NVIDIA AD102 GPU has 76.3B transistors
The most transistor-packed consumer GPU yet.
Not everything was revealed by NVIDIA today, but luckily for us, some board partners were quick to share the missing details.
Today, NVIDIA introduced its Ada Lovelace architecture for next-gen gaming, workstation and server GPUs. This architecture is based on custom TSMC 4N process and brings higher raster, raytracing and AI-accelerated computing over Ampere.
The AD102 GPU has 76.3 billion transistors and 608.4 mm². This means that the transistor density of 125.5 million per mm² is 2.78x higher than GA102 “Ampere” GPU manufactured by Samsung using 8N node. Those details were confirmed by PNY for RTX 6000 Ada workstation graphics card.
NVIDIA Ada features Shader Execution Reordering (SER) which supposedly makes raster operations faster, providing up to 25% better gaming performance. Furthermore, Ada features next-generation RT Core (Gen3) and Tensor cores (Gen4). The latter delivers up to 1400 TFLOPS, 4,375 times more than Ampere’s 3rd generation cores.
The AD102 GPU will not be used by RTX 4090 gaming card, RTX 6000 ADA workstation and L40 Server GPU. With 2048 CUDA core remaining budget, it is very likely that NVIDIA will introduce an even faster gaming card later.
|NVIDIA Flagship Gaming GPUs|
|Node||TSMC 4N||SAMSUNG 8N||TSMC 12nm||TSMC 12nm||TSMC 16nm|
|Die Size||608 mm²||628 mm²||754 mm²||815 mm²||471 mm²|
|Trans. Density||125.5M TRAN/mm2||45.1M TRAN/mm2||24.7M TRAN/mm2||25.9M TRAN/mm2||25.5M TRAN/mm2|
|Tensor Cores||576 Gen4||336 Gen3||576 Gen2||640||–|
|RT Cores||144 Gen3||84 Gen2||72 Gen1||–||–|
|Memory Bus||GDDR6X 384-bit||GDDR6X 384-bit||GDDR6 384-bit||HBM2 3072-bit||GDDR6X 384-bit|