Tesla Autopilot hardware#Hardware 2
{{Short description|Hardware used in Tesla autonomous driving systems}}
Tesla Autopilot, an advanced driver-assistance system for Tesla vehicles, uses a suite of sensors and an onboard computer. It has undergone several hardware changes and versions since 2014, most notably moving to an all-camera-based system by 2023, in contrast with ADAS from other companies, which include radar and sometimes lidar sensors.
Initially, the ADAS used a combination of cameras capturing the visual spectrum, forward-facing radar, ultrasonic proximity sensors, and a Mobileye EyeQ3 computer as Hardware 1, fitted to Model S vehicles starting in October 2014. After Mobileye ended its partnership with Tesla in 2016, Tesla began shipping cars equipped with an Nvidia Drive PX 2 computer and an increased number of cameras as Hardware 2. In 2019, Tesla shifted to a computer using a custom "FSD Chip" designed by Tesla, branded as Hardware 3. Starting in 2021, Tesla stopped installing the radar sensor in new vehicles, and the ADAS was updated to drop radar support. In 2022, Tesla announced it also would drop support for the ultrasonic sensors, moving the ADAS to an all-visual system. The most recent sensor and computer implementation is Hardware 4, which began shipping in January 2023.
Physical versions
{{See also|Tesla Autopilot#Safety statistics and concerns}}
=Hardware 1{{anchor|HW1|HW1.0}}=
{{multiple image
| align = right
| direction = vertical
| image1 = Tesla Model S SAO 2016 9208.jpg
| caption1 = Tesla Model S with HW1
| image2 = Tesla Model S with Autopilot Hardware 1 3.jpg
| caption2 = Hardware 1 uses a single front camera, mounted in the triangular-shaped cutout at the top center of the windshield.
}}
Vehicles manufactured after late September 2014 are equipped with a single camera mounted at the top of the windshield, forward looking radar{{cite news |last=Isidore |first=Chris |date=July 18, 2016 |title=Elon Musk says Autopilot upgrade could be coming |url=https://money.cnn.com/2016/07/18/technology/elon-musk-tesla-autopilot-upgrade/index.html?iid=Lead |access-date=July 19, 2016 |work=CNN |location=US}}{{cite web |last=Musk |first=Elon |date=July 17, 2016 |title=Twitter |url=https://twitter.com/elonmusk/status/754769913814941696 |access-date=July 19, 2016}} in the lower grille, and 12 ultrasonic acoustic location sensors in the front and rear bumpers that provide a 360-degree view around the car.{{cite news |author=LeBeau, Phil |date=October 15, 2015 |title=Tesla rolls out autopilot technology |url=https://www.cnbc.com/2015/10/14/tesla-rolls-out-autopilot-technology.html |access-date=16 February 2023 |work=CNBC}} The computer is the Mobileye EyeQ3;{{cite web |last=Lambert |first=Fred |date=January 2, 2017 |title=Lucid Motors' autonomous tech in its all-electric sedan will be powered by Tesla's former partner Mobileye |url=https://electrek.co/2017/01/02/lucid-motors-autonomous-tech-all-electric-sedan-mobileye/ |access-date=January 2, 2017 |work=Electrek}} as implemented, this chip is built on a 40 nm process with a TDP of 2.5 W and a clock speed of 500 MHz.{{cite news |url=https://wccftech.com/tesla-autopilot-story-in-depth-technology/ |title=Exclusive: The Tesla AutoPilot – An In-Depth Look At The Technology Behind the Engineering Marvel |first=Usman |last=Pirzada |date=December 3, 2015 |work=wccftech |access-date=17 June 2024}}{{rp|5}} This equipment allows suitably equipped Tesla Model S and Model X vehicles to detect lane markings, obstacles, and other vehicles, enabling advanced driver-assistance functions branded Autosteer (automatic lane-keeping), Auto lane change, Autopark (parallel parking robot), and Side-collision warning. The EyeQ3 used a neural network approach, relying primarily on inputs from the camera to recognize and label objects to determine which potential areas in the camera's field of view are unoccupied and possible for the vehicle to travel through.{{rp|4}}
Auto lane change can be initiated by the driver turning on the lane changing signal when safe (due to the ultrasonic sensors' 16-foot limited range capability), and then the system completes the lane change.{{Cite web |last=Lavrinc |first=Damon |date=October 9, 2014 |title=Tesla Model S Auto Pilot Does Lane Changes For You |url=http://jalopnik.com/tesla-model-s-auto-pilot-does-lane-changes-for-you-1644644652 |access-date=August 23, 2017 |website=Jalopnik}} In 2016 the system did not detect pedestrians or cyclists,{{cite news |last=Abuelsamid |first=Sam |date=July 1, 2016 |title=Tesla Autopilot Fatality Shows Why Lidar And V2V Will Be Necessary For Autonomous Cars |url=https://www.forbes.com/sites/samabuelsamid/2016/07/01/first-tesla-autopilot-fatality-demonstrates-why-lidar-and-v2v-probably-will-be-necessary/ |access-date=July 1, 2016 |work=Forbes}} and while Autopilot detects motorcycles,{{cite web |last=Lambert |first=Fred |date=November 15, 2016 |title=Tesla Autopilot significantly improved pedestrian detection in v8 update tests show, now renders humans |url=https://electrek.co/2016/11/15/tesla-autopilot-pedestrian-detection-v8-renders-humans/ |access-date=December 21, 2016 |publisher=Electrek}} there have been two instances of HW1 cars rear-ending motorcycles.