Please note that this post is tagged as a rumor.
NVIDIA next-gen GeForce RTX 40 series to use TSMC N5 process instead of Samsung
Industry insider provides confirmation on NVIDIA Lovelace architecture process architecture.
Kopite7kimi is a leaker who correctly predicted almost the whole NVIDIA Ampere lineup many months in advance. Naturally, each word the user shares on Twitter is meticulously analyzed. The same applies to his recent NVIDIA leaks.
The user now reaffirms the rumors from July, when Greymon55 (another leaker although better known for AMD content) revealed that NVIDIA is planning to use TSMC N5 process technology for their next-gen gaming architecture known as Ada Lovelace. Now both leakers now agree that NVIDIA Lovelace is using TSMC rather than Samsung.
of course TSMC N5
— kopite7kimi (@kopite7kimi) August 26, 2021
The same leaker revealed that Lovelace GPUs will actually a bit sooner than previously expected, although no dates are given it is rumored to launch by the end of next year. NVIDIA may want to keep its 2-year cadence of new gaming GPUs, which would be around September 2022.
Ada Lovelace will come out a little bit earlier. Keep patient.
— kopite7kimi (@kopite7kimi) August 26, 2021
Unlike AMD, NVIDIA is not planning to use a multi-chip-module/chiplet design for its upcoming series. Instead, NVIDIA will keep Lovelace monolithic, at least the flagship GPU known as AD102 is. So far there hasn’t been much information on any other Lovelace GPUs such as AD104 or AD106.
The rumored specifications of NVIDIA AD102 GPU equip it with 18432 CUDA cores and performance nearly twice as high as GA102. The AD102 would be a large 600mm² die consuming 400-450W using a smaller and more efficient node.
| Next-gen Flagship GPU Comparison | |||
|---|---|---|---|
| VideoCardz.com | |||
| Fabrication Node | TSMC 5nm | TSMC 5nm + 6nm | |
| Architecture | NVIDIA Ada Lovelace | AMD RDNA3 | |
| GPU Package | Monolithic | Multi-Chip-Module (MCM) | |
| Estimated GPU Size | ~600mm² | ~800mm² (or more) | |
| Graphics Dies | 1 | 2 | |
| GPU Mega Clusters | 12 Graphics Processing Clusters (GPC) | 6 Shader Engines | |
| GPU Super Clusters | 72 Texture Processing Clusters (TPC) | 60 RDNA Workgroups (WGP) | |
| GPU Clusters | 144 Streaming Multiprocessors (SM) | – | |
| GPU FP32 Cores | 18432 CUDA Cores | 15360 Stream Processors | |
| GPU Clock | ~ 2.2 GHz | ? | |
| Memory & Bus | GDDR6X 384-bit | GDDR6 256-bit | |
| Large L3 Cache | ? | 256-512MB Infinity Cache | |
| Estimated Power Consumption | ~ 420-450 watts | ~ 450-480 watts | |
| Release Date | Expected Q4 2022 | Expected 2H 2022 | |
| Rumored Flagship SKU | GeForce RTX 4090 | Radeon RX 7900XT | |
| Rumored Performance | 2x GA102 | 2.5x NAVI 21 | |
| Based on rumors | |||
Source: @kopite7kimi via Wccftech
