After it was confirmed that NVIDIA is no longer launching the RTX 50 Super series, marking the first time in history that it breaks the «ritual of annual releases», They do not plan to remain with their arms crossed in front of AMD’s RadeonsArc from Intel and the Lisuan.
Perhaps patience has not lasted long enough and soon they will bring to the market two important products during the CTC conference in March 2026 of the company.
The next generation of «Feynman» GPU architecture from Jensen Huang’s company is just around the corner and is going to be key in the performance improvement for graphics cards GeForce RTX 60 o 70.
Do you want to know what it is about? For both large companies and 3D modeling professionals and gamers, they are surely interested in what is going to happen in the computer industry.
NVIDIA already has the powerful Feynman architecture for the upcoming GeForce RTX 60 and 70 series
NVIDIA prepares for CGT 2026 which will take place from March 16 to 19, according to the official website. Why is it so important? Well, it is an event in which the company reveals its new technological advances and products together with the CEO, Jensen Huang.
According to the media Techbang, the aforementioned leading company in GPUs would be planning to officially announce the new Feynman architectureapparently a name inspired by the American theoretical physicist Richard Feynman (1918-1988).
This internal design that defines how the chip works is going to be an important revolution in the next generations of graphics cards, probably the RTX 60 and 70 series or special artificial intelligence accelerators.
Its composition is a giant technological step because it incorporates a TSMC A16 node that is considered the world’s first 1.6 nanometer (nm) processsomething that improves how electricity flows to eliminate interference and increase energy efficiency to more than 20% compared to 2nm.
In this way, this type of processor can process complex calculation systems for artificial intelligence without being an electrical problem due to its rear power supply network (SPR).
Of course, it requires high consumption, approximately 1,000 W to 2,000 W (dual chip). Because of this, there is the concern of increased heat that will likely be mitigated with pure liquid cooling to prevent overheating.
Of course, thought has also been given to the issue of the large investment it requires and the impact on future electrical infrastructure. Manufacturing costs will be managed with Intel’s EMIB-T packaging and without depending on TSMC’s CoWoS capability.
This is in line with speculation that NVIDIA is also taking a focus on training and reasoning tasks for AI models. Likewise, it is still rumored that it could integrate «LPU technology for inference.»
The current GeForce RTX 50 cards have the Blackwell architecture, while its successor would be Rubin with a launch scheduled for this year. Feynman would arrive after Rubin on the market between 2028 and 2029.