{{cite web |date=October 21, 2016 |title=Motorcycle rear-ending raises questions on Tesla vehicle type approval in Europe |url=http://newatlas.com/tesla-autopilot-fema/46045/ |access-date=December 21, 2016 |publisher=New Atlas}}
Tesla released a new version of Autopilot in September 2016 that changed the object detection algorithm to more fully use the radar sensor; previously, primary obstacle detection responsibilities fell to the cameras and the radar was used in a secondary role to confirm their presence, but was not given the authority to initiate emergency braking alone. After the update, the radar data was given an equal role in object detection and made capable of identifying "dense obstacles," including "other vehicles, moose, or even alien spaceships," according to Musk. He added that Tesla "believes it would have" prevented the fatal May 2016 underride crash in Williston, Florida, in which Autopilot failed to detect a white trailer against the sky.{{cite news |author=Bigelow, Pete |date=September 11, 2016 |title=Tesla Upgrades Role of Radar in Autopilot Technology |url=https://www.caranddriver.com/news/a15345997/tesla-upgrades-role-of-radar-in-autopilot-technology-says-this-version-may-have-prevented-fatal-accident/ |access-date=16 February 2023 |work=Car and Driver}}
Mobileye ended its partnership with Tesla in 2016, stating that Tesla was "going to hurt the interests of [Mobileye] and hurt the interests of an entire industry, if a company of our reputation will continue to be associated with this type of pushing the envelope in terms of safety".{{cite news |author=Bigelow, Pete |date=September 15, 2016 |title=Autopilot Component Supplier: Tesla Was 'Pushing the Envelope in Terms of Safety' |url=https://www.caranddriver.com/news/a15345743/autopilot-component-supplier-tesla-was-pushing-the-envelope-in-terms-of-safety/ |access-date=16 February 2023 |work=Car and Driver}} Tesla responded that Mobileye backed away after learning Tesla was developing its own vision-based sensor system.{{cite news |date=September 16, 2016 |title=Tesla says Mobileye balked after learning carmaker to make own cameras |url=https://www.cnbc.com/2016/09/16/tesla-says-mobileye-balked-after-learning-carmaker-to-make-own-cameras.html |access-date=16 February 2023 |work=CNBC |agency=Reuters}} Speculation immediately following the announcement included partnering with Nvidia and potentially designing its own ADAS computer.{{cite news |url=https://wccftech.com/tesla-autopilot-future/|title=The Future of Tesla Autopilot — What Happens After Mobileye? |first=Usman |last=Pirzada |date=July 29, 2016 |work=wccftech |access-date=17 June 2024}} After Hardware 2 was released, an upgrade from Hardware 1 to Hardware 2 was not offered as it would have required substantial work and cost.{{cite web |last1=Lambert |first1=Fred |date=January 23, 2017 |title='Tesla will not stop innovating' and introduce 'major updates every 12 to 18 months', says Elon Musk |url=https://electrek.co/2017/01/23/tesla-major-updates-every-12-to-18-months/ |access-date=January 23, 2017 |work=Electrek |quote=require "stripping down the entire car and replacing 300+ parts"}}
= {{Anchor|Hardware 2|HW2|HW2.0}}Hardware 2 =
HW2, included in vehicles manufactured after October 2016, includes an Nvidia Drive PX 2{{cite web |last=Shapiro |first=Danny |date=October 20, 2016 |title=Tesla Self-Driving Car Built on NVIDIA DRIVE PX 2 – NVIDIA Blog |url=https://blogs.nvidia.com/blog/2016/10/20/tesla-motors-self-driving/ |url-status=dead |archive-url=https://web.archive.org/web/20161022111919/https://blogs.nvidia.com/blog/2016/10/20/tesla-motors-self-driving/ |archive-date=October 22, 2016 |access-date=October 21, 2016 |work=The Official NVIDIA Blog}} GPU for CUDA based GPGPU computation.{{cite web |last=Lambert |first=Fred |date=October 10, 2016 |title=Tesla is about to increase its lead in semi-autonomous driving w/ 'Tesla Vision': computer vision based on NVIDIA's parallel computing |url=https://electrek.co/2016/10/10/tesla-vision-autopilot-autonomous-driving-nvidia-cuda/ |access-date=October 10, 2016 |work=Electrek}}{{cite news |last1=Geuss |first1=Megan |date=October 20, 2016 |title=Teslas will now be sold with enhanced hardware suite for full autonomy |url=https://arstechnica.com/cars/2016/10/tesla-says-all-its-cars-will-ship-with-hardware-for-level-5-autonomy/ |access-date=October 20, 2016 |publisher=Ars Technica}} Tesla claimed that the hardware was capable of processing 200 frames per second.