NVIDIA AD102 “Ada” GPU for GeForce RTX 40 Series May Not Have PCIe Gen5 Support

Please note that this post is tagged as one rumor.

NVIDIA AD102 without PCIe Gen5 support?

Kopite7Kimi, a trusted niche covering NVIDIA hardware, today confirmed that the next-generation AD102 GPU does not have a PCIe Gen5 interface.

So far, it’s only a rumor, but if this were true, the GeForce RTX 40 series would not be a bottleneck for next-generation desktop systems. The AD102 GPU would simply not support any faster interface. Right now, only Intel Alder Lake supports PCI Express 5.0, but around the same time that NVIDIA is expected to introduce the Ada GPU series, AMD is also preparing its Ryzen 7000 “Raphael” platform, which most likely already has Gen5 support.

The “Ada Lovelace” architecture, set to launch later this year, was rumored to have PCIe Gen5 interface support, primarily because it already has Intel ATX 3.0 / PCIe Gen5 spec 16-pin unlocking connectors to 600W power per cable.

The rumor of lack of Gen5 support is particularly surprising, considering that back in March this year, NVIDIA introduced the Hopper architecture, which was the first GPU architecture to officially support PCIe Gen5. This is how NVIDIA described the new standard for its data center H100 accelerator:

PCIe Gen 5 provides 128 GB / sec. total bandwidth (64 GB / sec in each direction) compared to 64 GB / sec. total bandwidth (32 GB / sec in each direction) in Gen 4 PCIe. PCIe Gen 5 enables the H100 to interface with the highest performing x86 CPUs and SmartNICs or computing devices (DPUs).

In addition, Kopite mentioned that the structure of single-precision (FP32) core configuration may not be as simple as Amperes. It should not be ruled out that the CUDA core number will not change again with the new generation of the GeForce RTX series. An array of different integer and floating FP32 units may increase the number of cores, computing power, or completely introduce a new core type. It is simply too early to say.

NVIDIA usually keeps GPU architecture details under a closed lid, even until the last minute. In fact, when the GA102 GPU was launched, many desktop partners had no idea that the FP32 kernel number needed to be doubled. We still found incorrect key figures in official data sheets and product descriptions for many days after launch.

Next Generation Flagship GPU Comparison (RUMORED)
VideoCardz.com GeForce RTX 4090 – Class Radeon RX 7900 – Class
Fabrication Node TSMC N5 TSMC N5 / N6
Architecture NVIDIA available AMD RDNA3
GPU package Monolithic Multi-Chip Module (MCM)
Estimated GPU size ~ 600 mm² ~ 800 mm²
Graphics die 1 2 GCD + 4 MCD + 1 IOD
GPU Mega Clusters 12 Graphics Processing Clusters (GPC) 2 × 3 Shader motors
GPU Super Clusters 72 Texture Processing Clusters (TPC) 2 × 30 RDNA Workgroups (WGP)
GPU clusters 144 streaming multiprocessors (SM)
FP32 cores 18432 CUDA Cores 15360 Stream processors
GPU ur ~ 2.2 GHz ~ 2.5 GHz
Memory type 24 GB GDDR6X TBC GB GDDR6
Memory & Bus 21 Gbps 384-bit TBC 256-bit Gbps
Cache 96 MB (L2 Cache) 256 or 512 MB Infinity Cache
Interface PCIe Gen4 (?) TBC
Power consumption 600W (TDP) 500W (TBP)
Release date Q3 / Q4 2022 Q3 / Q4 2022

Source: @ kopite7kimi

Leave a Reply

Your email address will not be published.