Inno3D Lets Slip NVIDIA's GeForce RTX 5090 GPU Launch Timing

Inno3D GeForce RTX graphics card.
At this point, it's not much of a secret that NVIDIA will introduce its GeForce RTX 50 series at the Consumer Electronics Show (CES) in January. After all, NVIDIA co-counder and CEO Jensen Huang is scheduled to deliver a keynote at the event, which is a telltale sign, and there have been numerous leaks and rumors pointing to a CES launch. Adding to the pile, one of NVIDIA's GPU partners confirmed as much in a recent livestream that's since been hidden from the public on YouTube.

Hosted by Pichau, a retailer in Brazil, the 11-hour livestream brought on several guests, one of which was a representative from Inno3D Brazil. According to a person who watched the broadcast, viewers were pinging the hosts with questions, one of which asked when NVIDIA will launch its highly-anticipated GeForce RTX 5090 based on the company's Blackwell architecture.

Surprisingly, the Inno3D rep indicated it would be available "in just three weeks," according to what the viewer told Videocardz. The viewer went on to say that the revelation "caused a big reaction in the chat and among the live presenters," which led to the representative nervously correcting herself by saying, "Officially, the 5090 will only be announced at CES 2025 in January."

That's about as close to a confirmation as you can get without it actually being official. Hopefully, the rep doesn't find herself in too much hot water, as the 5090's launch is the worst kept secret in tech right now.

What's more interesting is the reveal that the 5090 will be out in the wild in three weeks, according to what the viewer witnessed, which falls around mid-December, well ahead of CES. That's not likely a launch time frame, but when NVIDIA's partners will have custom cards ready. It could also be when reviewers will have their hands on NVIDIA's next-gen flagship. Whatever the case, a three-week time frame indicates that the GeForce RTX 5090 is on track to launch as scheduled, without any last-minute hiccups.

It will be interesting to see how the specs and pricing shake out. Based on previous rumors, we're looking at a monster card with 21,760 CUDA cores and 32GB of GDDR7 memory tied to a 512-bit bus for around 1.52TB/s of memory bandwidth.