NVIDIA Ada GPUs Smaller, More Performance: AD106 and AD107


Images and specifications for NVIDIA’s upcoming AD106 and AD107 graphics processors have surfaced. New Ada-architecture GPUs will power the low-end laptops we expect to be released soon.

MLID While some images were published by TechPowerUp provided detailed specifications, including die sizes and transistor counts. Based on the Ada Lovelace architecture, GPUs are smaller in size than their Ampere counterparts. As you know, the previous generation used Samsung’s 8LPP (8nm class technology derived from a 10nm class technology) technique. NVIDIA now manufactures its new graphics processors with TSMC’s 4N (4nm class) fabrication technology.

AD102 AD103 AD104 AD106 AD107
Mold Size
Transistor ? ?
SM
CUDA Core
Memory Bus Width
L2 Cache
PCIe

4.0 x8

We can accept the decrease in GPU size as normal. However, the new GPUs offer much higher performances with their small size. Small chip sizes will also reduce costs. On the other hand, chip designers who want to achieve the needed performance targets should have had a harder time squeezing all the necessary interfaces (like memory and display outputs) into a single chip.


Ada GPUs Smaller, More Performance - AD106 and AD107

NVIDIA introduced its GeForce RTX 4000 series notebook graphics processors earlier this year. Laptops powered by the entry-level AD107 and mid-range AD106 GPUs will start at $999 and $1,500 respectively.

NVIDIA GeForce RTX 40 Island Laptop GPU Series

GPU GeForce RTX 4090 GeForce RTX 4080 GeForce RTX 4070 GeForce RTX 4060 GeForce RTX 4050
Production
technology
TSMC 4N TSMC 4N TSMC 4N TSMC 4N TSMC 4N
Architectural Ada Lovelace Ada Lovelace Ada Lovelace Ada Lovelace Ada Lovelace
GPU AD103 AD104 AD106 AD107 AD107
GPU SKU GN21-X11 GN21-X9 GN21-X6 GN21-X4 GN21-X2
CUDA Cores 9728 7424 4608 3072 2560
RT Cores 3rd Generation 3rd Generation 3rd Generation 3rd Generation 3rd Generation
Tensor Cores 4th Generation 4th Generation 4th Generation 4th Generation 4th Generation
Memory Capacity 16 GB 12GB 8GB 8GB 6GB
Data Bus 256-bit 192-bit 128-bit 128-bit 96-bit
Frequency Rate 1455 – 2040MHz 1350 – 2280 MHz 1230 – 2175 MHz 1470 – 2370 MHz 1605 – 2370 MHz
TBP 80 – 150W (+25W DB) 60 – 150W (+25W DB) 35 – 115W (+25W DB) 35 – 115W (+25W DB) 35 – 115W (+25W DB)
Microsoft DirectX 12 Ultimate
Supported Supported Supported Supported Supported
NVIDIA DLSS DLSS 3 (Super Resolution + Frame Generation) DLSS 3 (Super Resolution + Frame Generation) DLSS 3 (Super Resolution + Frame Generation) DLSS 3 (Super Resolution + Frame Generation) DLSS 3 (Super Resolution + Frame Generation)

NVIDIA Refuses To Disclose The Delayed RTX 3090 Ti

Fatih, who has been intertwined with technology and games from a young age, is happy to conduct research and convey his experiences to people.


Related News

PC death: Chip maker Intel ‘plans to lay off thousands of employees’ as rising popularity of smartphones and tablets dampens demand for traditional computers

Chipmaker Intel is reportedly planning major layoffs, which could number into the thousands, in the face of a slowdown in the personal computer market.

Countries with the Best Internet Service Announced

BroadbandChoices, the latest Internet Accessibility Index in the report published his research results based on metrics such as download and upload speed,

Get the Accessories That Power Your Gadgets with the Best Prime Days Deals on Anker Products

We love Anker's products. Not only does the company offer incredible accessories to complement our daily drivers like our smartphones and laptops, but they

Converting a Twitter thread to a PDF is that simple

Did you find a thread on Twitter interesting and want to have it saved in a PDF document? You don't need to copy it manually, because exists a tool that

Thermal cameras and AI can be used to crack passwords, new study warns

A password might not be enough to protect a device from hackers. A new study has shown how criminals can use thermal cameras to retrieve passwords typed in by

Tensor G2 confirmed as 5nm processor, not 4nm as expected

As per the new information, the Tensor G2 is actually built on a 5nm manufacturing process and not a 4nm one as was expected before the launch.