From WikiChip
Drive Xavier - Nvidia
< nvidia
Revision as of 01:19, 21 August 2018 by David (talk | contribs) (typo hours -> years)

Edit Values
Drive Xavier
General Info
DesignerNvidia
ManufacturerTSMC
MarketArtificial Intelligence
IntroductionJanuary 8, 2018 (announced)
General Specs
FamilyDrive
Microarchitecture
ISAARMv8 (ARM)
Process12 nm
Transistors9,000,000,000
TechnologyCMOS
Die350 mm²
Word Size64 bit
Cores8
Threads8
Electrical
TDP30 W
TDP (Typical)20 W

Drive Xavier is a 64-bit ARM high-performance autonomous machine neural processor designed by Nvidia and introduced in 2018. The Drive Xavier is incorporated into Nvidia's Drive Pegasus autonomous computer.

Overview

The Drive Xavier is an autonomous machine system on chip designed by Nvidia and introduced at CES 2018. Silicon came back in the last week of December 2017 with sampling started in the first quarter of 2018. NVIDIA plans on mass production by the end of the year. NVIDIA reported that the product is a result of $2 billion R&D and 8,000 engineering years. The chip is said to have full redundancy and diversity in its functional blocks. That is, the SoC can continue to operate properly even after a fault is detected.

Architecture

CPU

Main article: Carmel core

The chip features eight Carmel core, Nvidia's own custom 64-bit ARM cores. Those cores have full ECC and parity as well as dual-execution (unknown if lockstep or something a bit different) allowing all code to execute twice for redundancy reasons.

GPU

Main article: Volta

The chip incorporates a Volta GPU with 512 CUDA Cores capable of operating in 64-bit and 32-bit floating point as well as 8-bit integer. This allows the various deep learning artificial neural networks types to run efficiently in the format most suitable for them. This translates to 1.3 CUDA TOPS (32-bit FP) and another 20 Tensor Core TOPS (16-bit FP).

Accelerators

The Drive Xavier incorporates a Programmable Vision Accelerator (PVA) for processing computer vision. It is capable of 1.6 TOPS and the ability to do stereo disparity (e.g., processing parallax between two camera to obtain useful information such as depth), optical flow (e.g., direction and speed of vectors), and image processing. Additionally, since the chip is expected to be connected to a network of camera (e.g., side, front, inside), the chip is capable of doing real time encoding for all camera in high dynamic range.

Video Processing
Encode Decode
1.2 GPIX/s 1.8 GPIX/s

The chip has an ISP with native full-range HDR support and tile rendering capable of processing at 1.5 GPIX/s.

Deep Learning Accelerator

The chip incorporates a deep learning accelerator (DLA) that implements a number of specific set of deep learning functions common to many applications. This allows them to read the highest possible energy efficiency for those operations. The DLA has a peak performance of 5 TOPS for 16-bit integers or 10 TOPS for 8-bit integer.

Memory controller

[Edit/Modify Memory Info]

ram icons.svg
Integrated Memory Controller
Max TypeLPDDR4-2133
Supports ECCYes
Channels8
Width32 bit
Max Bandwidth127.1 GiB/s
130,150.4 MiB/s
136.473 GB/s
136,472.586 MB/s
0.124 TiB/s
0.136 TB/s

I/O

  • 16 CSI channels
    • 109 Gbps total bandwidth
  • 1x gE interface
  • 1x 10gE interface

Die

  • 9,000,000,000 transistors
  • 350 mm² die size
  • TSMC's 12FFN

nvidia xavier die shot.png


nvidia xavier die shot (annotated).png

Documents

Facts about "Tegra Xavier - Nvidia"
core count8 +
designerNvidia +
die area350 mm² (0.543 in², 3.5 cm², 350,000,000 µm²) +
familyDrive +
first announcedJanuary 8, 2018 +
full page namenvidia/tegra/xavier +
has ecc memory supporttrue +
instance ofmicroprocessor +
isaARMv8 +
isa familyARM +
ldate3000 +
manufacturerTSMC +
market segmentArtificial Intelligence +
max memory bandwidth127.1 GiB/s (130,150.4 MiB/s, 136.473 GB/s, 136,472.586 MB/s, 0.124 TiB/s, 0.136 TB/s) +
max memory channels8 +
nameDrive Xavier +
process12 nm (0.012 μm, 1.2e-5 mm) +
supported memory typeLPDDR4-2133 +
tdp30 W (30,000 mW, 0.0402 hp, 0.03 kW) +
tdp (typical)20 W (20,000 mW, 0.0268 hp, 0.02 kW) +
technologyCMOS +
thread count8 +
transistor count9,000,000,000 +
word size64 bit (8 octets, 16 nibbles) +