{{Cite news |last=Wang |first=Brian |date=August 1, 2018 |title=Tesla claims to have 'world's most advanced computer for autonomous driving' with Autopilot 3.0 update coming next year |url=https://electrek.co/2018/08/01/tesla-chip-most-advanced-computer-autonomous-driving-autopilot-hardware-3-update/ |access-date=August 9, 2018 |work=Electrek |language=en-US}} Elon Musk called HW2 "basically a supercomputer in a car", referring to its capacity of up to 12 trillion operations per second.{{cite news |author=Johnson, Davey G. |date=October 19, 2016 |title=Tesla Announces Major Self-Driving Hardware Revision and Future Level 5 Capability |url=https://www.caranddriver.com/news/a15344452/tesla-announces-major-self-driving-hardware-revision-and-future-level-5-capability/ |access-date=16 February 2023 |work=Car and Driver}} The Autopilot computer hardware, housed just above the glovebox, is replaceable to allow for future upgrades.{{cite web |last1=Lambert |first1=Fred |date=October 20, 2016 |title=A look at Tesla's new Autopilot hardware suite: 8 cameras, 1 radar, ultrasonics & new supercomputer |url=https://electrek.co/2016/10/20/tesla-new-autopilot-hardware-suite-camera-nvidia-tesla-vision/ |access-date=October 20, 2016 |work=Electrek |quote=Musk did say that the new vehicles will eventually be able to upgrade the new onboard Autopilot computer since the access has been made relatively easy}}{{cite news |author=Lambert, Fred |date=January 20, 2017 |title=First picture of Tesla's new NVIDIA onboard supercomputer for Autopilot installed in a car |url=https://electrek.co/2017/01/20/first-picture-of-teslas-new-nvidia-onboard-supercomputer-for-autopilot-installed-in-a-car/ |access-date=16 February 2023 |work=Electrek}} Tesla claimed the HW2 suite of sensors and computation provided the necessary equipment to allow FSD at SAE Level 5.{{cite web |title=Autopilot: Full Self-Driving Hardware on All Cars |url=https://www.tesla.com/autopilot/ |access-date=October 21, 2016 |publisher=Tesla Motors}}
File:20220213 Tesla Model S75D 850 9151.jpg
The hardware includes eight cameras covering an aggregate view of 360° around the car and 12 ultrasonic sensors, in addition to forward-facing radar with enhanced processing capabilities. The radar is able to observe beneath and ahead of the vehicle in front of the Tesla;{{Cite web |date=December 28, 2016 |title=Video shows Tesla's new radar detect accident – and avoid it |url=https://www.mercurynews.com/2016/12/28/video-shows-teslas-new-radar-detect-accident-and-avoid-it/ |access-date=June 23, 2020 |website=The Mercury News |language=en-US}} the radar can see vehicles through heavy rain, fog or dust.{{Cite news |last=Gitlin |first=Jonathan M. |date=September 11, 2016 |title=Tesla is all about autopilot and radar in firmware 8 |url=https://arstechnica.com/cars/2016/09/tesla-is-all-about-autopilot-and-radar-in-firmware-8/ |access-date=November 21, 2016 |work=Ars Technica}} The eight cameras are mounted in various locations around the vehicle: three forward-facing, next to the central rearview mirror mounted on the windshield; two front/side cameras, one each mounted in the left and right B-pillars; two rear/side cameras, mounted in the left and right front fender turn-signal repeaters; and one rear camera, above the license plate.
When "Enhanced Autopilot" was enabled in February 2017 by the v8.0 (17.5.36) software update, testing showed the system was limited to using one of the eight onboard cameras—the main forward-facing camera.{{cite news |last=Lambert |first=Fred |date=February 28, 2017 |title=Tesla Autopilot is currently using only 1 out of 8 cameras of the new hardware suite, still very much in 'beta' |url=https://electrek.co/2017/02/28/tesla-autopilot-cameras-autosteer-beta/ |access-date=October 3, 2018 |work=electrek}} The v8.1 software update released a month later enabled a second camera, the narrow-angle forward-facing camera.{{cite news |last=Lambert |first=Fred |date=March 30, 2017 |title=Tesla's Autopilot 2.0 is now using 2 out of 8 cameras with the new update |url=https://electrek.co/2017/03/30/tesla-autopilot-2-0-camera-8-1-update/ |access-date=October 3, 2018 |work=electrek}} With all eight cameras enabled, data extracted from Autopilot in debugging mode showed the cameras provide a black-and-white feed to the computer, possibly to improve image processing speed.{{cite news |author=Lambert, Fred |date=May 16, 2017 |title=Here's what Tesla's Autopilot 2.0 can see with its 8 cameras |url=https://electrek.co/2017/05/16/tesla-autopilot-2-0-can-see/ |access-date=16 February 2023 |work=Electrek}}
