RTX 4070 Super DLSS 5 support and VRAM evidence

Current DLSS 5 status

RTX 4070 Super DLSS 5 support is unconfirmed as of 2026-10-10. NVIDIA has published no model-specific statement covering Ada Lovelace, so this record stays unconfirmed rather than supported; the announced scope names RTX 50-series Blackwell parts. Card record: 12 GB of VRAM, 7,168 CUDA cores, a 192-bit memory bus, 220 W board power, launched 17 January 2024. Unconfirmed is not a rejection: it means waiting for NVIDIA's own model list, because in-game availability also depends on drivers and per-title implementations.

Unconfirmed by NVIDIA

Model note：DLSS 5 unconfirmed

Evidence source：https://www.nvidia.com/en-us/geforce/news/dlss-5-3d-guided-neural-rendering/

Independent evidence page; this is not an NVIDIA compatibility guarantee.

Model record

What the source does and does not prove

The linked source does not resolve this model either way, and that is exactly what the record says. “Unconfirmed” is a prompt to wait for an official statement, not a soft rejection.

Hardware specification of RTX 4070 Super

These values come from the published GeForce RTX 40 Series specification table, which aggregates NVIDIA’s own product pages and architecture whitepapers. They describe the card itself; the DLSS 5 status above rests on a different source.

Launch date, die code, core count, bus width and board power are hardware facts. This page publishes no benchmark score and no ranking between models — more CUDA cores is not on its own a statement about DLSS 5 performance.

Vendor cards and laptop versions of RTX 4070 Super can differ in clocks, memory capacity and power limit, so match the exact variant before you buy.

Hardware specification source：https://en.wikipedia.org/wiki/GeForce_RTX_40_series — “GeForce RTX 40 Series” specification table on Wikipedia, which cites NVIDIA’s product pages and architecture whitepapers. Retrieved 2026-10-10.

Local DLSS5 Studio on RTX 4070 Super

This card carries 12 GB: above the 8 GB minimum, 4 GB short of the 16 GB recommendation for 1080p and 4K video.

We publish no measured run for RTX 4070 Super; our only measurement so far is the RTX 4090 (720p → 2560×1440 H.264 in 87.4 s (2.43 s/frame)). We do not extrapolate it to other cards, and this page does not claim one.

Feature scope for GeForce RTX 40 Series

These rows come from the same data set as the model table. A “yes” describes the published generation-level capability, not a guarantee that a given game enables it on this card.

Nearest models in the same series

• RTX 4070 Ti · 12 GB · 7,680 CUDA cores
• RTX 4070 · 12 GB · 5,888 CUDA cores
• RTX 4070 Ti Super · 16 GB · 8,448 CUDA cores
• RTX 4080 · 16 GB · 9,728 CUDA cores

The closest record in this series is RTX 4070 Ti (12 GB). Against RTX 4070 Super, RTX 4070 Ti launched 12 months earlier, with 512 more CUDA cores and 65 W higher board power. Both share the same DLSS 5 status in this data set, and we publish no performance benchmark and no ranking between them.

Before you buy or upgrade

1. Confirm the exact RTX 4070 Super variant and its VRAM capacity — vendor and laptop versions differ.
2. Open the linked source and check whether your target game and driver publish a compatible path.
3. Treat “unconfirmed” as a prompt to wait for evidence; never infer support from a similar model.

GPU detail FAQ