NVIDIA GM20B

ODNX02-A2

TM660M-A2

TM670D-A1

Fritzchens Fritz

Die Shot

Block Diagram

SMM Diagram

NVIDIA's GM20B GPU uses the Maxwell 2.0 architecture and is made using a 20 nm production process at TSMC. With a die size of 118 mm² and a transistor count of 2,000 million it is a small chip. GM20B supports DirectX 12 (Feature Level 12_1). For GPU compute applications, OpenCL version 1.2 and CUDA 5.3 can be used. It features 256 shading units, 16 texture mapping units and 16 ROPs.

Further reading: Tegra X1 Architecture Whitepaper

Graphics Processor

Released: Jan 4th, 2015

GPU Name: GM20B

Codename: NV12B

Architecture: Maxwell 2.0

Foundry: TSMC

Process Size: 20 nm

Transistors: 2,000 million

Density: 16.9M / mm²

Die Size: 118 mm²

Graphics Features

DirectX: 12 (12_1)

OpenGL: 4.6

OpenCL: 1.2

Vulkan: 1.1

CUDA: 5.3

Shader Model: 6.0

PureVideo HD: VP9

VDPAU: Feature Set I

Render Config

Shading Units: 256

TMUs: 16

ROPs: 16

SMM Count: 2

GPCs: 1

L1 Cache: 48 KB per SMM

L2 Cache: 256 KB

Max. TDP: 15 W

SFUs: 128

All Maxwell 2.0 GPUs

NVIDIA GPU Architecture History

2023 Hopper
2022-2024 Ada Lovelace
2020-2024 Ampere
2018-2022 Turing
2017-2020 Volta
2016-2021 Pascal
2014-2019 Maxwell 2.0
2014-2017 Maxwell
2013-2015 Kepler 2.0
2012-2018 Kepler
2010-2016 Fermi 2.0
2010-2013 VLIW Vec4
2010-2016 Fermi
2007-2013 Tesla 2.0
2006-2010 Tesla
2003-2013 Curie
2003-2005 Rankine
2001-2003 Kelvin
1999-2005 Celsius
1998-2000 Fahrenheit

Graphics cards using the NVIDIA GM20B GPU

Name	Chip	Memory	Shaders	TMUs	ROPs	Base Clock	Boost Clock	Memory Clock
NVIDIA Switch GPU 20nm	ODNX02-A2	4 GB	256	16	16	384 MHz	768 MHz	1600 MHz
NVIDIA Jetson TX1	TM670D-A1	4 GB	256	16	16	998 MHz		600 MHz
NVIDIA Jetson Nano	TM660M-A2	4 GB	128	16	16	640 MHz	921 MHz	1600 MHz