The Tesla Model 3, introduced in 2017, and related Model Y, introduced in 2019, are equipped with an additional driver-facing in-cabin camera. This was disabled at launch and was intended to monitor the cabin remotely while the owner was operating the vehicle as an autonomous robotaxi,{{cite news |author=Fingas, Jon |date=April 12, 2020 |title=Elon Musk explains why Tesla's Model 3 has an in-cabin camera |url=https://www.engadget.com/tesla-model-3-cabin-camera-explained-222017515.html |access-date=16 February 2023 |work=engadget}} but was activated in May 2021 to monitor driver attentiveness while using Autopilot in vehicles without radar sensors.{{cite news |author=Kolodny, Lora |date=May 28, 2021 |title=Tesla starts using cabin cameras to make sure drivers are paying attention |url=https://www.cnbc.com/2021/05/28/tesla-starts-using-cabin-cameras-for-driver-monitoring.html |access-date=16 February 2023 |work=CNBC}}
== {{Anchor|Hardware 2.5|HW2.5}}Hardware 2.5 ==
File:Tesla_Autopilot_HW2.5_and_Infotainment_Boards_(50938493268).jpg
In August 2017, Tesla announced that HW2.5 included a secondary processor node to provide more computing power and additional wiring redundancy to slightly improve reliability; it also enabled dashcam and "sentry mode" capabilities.{{Cite web |date=October 5, 2018 |title=Discover Software Version 9.0 |url=https://www.tesla.com/support/software-v9#dashcam |access-date=September 21, 2019 |website=www.tesla.com |language=en |quote="Dashcam is only available for Model S and Model X cars manufactured after August 2017..."}} During this time, the supplier for the system's radar components was changed from Bosch to Continental,{{Cite web |date=November 2019 |title=The Ultimate Guide to Tesla Autopilot |url=https://www.currentautomotive.com/the-ultimate-guide-to-tesla-autopilot/ |website=Current Automotive}} using the ARS4-B unit.{{cite web |author=Jeffs, James |date=January 6, 2022 |title=Tesla Dropping Radar Was a Mistake, Here is Why |url=https://www.idtechex.com/en/research-article/tesla-dropping-radar-was-a-mistake-here-is-why/25619 |access-date=23 March 2023 |website=ID Tech Ex}}
Tesla will upgrade HW2.5 vehicles to HW3 at no extra cost if the owner purchases Full Self-Driving.
= Hardware 3{{anchor|HW3|HW3.0}} =
According to Tesla's director of Artificial Intelligence (AI) Andrej Karpathy, Tesla had, as of Q3 2018, trained large neural networks but they could not be deployed to Tesla vehicles built up to that time due to insufficient computational resources. HW3 was designed to run these neural networks.{{Cite news |last=SeekingAlpha |first=Brian |date=October 25, 2018 |title=Tesla Q3 2018 Conference Call Transcript |url=https://seekingalpha.com/article/4214138-tesla-inc-tsla-ceo-elon-musk-q3-2018-results-earnings-call-transcript?part=single |access-date=February 13, 2021 |work=SeekingAlpha |language=en-US}} Overall, Tesla claims HW3 has 2.5× improved performance over HW2.5, with 1.25× higher power and 0.2× lower cost.{{Cite web |last=Cutress |first=Dr Ian |title=Hot Chips 31 Live Blogs: Tesla Solution for Full Self Driving |url=https://www.anandtech.com/show/14766/hot-chips-31-live-blogs-tesla-solution-for-full-self-driving |access-date=August 21, 2019 |website=www.anandtech.com}}
HW3 is based on a custom Tesla-designed system on a chip called "FSD Chip",{{cite news |date=April 24, 2019 |title=Tesla and Nvidia spar over 'best' FSD chip |url=https://www.eenewseurope.com/en/tesla-and-nvidia-spar-over-best-fsd-chip/ |access-date=15 February 2023 |work=eeNews Europe}} fabricated using a 14 nm process by Samsung.{{Cite web |last=Coldewey |first=Devin |date=April 22, 2019 |title=Tesla vaunts creation of 'the best chip in the world' for self-driving |url=https://techcrunch.com/2019/04/22/tesla-vaunts-creation-of-the-best-chip-in-the-world-for-self-driving/}} Jim Keller and Pete Bannon, among other architects, have led the project since February 2016.{{Cite web |last=Lambert |first=Fred |date=January 4, 2019 |title=Tesla leaks info about new self-driving computer in latest software update |url=https://electrek.co/2019/01/04/tesla-leaks-hardware-3-self-driving-computer/ |archive-url=https://web.archive.org/web/20190104231142/https://electrek.co/2019/01/04/tesla-leaks-hardware-3-self-driving-computer/ |archive-date=January 4, 2019 |access-date=January 4, 2019 |website=Electrek |language=en-US}} FSD Chip features twelve ARM Cortex-A72 CPUs operating at 2.6 GHz, two systolic arrays (not unlike the approach of TPU){{cite journal |last1=Yüzügüler |first1=Ahmet |last2=Sönmez |first2=Canberk |date=2023-03-01 |title=Scale-out systolic arrays |url=https://dl.acm.org/doi/pdf/10.1145/3572917 |journal=ACM Transactions on Architecture and Code Optimization |volume=20 |issue=2 |pages=1–25 |doi=10.1145/3572917|arxiv=2203.11540 }} operating at 2 GHz and a Mali GPU operating at 1 GHz. Tesla claimed that FSD Chip processes images at 2,300 frames per second (fps), which is a 21× improvement over the 110 fps image processing capability of HW2.5.