPER refresh is the massive 50% increase in VRAM capacity across all models. According to recent leaks, the RTX 5070 SUPER will feature 18GB of ultra-fast GDDR7 memory, while the RTX 5070 Ti SUPER and RTX 5080 SUPER are expected to pack 24GB of GDDR7 VRAM each. This bump in memory addresses one of the most pressing needs for 4K gaming, AI workloads, and content creation, where larger memory pools significantly improve performance and future-proofing.
Alongside the increased memory, these cards will come with higher boost clock speeds compared to their non-SUPER counterparts. This performance boost does come at a cost, however. Potential buyers should expect a higher power draw, as early estimates suggest TGP increases of 10–17%, which aligns with the upgraded specifications. The CUDA core count remains unchanged for most models, except for the RTX 5070 SUPER, which is rumored to get a minor increase in core count.
If the rumored timeline holds, this would be the fastest SUPER refresh in NVIDIA’s history, arriving less than a year after the initial RTX 50 series launch. Pricing details currently remain under wraps, and we’ll likely learn more in this regard as the official announcement approaches.
The GeForce RTX 50 series is NVIDIA’s latest line of consumer graphics cards, following the GeForce 40 series. Officially revealed during CES 2025, the series launched on January 30, 2025, with the release of the RTX 5080 and RTX 5090. Built on the new Blackwell architecture, these GPUs incorporate fourth-generation RT Cores for advanced real-time ray tracing and fifth-generation Tensor Cores optimized for deep learning tasks.
