NVIDIA Rumor & News Roundup: Rumored Specs for RTX 5070, 5070 Ti, and 5080 SUPER Emerge, Transformer Model Uses Less VRAM, RTX 5090 Grows to DD Overseas

The rumor mill is in full motion with details regarding NVIDIA’s eventual GPU refresh lineup, plus DLSS now uses less VRAM, sort of. It’s no secret that NVIDIA normally does a refresh launch of some of its GPUs each generation. These can sometimes arrive in the form of Ti or SUPER branding, and in the case of the GeForce RTX 4070 Ti SUPER, both. That being said, hardware industry information leakers are holding steadfast that NVIDIA will once again repeat its naming nomenclature for its Blackwell-based GeForce RTX 50 series of GPUs. The most notable changes are increased VRAM due to the availability of 3 GB GDDR7 modules and raised power limits; it was not mentioned if GPU core clock frequencies will be increased as well. It’s said that NVIDIA could announce these variants at CES 2026, which would also be par for the course given its semi-annual cadence refresh launch cycles. The latest rumors regarding the variants came via Kopite (1, 2, 3).

GeForce RTX 5080 SUPER

Starting at the top, the GeForce RTX 5080 appears to be getting the least amount of upgrades, with only an increase in memory and power draw. While an increased amount of VRAM has long been expected, something suspected given that the mobile RTX 5090 (aka scaled down desktop RTX 5080) features 24 GB GDDR7 on a 256-bit bus, it has yet to be confirmed. Power draw is reportedly being raised from 360 W to 415 W. All other specs are said to be identical.

GeForce RTX 5070 Ti SUPER

With a name that assuredly doesn’t roll off the tongue for any human, and can make most veteran PC enthusiasts cringe, the GeForce RTX 5070 Ti SUPER is said to also get an increased memory to 24 GB GDDR. This shouldn’t come as a surprise either, since it too uses a scaled-down version of the die used for the GeForce RTX 5080. Power draw will reportedly be increased from 300 W to 350 W. Core counts, like the aforementioned GeForce RTX 5080 SUPER, are said to remain the same as the current model.

GeForce RTX 5070 SUPER

If these rumors pan out to be true, the GeForce RTX 5070 SUPER will be the big winner in NVIDIA’s refresh cycle. If this happens, it will also mirror what occurred with the GeForce RTX 4070 SUPER, which featured similar improvements over its predecessor. VRAM is said to be increased from 12 GB GDDR7 on a 192-bit bus to 18 GB GDDR7 on a 192-bit bus. Core counts will reportedly go from 6144 to 6400 with power draw having only a slight increase from 250 W to 275 W.

DLSS Transformer

NVIDIA has released a new SDK for DLSS, in which it has indicated that the feature can use up to 20% less memory than the previous version. However, this pertains to the software implementation and not necessarily VRAM usage from apps or games. Testing done at 1080p saw VRAM consumption drop to 87.8 MB from 106.9 and reportedly was similar when tested (per NVIDIA via VideoCardz) at 1440p (143.54 MB from 181.11 MB), 4K (307.37 MB from 387.21), and 8K (1,225.17 MB from 1,517.60 MB).

GeForce RTX 5090 DD

Lastly, NVIDIA, in order to meet export requirements, has released another version of the RTX 5090 for China. This time around, it’s doubling down on naming with a new change expanding it to be the GeForce RTX 5090 DD. The only notable change, per VideoCardz, is that memory has been reduced from 32 GB GDDR7 to 24 GB GDDR7 on a 384-bit bus compared to the 512-bit bus found on the non-Chinese version.

Join the discussion in our forums...