With less than one month before CES 2025, Nvidia has dropped the first teaser for its upcoming RTX 50 family of GPUs - employing the Blackwell architecture. According to rumors, Nvidia CEO Jensen Huang is set to take the stage and drop the curtains on the next-gen RTX 5090 and RTX 5080 GPUs on January 6th at 6:30 PM PT. Also, at the TGA (The Game Awards) 2024. a cinematic trailer for the upcoming Witcher IV was unveiled, and said to be rendered on an unannounced RTX 50 GPU - likely the RTX 5090.
Nvidia has kicked off a new "GeForce Hype Meter" social media campaign, from now until Jensen's keynote next month. Stamped with an eerily familiar "the more you engage, the better it gets" slogan, users are urged to participate by interacting with their posts for special prizes. To warm up gamers for the RTX 50 unveil, Nvidia is bringing back GeForce LAN dubbed "GeForce LAN 50" starting January 4th at 4:30 PM PT - 50 representing the number of hours before the keynote and well, the RTX 50 series.
It all starts now…🟢 CELEBRATE 25 Years of GeForce🟢 PLAY along & join our GeForce LAN 50🟢 WATCH our NVIDIA Keynote at CES 2025 on 1/6/25#GeForceGreats → https://t.co/LGRaII9HUV pic.twitter.com/enMKzdr5OKDecember 13, 2024
CD Projekt also hinted at Nvidia's RTX 50 Blackwell GPUs at The Game Awards last night. The footnote at the start of the trailer for Witcher IV reads, "Cinematic trailer pre-rendered in Unreal Engine 5 on an unannounced Nvidia GeForce RTX GPU", and it doesn't take much to infer that this alludes to the upcoming RTX 5090.
Blackwell should offer several enhancements over Ada Lovelace across gaming, productivity, and AI acceleration. Akin to its server counterpart, we may see native support for FP4 and FP6 data types on the Tensor cores and better ray-tracing capabilities. Nvidia is likely to leverage TSMC's 4NP (5nm-grade) node for Blackwell desktop - which is 30% denser than 4N if we go by server-class Blackwell products (B100/B200).
Going over the specifications, the RTX 5090 will employ the GB202 GPU said to be Nvidia's largest consumer-grade chip since 2018 with 170-enabled SMs (out of 192 in total) or 21,760 CUDA cores. In addition, leaks point toward 32GB of GDDR7 memory and a large 512-bit interface as the RTX 5090 is expected to guzzle almost 600W of power. The GB203-powered RTX 5080 might have a die size less than half that of the RTX 5090 with just 84 SMs or 10,752 CUDA cores, and 16GB of memory though the power rating still stands strong at 400W - just in time to keep you cozy throughout winter.
Rumors allege that the RTX 50 family will debut with the RTX 5090 and RTX 5080 in January. The RTX 5070 and 5070 Ti are expected to launch in February followed by the RTX 5060 series in April. As it stands, we have no information regarding the pricing structure of these GPUs so it is best to wait for the official reveal.
