NVIDIA Next Generation Ada Lovelace GPUs Will Use TSMC's 5nm Process Instead Of Samsung's

1 of the about famous leakers for NVIDIA GPUs has but confirmed a key detail about the upcoming next-generation Ada Lovelace GPUs from the company. The make new architecture will exist built on TSMC's 5nm procedure. This is massive news if true - and permit's talk about that for a 2nd. Kopite has leaked pretty much all the details of the Turing and Ampere architectures and has historically been extremely accurate. That said, every bit ever with leaks and before multiple confirmations, a grain of table salt never injure anyone.

NVIDIA Ada Lovelace GPUs will "come up out a lilliputian bit earlier" according to manufacture insider

Kopite had already previously confirmed that the Ada Lovelace compages would exist built on the 5nm process, but it was assumed that this would be a Samsung node. NVIDIA had made the (correct) decision to transition to Samsung for foundry services because it had (accurately) forecasted the need that would exist hitting TSMC. Having to share resource with Apple tree and AMD was a risk and NVIDIA's decision to become with Samsung appears to have been mostly validated.

of course TSMC N5

— kopite7kimi (@kopite7kimi) Baronial 26, 2021

Withal, if this leak is correct then NVIDIA is going back to TSMC for their 5nm process - and likely means at least one flagship GPU series will be produced at TSMC - which is business that Samsung Foundries simply lost out on. Fifty-fifty if they build the high end chips on TSMC and keep the low end at Samsung (which is where the volume unremarkably lies) it is still interesting because it means that Jensen has lost organized religion in Samsung'due south ability to keep step with the haemorrhage border nodes.

Ada Lovelace will come out a lilliputian bit earlier. Continue patient.

— kopite7kimi (@kopite7kimi) Baronial 26, 2021

Kopite has also stated that Ada Lovelace GPU volition be arriving sooner than expected, although since they have not confirmed the timeline in the beginning place, nosotros still don't have a concrete date. Rumors accept put the inflow of Ada Lovelace GPUs between Q2 2022 and Q4 2022 so if Kopite is expecting NVIDIA to push dorsum the production and commence mass production a bit early, the Q2 2022 timeframe might be the i to continue in mind going forrad.

Recap on NVIDIA ADA GPU architecture

In many ways, Ada Lovelace can exist thought of every bit the world'due south get-go figurer enthusiast. She is the showtime person to have realized that the Analytical Engine proposed by Charles Babbage had applications across pure calculation and also published what is thought to exist the first algorithm (becoming the offset calculator programmer) intended to be carried past such a machine. This was nearly half a century before Alan Turing would terminate their work and invent the general-purpose computer during the globe war.

NVIDIA has been known to base its architectures on prominent physicists, mathematicians, and scientists and Ada Lovelace is no different. Videocardz actually managed to notice a major hint in NVIDIA's own merchandise store that appears to ostend this rumor near Lovelace architecture being the adjacent generation of GPUs from the company. If yous look at the heroes showcased during GTC's 2018 keynote you notice not merely Ada Lovelace but what are potentially all time to come architectural codenames from NVIDIA. Jensen might have sneakily left the unabridged hereafter roadmap (every bit far as codenames go) in the GTC'eighteen keynote.

NVIDIA CUDA GPU (RUMORED) Preliminary:

GPU	TU102	GA102	AD102
Architecture	Turing	Ampere	Ada Lovelace
Process	TSMC 12nm NFF	Samsung 8nm	5nm
Graphics Processing Clusters (GPC)	6	7	12
Texture Processing Clusters (TPC)	36	42	72
Streaming Multiprocessors (SM)	72	84	144
CUDA Cores	4608	10752	18432
Theoretical TFLOPs	sixteen.i	37.half dozen	~xc TFLOPs?
Memory Blazon	GDDR6	GDDR6X	GDDR6X
Memory Bus	384-flake	384-bit	384-fleck
Memory Capacity	11 GB (2080 Ti)	24 GB (3090)	24 GB (4090?)
Flagship SKU	RTX 2080 Ti	RTX 3090	RTX 4090?
TGP	250W	350W	450-650W?
Release	Sep. 2018	Sept. twenty	2022 (TBC)

The NVIDIA AD102 "ADA GPU" appears to have 18432 CUDA Cores based on the preliminary specs (which can change) provided by Kopite. This is about twice the cores present in Ampere which was already a massive step upwards from Turing. The but fashion this is even possible is because NVIDIA is apparently edifice this on the 5nm process which has significant die area and ability reduction. Interestingly, if you assume a clock speed of 1.75 GHz y'all can also go the peak single-precision performance of the ADA 102 GPU: 64 TFLOPs.

According to Kopite the ADA architecture volition characteristic a much bigger L2 Cache (both Turing and Ampere have 6 MB of cache) which means this might be a major architectural revision (as Turing was to Pascal and Pascal was to Fermi/Kepler) instead of just the usual process shrink.