{{Cite web |last=Lambert |first=Fred |date=December 27, 2018 |title=Tesla to start testing new Autopilot Hardware 3 in employee vehicles |url=https://electrek.co/2018/12/27/tesla-autopilot-hardware-3-employee-vehicles/ |access-date=January 4, 2019 |website=Electrek |language=en-US}} The firm described FSD Chip as a "neural network accelerator" custom-designed for Tesla AI processing. Each of the two systolic arrays on a single FSD Chip are capable of 36 trillion operations per second, and there are two FSD Chips for redundancy.{{Cite web |last=Hollister |first=Sean |date=April 22, 2019 |title=Tesla's new self-driving chip is here, and this is your best look yet |url=https://www.theverge.com/2019/4/22/18511594/tesla-new-self-driving-chip-is-here-and-this-is-your-best-look-yet |access-date=August 24, 2020 |website=The Verge |language=en}}
{{multiple image
| direction = vertical
| align = left
| title = HW2/2.5/3 cameras
| image1 = Tesla Autopilot camera 5.jpg
| caption1 = Front triple camera array
| image2 = Tesla Model 3 rear camera.jpg
| caption2 = Model 3 rear camera
| image3 = Tesla Model 3 Side Badge (35857851690).jpg
| caption3 = Model 3 side/rear camera in front fender badge / marker
| image4 = Tesla Autopilot camera.jpg
| caption4 = Side/front (oval cutout on B-pillar) camera
}}
HW3 cars are equipped with eight cameras, in the same locations and covering the same directions as the HW2 and HW2.5 cameras. Each of the eight cameras supplied with HW3 use the same AR0136A image sensor supplied by Onsemi, which has a maximum resolution of 1280×960 (1.2-megapixel) and a 3.75 μm pixel size.{{cite news |author=Yoshida, Junko |date=June 24, 2020 |title=Teardown: Tesla's Hardware Retrofits for Model 3 |url=https://www.eetasia.com/teslas-hardware-retrofits-for-model-3/ |access-date=16 February 2023 |work=EE Times Asia |url-access=subscription |archive-url=https://web.archive.org/web/20200819064648/https://www.eetasia.com/teslas-hardware-retrofits-for-model-3/ |archive-date=August 19, 2020 |url-status=live}} Initial versions of HW3 also included a Continental ARS4-B radar module. The HW3 system board is the same physical size as the HW2.5 board, but carries more components. This makes it possible to upgrade from HW2 / HW2.5 to HW3, which was not possible during the prior generational change from HW1 to HW2/2.5.
The company claimed that HW3 was necessary for FSD, but not for "enhanced Autopilot" functions.{{Cite news |last=Wang |first=Brian |date=August 8, 2018 |title=Tesla's new Autopilot hardware upgrade will be free for owners with the $3,000 to $5,000 'Full Self-Driving' package |url=https://electrek.co/2018/08/08/tesla-autopilot-hardware-upgrade-free-with-full-self-driving-package/#more-72292 |access-date=August 9, 2018 |work=Electrek |language=en-US}} The first availability of HW3 was April 2019.{{Cite web |date=December 31, 2018 |title=Tesla Invites More Employees To Beta Test Hardware 3 Autonomous Driving Upgrade |url=https://cleantechnica.com/2018/12/30/tesla-invites-more-employees-to-beta-test-hardware-3-autonomous-driving-upgrade/ |access-date=January 4, 2019 |website=CleanTechnica |language=en-US}} Customers with HW2 or HW2.5 who purchased the FSD package were promised an upgrade to HW3 without cost;{{Cite web |last=Alvarez |first=Simon |date=October 16, 2018 |title=Tesla owners who ordered Full Self-Driving will get free HW3 upgrade, says Elon Musk |url=https://www.teslarati.com/tesla-full-self-driving-free-hw3-upgrade-elon-musk/ |access-date=February 4, 2019 |website=TESLARATI.com |language=en-US}} however, when FSD Beta was released in 2021, Tesla owners who had subscribed to FSD and owned a vehicle with HW2 or HW2.5 were required to pay US$1500 to upgrade to HW3. Owners who had purchased FSD with a one-time fee as part of their initial vehicle acquisition were eligible for a free upgrade to HW3.{{cite news |url=https://www.autoweek.com/news/green-cars/a37059597/tesla-fsd-subscribers-might-need-dollar1500-in-new-hardware/ |title=Tesla FSD Subscribers Might Need $1500 In New Hardware |first=Jay |last=Ramey |date=July 19, 2021 |work=Autoweek |access-date=17 June 2024}}{{cite web |url=https://www.tesla.com/support/full-self-driving-computer |title=Full Self-Driving Computer Installations |publisher=Tesla Motor Company |quote=If you purchased Full Self-Driving capability and have Autopilot computer 2.0 or 2.5, you are eligible for a complimentary upgrade to FSD computer. Certain early production vehicles may also receive complimentary camera replacements. A complimentary hardware upgrade is not available for Full Self-Driving capability subscriptions; however, you may be eligible to upgrade at your cost. |access-date=17 June 2024}} Within days, Tesla reduced the cost of the upgrade to US$1000.