NVIDIA Secures TSMC A16 Node for Next-Generation "Feynman" GPUs | TechPowerUp
Friday, August 14th 2026
NVIDIA Secures TSMC A16 Node for Next-Generation "Feynman" GPUs
by AleksandarK
Yesterday, 04:56
Discuss (10 Comments)
According to supply chain sources and those close to TSMC, NVIDIA is already working on prototyping the replacement for its "Rubin" and "Rubin Ultra" GPU designs with the next-generation "Feynman" GPU microarchitecture. As noted in the latest report from DigiTimes, NVIDIA is opting for TSMC's A16 node, bypassing the N2 family of nodes entirely in favor of the more advanced 1.6 nm A16 design. Mass production is scheduled for the second half of 2028. The "Feynman" designs will rely on the backside power delivery in the A16 node, as well as on packaging technologies like 3D chiplets using SoIC, CoWoS-L 2.5D integration, and next-generation panel-level packaging.
Initially, NVIDIA will use SoIC 3D stacking technology to stack multiple chiplets into a single package, which will shorten the travel time between chips and reduce latency. These 3D chips will then be packaged side-by-side in a CoWoS-L or CoPoS packaging, creating a multi-kilowatt design capable of processing dozens of PetaFLOPS. For memory, NVIDIA will work with its partners to develop customized HBM, likely HBM4E, by the time "Feynman" reaches mass production. This customized memory includes a specialized base die that can contain logic elements like a memory controller or packet processing unit. Think of it as a way to pre-process data before it enters memory, or as a simple speedup, depending on what NVIDIA aims to achieve.<br>The final piece of the puzzle will be co-packaged optics (CPO), where NVIDIA will introduce optical interconnects between GPUs to reduce latency and power requirements imposed by traditional copper connections. The current "Blackwell" NVL72 rack has a total bandwidth of 130 TB/s between GPUs. "Rubin" doubles that to 260 TB/s, while "Rubin Ultra" also doubles the total bandwidth to 520 TB/s. For "Feynman," the only way to increase system performance is to implement CPO and harness the speed of light to deliver over 1,000 TB/s total system bandwidth, breaking into the petabyte range. All of this will be achieved using TSMC's COUPE technology.
Source:<br>DigiTimes
Related News
Tags:
2.5D
2028
3D
A16
Blackwell
chiplets
CoPoS
CoWoS
CoWoS-L
CPO
Feynman
GPU
HBM
HBM4
HBM4E
NVIDIA
NVL72
power delivery
Rubin
Rubin Ultra
SoIC
TSMC
Ultra
Mar 30th 2026 NVIDIA Readies Rubin-based GeForce RTX 60-series with Massive RT Performance Gains (162)
Jan 16th 2026 NVIDIA Reportedly Ends GeForce RTX 5070 Ti Production, RTX 5060 Ti 16 GB Next (210)
Sep 18th 2025 NVIDIA Buys $5B Worth of Intel, RTX iGPUs Coming to x86, Shares up 25% (256)
Sep 10th 2025 NVIDIA GeForce RTX 50-Series SUPER GPUs Could Arrive at CES 2026 (69)
Jul 10th 2026 NVIDIA Readies GeForce RTX 5090 SE Graphics Card (161)
Dec 31st 2025 Leaks Predict $5000 RTX 5090 GPUs in 2026 Thanks to AI Industry Demand (124)
Sep 3rd 2025 NVIDIA Discrete GPU Market Share Dominance Expands to 94%, Notes Report (237)
Sep 29th 2025 March-April Release of NVIDIA GeForce RTX 50-series SUPER Lineup, Possible CES Reveal (73)
May 26th 2026 NVIDIA GeForce Graphics Drivers 610.47 WHQL Drops Control Panel Support (145)
Feb 19th 2026 AMD Ryzen 10000 "Olympic Ridge" to Debut with 6/8/10/12/16/20/24-Core "Zen 6" SKUs (126)
Add your own comment<br>10 Comments on NVIDIA Secures TSMC A16 Node for Next-Generation "Feynman" GPUs
#1
ALL TSMC capacity for chips on the A16 node just gone ... poof.
#2
I think it was one of Feynman's books where I first read the term "Cargo Cult"
#3
john_ALL TSMC capacity for chips on the A16 node just gone ... poof.<br>That's fine, you know that all these things are going to be aimed at the big AI players, which leaves the other lines available for consumer lines... BUT, will Nvidia do that? Nope, they rather sell 1060/3060/5060 and then sell you cloud gaming, and people are dumb enough to do that
#4
john_ALL TSMC capacity for chips on the A16 node just gone ... poof.<br>Intel and AMD have placed orders on N2, same fab. Just not the back side delivery variant. For us it means 60,70 series on N3, different fab. Good riddance if AI leaves the N3 alone so that i can finally upgrade.
#5
N/AIntel and AMD have placed orders on N2, same fab. Just not the back side delivery variant. For us it means 60,70 series on N3, different fab. Good riddance if AI leaves the N3 alone so that i can finally upgrade.<br>Hm, interesting thought. OK let's assume being on the first trailing node is our new normal as gamers, is that even a viable expectation? So we pause 1-2 years on 4nm as node tech progresses and we get nothing from the 3nm mode until AI accelerators move to 1.8-2nm. Can we expect any GPU volume at 3nm for our next-gen parts or is 3nm still good enough to sell out to AI and we're effed instead for 4 years?
#6
And Jacket Man strikes...