[go: up one dir, main page]
Scribd
Upload
21 views2 pages

Pytorch (Linux Group) Sram Storage

The document discusses and compares Tesla's Dojo supercomputer and Huawei's Harmony OS for training self-driving vehicles using artificial intelligence. It provides a table explaining the technical capabilities of each system, such as the chips used, software, memory storage, and design. Dojo uses Tesla-designed D1 chips in a tile-packed array, while Harmony OS uses a Huawei 1C300B chip in a four-layer structure. Both systems are designed to train machine learning models from sensor data to enable autonomous driving functionality.

Uploaded by

pamin08
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
21 views2 pages

Pytorch (Linux Group) Sram Storage

The document discusses and compares Tesla's Dojo supercomputer and Huawei's Harmony OS for training self-driving vehicles using artificial intelligence. It provides a table explaining the technical capabilities of each system, such as the chips used, software, memory storage, and design. Dojo uses Tesla-designed D1 chips in a tile-packed array, while Harmony OS uses a Huawei 1C300B chip in a four-layer structure. Both systems are designed to train machine learning models from sensor data to enable autonomous driving functionality.

Uploaded by

pamin08
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

Huawei may or may not want to compete in Type II, but the company seemed to be positioning

themselves for Type III in the direction of integrating AI with work/life implications in terms of lifestyle
changes. In that sense perhaps, Tesla also will be doing its own ecosystem positioning by bundling its
newly acquired social media platform twitter, now rebranded as “X”, though it’s use in China will be
limited, again, due to geo-political considerations. In USA, Driverless taxi company Waymo appeared to
be on the road for taxi services; but apparently experiencing set back in San Francisco. Tesla on the
other hand is furiously working on exploring the Full Self Driving using Big Data Analytics and Artificial
Intelligence (AI). Tesla is collecting on road data using sensors and cameras on their vehicles; that data is
then analyzed and utilized in identifying dangerous hazards on the road like other vehicles, persons, and
objects. This real time data collected through ultrasonic sensors like Camera, GPS, Radar, and Lidar is
then trained and analyzed using machine learning (ML). ML is an AI type where a car will learn and adapt
time to time with the data collection and reacts based on historical occurrences. It also enhances user
experience by learning driver’s habits and preferences and personalizes’ car’s settings and features.
Henceforth, the use of ML, big data and AI, Tesla cars are not only safe and productive but also an astute
vehicle that can be driven in any driving condition.
Dojo is Tesla’s own customized supercomputer developed from scratch for training and machine
learning the data gathered from it’s EV Cars. Dojo has Tesla inbuilt chips and the whole infrastructure is
designed by the company itself. This will improve the capability of Tesla to train its neural network using
the data collected from sensors of the car which is a sensitive towards vision tech and self-driving.
Unlike Nvidia’s GPU based supercomputer by Tesla, Dojo replaces six GPU box with a single tile making it
small and still less costly than GPUs. Tesla has its own powering and cooling system for Dojo Exapod
(Combined cabinet of system trays) as they once blew their local power grid to test Dojo cabinet.
Huawei is not left behind in terms of integrating AI for self-driving cars. The company started Intelligent
Automotive solution in 2019 with a goal that 4 in 10 cars will be autonomous by 2030. It is probably
entering Type III space leading to fundamental changes in how we utilize time daily. Huawei has their
own system, Harmony OS to compete with Tesla’s Dojo. Huawei started developing their own system
after Washington ban on Huawei to work with Google back in 2016. (Cite from 4)

Here is a table explaining technical capabilities of Huawei’s Harmony and Tesla’s Dojo:
Tesla-Dojo Huawei-Harmony
Chip D1 1C300B
Software PyTorch (Linux group) Unix like (Modified AOSP)
Memory SRAM [static random-access memory] storage RAM storage
storage
Design Tile packed in 5X5 array (25 D1 chips in a package) 4 layers (Kernel, system service,
framework and application)
Usage Specifically, to train Tesla car using Machine Learning Smartphones, Tablets, TVs and
Smart watches
References:
1. https://electrek.co/2022/10/01/tesla-dojo-supercomputer-tripped-power-grid/

2. https://medium.com/dare-to-be-better/how-tesla-is-using-ai-and-big-data-analytics-in-their-self-
driving-cars-7072e410c1b8#:~:text=Tesla's%20Autopilot%20system%20uses%20a,algorithms%20to
%20make%20driving%20decisions

3. https://www.cnbc.com/2023/11/29/huawei-pushing-its-tech-into-evs-with-cars-automaker-
partnerships.html

4. https://technode.com/2022/03/07/insights-we-tested-huawei-harmonyos-on-a-new-chinese-electric-
vehicle/
5. https://en.wikipedia.org/wiki/Tesla_Dojo#Software
6. https://en.wikipedia.org/wiki/HarmonyOS

Reference Terminologies:

1. D1 chip: Each node (computing core) of the D1 processing chip is a general purpose 64-bit CPU with a
superscalar core. It supports internal instruction-level parallelism, and includes simultaneous
multithreading (SMT). It doesn't support virtual memory and uses limited memory protection
mechanisms.
2. Android Open Source Project is AOSP
3. 1C300B Chip: 1C300 is a cost-effective single-chip system based on the GS232 processor core of
Loongson. It has a built-in floating-point processing unit, supports multiple types of memory, supports
large-capacity MLC flash memory chips, and can be used in industrial control, Internet of Things and other
fields.

You might also like