{{cite news |url=https://www.businessinsider.com/tesla-full-self-driving-upgrade-fsd-price-500-off-2021-7 |title=Tesla knocked $500 off a hardware upgrade for its Full Self Driving subscription after customer backlash |first1=Isobel Asher |last1=Hamilton |first2=Tim |last2=Levin |date=July 21, 2021 |work=Business Insider |access-date=17 June 2024}} In 2022, a Tesla owner in the State of Washington won a default judgment against Tesla which ordered the company to upgrade his vehicle to HW3 without cost; in the ruling, the judge cited Tesla's promises that all cars sold since 2016 included the hardware for FSD.{{cite news |url=https://electrek.co/2022/12/12/tesla-ordered-upgrade-self-driving-computer-for-free-false-advertising/ |title=Tesla ordered to upgrade self-driving computer for free due to 'false advertising' |first=Fred |last=Lambert |date=December 12, 2022 |work=electrek |access-date=17 June 2024}}
== {{Anchor|Tesla Vision}}Tesla Vision ==
File:TeslaVision_eight-camera_system.svg
In late May 2021, Elon Musk posted to Twitter that "Pure Vision Autopilot" was starting to be implemented.{{cite tweet |number=1397347380874203136 |user=elonmusk |title=Pure vision Autopilot is now rolling out in North America. There will be an update of this production release in 2 weeks, then FSD beta V9.0 (also pure vision) a week later. FSD subscription will be enabled around the same time. |date=May 25, 2021}} The system, which Tesla brands "Tesla Vision", eliminates the forward-facing radar starting in May 2021 from the Autopilot hardware package on Model 3 and Model Y vehicles built for the North American market.{{Cite web |last=Lambert |first=Fred |date=2021-05-25 |title=Tesla announces transition to 'Tesla Vision' without radar, warns of limitations at first |url=https://electrek.co/2021/05/25/tesla-vision-without-radar-warns-limitations-first/ |access-date=2022-11-01 |website=Electrek |language=en-US}} The Washington Post reported in March 2023 the immediate result was "an uptick in crashes, near misses and other embarrassing mistakes by Tesla vehicles suddenly deprived of a critical sensor."{{cite news |author=Siddiqui, Faiz |date=March 19, 2023 |title=How Elon Musk knocked Tesla's 'Full Self-Driving' off course |url=https://www.washingtonpost.com/technology/2023/03/19/elon-musk-tesla-driving/ |url-access=limited |access-date=23 March 2023 |newspaper=The Washington Post}} For vehicles without the forward radar, temporary limitations were applied to certain features such as Autosteer, and other features (Smart Summon and Emergency Lane Departure Avoidance) were disabled, but Tesla promised to restore the features "in the weeks ahead ... via a series of over-the-air software updates".
In response, Consumer Reports delisted the Model 3 from its Top Picks in 2021, and the Insurance Institute for Highway Safety (IIHS) announced plans to delist the Model 3 as a Top Safety Pick+,{{cite news |last=Barry |first=Keith |date=May 27, 2021 |title=Tesla Model 3 Loses CR Top Pick Status and IIHS Safety Award After Dropping Features |url=https://www.consumerreports.org/car-safety/tesla-model-3-loses-cr-top-pick-status-and-iihs-safety-award-a2421791602/ |access-date=May 28, 2021 |work=Consumer Reports}}{{cite news |last=Krisher |first=Tom |date=May 27, 2021 |title=Safety ratings yanked after Tesla pulls radar from 2 models |url=https://apnews.com/article/lifestyle-technology-business-3254fcec7f9a59b604442b6a73a4708d |access-date=May 28, 2021 |work=AP News}} but after further testing, both organizations restored those designations.{{Cite web |last=Chapman |first=Michelle |date=June 29, 2021 |title=Tesla Model 3 regains top safety pick designations |url=https://www.mercurynews.com/2021/06/29/tesla-model-3-regains-top-safety-pick-designations |access-date=August 26, 2021 |website=The Mercury News |language=en-US}} NHTSA rescinded the agency's check marks for forward collision warning, automatic emergency braking, lane departure warning, and dynamic brake support, applicable to Model 3 and Model Y vehicles built on or after April 27, 2021,{{Cite news |last=Shepardson |first=David |date=2021-05-27 |title=Tesla loses U.S. designation for some advanced safety features |url=https://www.reuters.com/technology/us-agency-briefed-tesla-radar-sensor-removal-2021-05-26/ |access-date=2022-10-30 |work=Reuters |language=en}} but {{As of|2022|October|lc=y}}, those check marks have been restored.{{Cite web |title=2022 TESLA MODEL 3 4 DR AWD {{!}} NHTSA |url=https://www.nhtsa.gov/vehicle/2022/TESLA/MODEL%203/4%20DR/AWD |access-date=2022-10-30 |website=www.nhtsa.gov |language=en}}{{Cite web |title=2022 TESLA MODEL Y 5-SEAT SUV RWD {{!}} NHTSA |url=https://www.nhtsa.gov/vehicle/2022/TESLA/MODEL%20Y%20%205-SEAT/SUV/RWD |access-date=2022-10-30 |website=www.nhtsa.gov |language=en}}
