RTX 50 SUPER Cards Are Sitting at Board Partners, Held Up by GDDR7 Supply

The Cards Are Built. They Just Can’t Ship.

Somewhere in a warehouse right now, RTX 50 SUPER graphics cards exist. They have been manufactured, boxed up, and shipped to NVIDIA’s board partners. And yet, nobody can buy one. According to TechPowerUp, citing VideoCardz, NVIDIA has physically delivered the RTX 50 SUPER silicon to its partners, but Team Green headquarters has put the launch on hold. The reason is straightforward and a little frustrating: there are not enough 3 GB GDDR7 memory modules to go around.

The 3 GB GDDR7 chips are central to what makes the SUPER lineup distinct. Rumor has long held that NVIDIA planned to use these denser modules to meaningfully bump memory capacity across the board. The RTX 5070 SUPER was expected to jump to 18 GB, while both the RTX 5070 Ti SUPER and the RTX 5080 SUPER were slated for 24 GB. Everything else, according to TechPowerUp, stays the same as the standard RTX 50 series. No extra CUDA cores, no architectural overhaul. Just more VRAM, and the kind of headline numbers that tend to move units.

AI Business Is Eating the Memory Supply

The problem is that those same 3 GB GDDR7 modules are in extremely high demand elsewhere. As TechPowerUp reports, NVIDIA is using them in its AI GPU portfolio, specifically the RTX PRO 6000 Blackwell and the Rubin CPX. The availability of these chips is well below current industry demand, and it is not hard to see why NVIDIA would direct limited supply toward products that carry higher margins than consumer gaming cards. The regular GeForce RTX 50 series, which uses the more widely available 2 GB GDDR7 modules, remains on shelves in the meantime.

For gamers who have been waiting on a SUPER refresh to push past the memory limitations of the current lineup, that is cold comfort. The 2 GB module cards are fine. They are not what people were hoping to upgrade to.

Power Numbers Add Up

While the launch itself is stalled, a separate breadcrumb surfaced through an unlikely source. The Seasonic PSU calculator, spotted by users and reported by TechPowerUp, lists power figures for the SUPER cards that help fill in the picture. The RTX 5080 SUPER shows a total graphics power rating of 415 W, which is 55 W above the RTX 5080’s 360 W figure in the same calculator. The RTX 5070 Ti SUPER comes in at 350 W, up 50 W from the 5070 Ti’s 300 W. The RTX 5070 SUPER lands at 275 W, a 25 W increase over the base 5070’s 250 W.

Those bumps are consistent with what you would expect from the denser memory configuration. Running 3 GB modules draws more power than 2 GB modules, and TechPowerUp notes that NVIDIA might also introduce slight frequency improvements on top of that to give the SUPER branding more to point to than just a memory upgrade.

What This Means for Buyers

The realistic takeaway here is that the RTX 50 SUPER lineup is real, further along in production than many expected, and stuck at the final hurdle. NVIDIA controls the memory allocation, and right now the AI side of the business is winning that argument. Until the supply of 3 GB GDDR7 chips catches up with demand, or until NVIDIA decides to shift priorities, the cards will sit with board partners and the launch date will remain unannounced.

For anyone who has been holding off on a GPU purchase specifically to see what the SUPER refresh brings, the news is genuinely mixed. The hardware exists. The specifications look meaningful, particularly the jump to 24 GB on the Ti and 5080 class. The wait, though, is entirely out of the buyer’s hands.