In December 2021, the New York Times reported that Musk had made a unilateral decision to pursue the camera-only approach and had "repeatedly told members of the Autopilot team that humans could drive with only two eyes and that this meant cars should be able to drive with cameras alone." Several autonomous vehicle experts have denounced the analogy.{{cite news |last1=Metz |first1=Cade |last2=Boudette |first2=Neal E. |date=December 6, 2021 |title=Inside Tesla as Elon Musk Pushed an Unflinching Vision for Self-Driving Cars |url=https://www.nytimes.com/2021/12/06/technology/tesla-autopilot-elon-musk.html |access-date=December 7, 2021 |newspaper=The New York Times |via=NYTimes.com |issn=0362-4331}} According to former employees, Tesla engineers attempted to convince Musk that removing the radar could lead to crashes if the cameras became obscured, but Musk "was unconvinced and overruled" their objections. Brad Templeton noted that LIDAR "will never fail to see a train or truck, even if it doesn't know what it is. It knows there is an object in front and the vehicle can stop without knowing more than that." In contrast, Tesla Vision relies on the "Autopilot labeling team",{{cite web |date=February 16, 2023 |title=In Response to False Allegations |url=https://www.tesla.com/blog/in-response-false-allegations |access-date=March 23, 2023 |website=Tesla Motors}} hundreds of Tesla employees that view short video clips recorded by the cameras to label visible signs and objects, which trains the machine vision interpreter.{{cite news |author=Kolodny, Lora |date=June 28, 2022 |title=Tesla is cutting about 200 Autopilot jobs and closing office in San Mateo, California |url=https://www.cnbc.com/2022/06/28/tesla-cutting-200-jobs-closing-autopilot-office-in-san-mateo.html |access-date=6 April 2023 |work=CNBC}}
Tesla announced in October 2022 they would remove ultrasonic sensors by 2023; vehicles without the ultrasonic sensors would be shipped without Autopark, Park Assist, Summon, and Smart Summon features initially.{{cite news |last=Ramey |first=Jay |date=October 6, 2022 |title=Tesla Is Now Dropping Ultrasonic Sensors as Well |url=https://www.autoweek.com/news/technology/a41544080/tesla-drops-ultrasonic-sensors-autopilot/ |access-date=29 November 2022 |work=Autoweek}} Vehicles shipped in 2023 featured Park Assist, however it relied on the cameras to determine the distance from objects. In late 2023, Tesla rolled out High Fidelity Park Assist for AMD-equipped vehicles. Throughout 2024, additional features were rolled out, including Summon (renamed to Dumb Summon) and the revamped Autopark and Smart Summon (named Actually Smart Summon).{{Cite web |last=Staff |first=Not a Tesla App |date=2023-12-16 |title=First Look at Tesla's New High Fidelity Park Assist [Video] |url=https://www.notateslaapp.com/news/1828/first-look-at-teslas-new-high-fidelity-park-assist-video |access-date=2024-10-09 |website=Not a Tesla App |language=en}}{{Cite web |last=Staff |first=Not a Tesla App |date=2024-03-21 |title=Tesla's New 'Tap to Park' Autopark Feature [Updated With Video] |url=https://www.notateslaapp.com/news/1967/tesla-releases-new-tap-to-park-autopark-feature-in-update-2024-2-11 |access-date=2024-10-09 |website=Not a Tesla App |language=en}}
= Hardware 4 =
File:Tesla Autopilot camera 4.jpg wires]]
Samsung manufactures the processor for Hardware 4 (HW4) in Hwasung, South Korea, on a 7 nm process. The custom System on a chip (SoC) is called "FSD Computer 2".{{cite news |author=Hwang, Jeong-Soo |date=September 23, 2021 |title=Samsung to make Tesla's HW 4.0 self-driving auto chip |url=https://www.kedglobal.com/semiconductors/newsView/ked202109230009 |access-date=15 February 2023 |work=The Korea Economic Daily}} According to a teardown of a production HW4 unit in August 2023, the board has 16 GB of RAM and 256 GB of storage, which are two and four times the RAM and storage in HW3, respectively.{{cite news |url=https://www.autoevolution.com/news/hardware-4-teardown-shows-how-tesla-shifted-resources-toward-the-autopilot-computer-219735.html |title=Hardware 4 Teardown Shows How Tesla Shifted Resources Toward the Autopilot Computer |first=Cristian |last=Agatie |date=18 August 2023 |work=Auto Evolution |access-date=15 June 2024}} Musk stated that HW4 computational capabilities are three to eight times more powerful than HW3.
In June 2022, Tesla filed an application with the US Federal Communications Commission (FCC) to use a new radar module in future vehicles.{{cite news |author=Agatie, Cristian |date=9 June 2022 |title=Tesla Backpedals on the Use of Pure Vision in Its Vehicles, Files To Use a New Radar |url=https://www.autoevolution.com/news/tesla-backpedals-on-the-use-of-pure-vision-in-its-vehicles-files-to-use-a-new-radar-190789.html |access-date=23 March 2023 |work=AutoEvolution}} Based on filings with the Chinese government and a prototype Tesla Model 3 spotted in December 2022 while undergoing testing in Santa Cruz, it is believed that HW4 could include radar, high-resolution cameras, and fans and/or heaters fitted to the cameras to prevent fogging.{{cite news |author=Agatie, Cristian |date=15 January 2023 |title=Tesla Prepares To Launch Autopilot HW4 Sensor Suite With Higher Resolution Cameras |url=https://www.autoevolution.com/news/tesla-prepares-to-launch-autopilot-hw4-sensor-suite-with-important-changes-under-the-hood-208434.html |access-date=23 March 2023 |work=AutoEvolution}} While automotive radar modules used in other manufacturer's cars have up to 200 channels (giving them capabilities similar to lidar), an updated filing with the FCC in 2023 indicates the new radar module that might be included with HW4 is not likely to be high-resolution.{{cite news |author=Agatie, Cristian |date=18 February 2023 |title=The New High-Definition Radar in Tesla's HW4 Is Not As Revolutionary as Promised |url=https://www.autoevolution.com/news/the-new-high-definition-radar-in-tesla-s-hw4-is-not-as-revolutionary-as-it-promised-to-be-210526.html |access-date=23 March 2023 |work=AutoEvolution}}
Tesla started shipping cars with HW4 in January 2023, starting with the refreshed Model S and Model Y;{{Cite web |last=Agatie |first=Cristian |date=2024-06-14 |title=Elon Musk Reveals the First Details About Hardware 5 Autopilot Computer and Sensors |url=https://www.autoevolution.com/news/elon-musk-reveals-the-first-details-about-hardware-5-autopilot-computer-and-sensors-235405.html |access-date=2024-06-15 |website=autoevolution |language=en}} however, FSD was not available initially.{{cite news |url=https://www.autoevolution.com/news/shocking-tesla-model-s-x-with-hw4-cannot-handle-full-self-driving-for-now-211404.html |title=Shocking: Tesla Model S/X With HW4 Cannot Handle Full Self-Driving for Now |date=6 March 2023 |first=Cristian |last=Agatie |work=Auto Evolution |access-date=16 June 2024}} It took six months before HW4-based cars ran camera-based software.{{cite news |url=https://cleantechnica.com/2023/08/10/first-days-with-tesla-model-s-plaid-with-hardware-4/ |title=First Days With Tesla Model S Plaid With Hardware 4 |first=Frank O. |last=Clark |date=August 10, 2023 |work=Clean Technica |access-date=15 June 2024}} Despite the increased image sensor resolution{{Cite web |title=HW4 (Tesla new self-driving hardware) {{!}} Blog or Die! |url=https://www.blogordie.com/2023/09/hw4-tesla-new-self-driving-hardware/ |access-date=2025-01-30 |language=en-US}} with HW4-equipped cars, initial software for HW4 ran the FSD software by emulating HW3, including downsizing the camera images.{{Cite web |last=Zlatev |first=Daniel |date=March 30, 2024 |title=Tesla emulates FSD 12 on Hardware 4.0 vehicles as Elon Musk says HW3 cars a priority |url=https://www.notebookcheck.net/Tesla-emulates-FSD-12-on-Hardware-4-0-vehicles-as-Elon-Musk-says-HW3-cars-a-priority.820609.0.html |access-date=June 15, 2024 |website=Notebookcheck}} FSD v13.2.1 is the first software version to use the native resolution of all cameras and as of December 2024 FSD v13 is only available for cars equipped with HW4.{{cite web | url=https://www.notateslaapp.com/software-updates/version/2024.39.15/release-notes | title=2024.39.15 (FSD 13.2.1) Official Tesla Release Notes - Software Updates }}
= Hardware 5 =
Musk announced Hardware 5 (HW5), being named AI5, during the Tesla annual meeting on June 13, 2024.{{Cite web |last=Ashraf |first=Anan |date=June 21, 2024 |title=Elon Musk Outlines Tesla Giga Texas Expansion Plans: New Data Center To Have Over 500MW Capacity In Next 18 Months To Power Own And Nvidia AI Hardware - Tesla (NASDAQ:TSLA) |url=https://www.benzinga.com/news/24/06/39431163/elon-musk-outlines-tesla-giga-texas-expansion-plans-new-data-center-to-have-over-500mw-capacity-in-n |access-date=2024-06-23 |website=Benzinga |language=English}} Musk stated it is scheduled for release in January 2026, and will be ten times more powerful than HW4. Musk also stated that it will use up to 800 watts when processing complex environments, versus up to 100 watts for HW3 and 160 watts for HW4.{{cite web |author=Patel, Dylan |date=27 June 2023 |title=Tesla AI Capacity Expansion – H100, Dojo D1, D2, HW 4.0, X.AI, Cloud Service Provider |url=https://semianalysis.com/2023/06/27/tesla-ai-capacity-expansion-h100/|access-date=26 January 2025 }}
Feature history
Hardware
Notes
{{notelist}}