A look at Tesla's Dojo supercomputer shared at Hot Chips 34

By Lennon Cihak
Tesla's Dojo training tile
Tesla's Dojo training tile
Tesla/YouTube

Tesla participated in Hot Chips 34 and shared a ton of mind-boggling information about their Dojo supercomputer and chip architecture.

The details shared about Dojo in Hot Chips 34 by Tesla’s Emil Talpes, who worked at AMD for about 17 years for Opteron processors, are only about the hardware and capabilities of the tiles and Dojo as a whole. The performance will be discussed at Tesla’s AI Day II on September 30th.

The goal with Dojo is, according to Musk, “to be really good at video training. We have probably the fourth or approaching the third most powerful computing center in the world for AI training. Our first goal with Dojo is to make it competitive and be more effective and neural net training than a whole bunch of GPUs.”

Since Tesla needs a lot of computing power to process the video data from the vehicles in its fleet, it has built a proprietary system-on-wafer solution. According to ServeTheHome, “Each D1 die is integrated onto a tile with 25 dies at 15kW. Beyond the 25 D1 dies, there are also 40 smaller I/O dies.”

All of the power and cooling is integrated directly on the Training Tile, which is capable of 10 TB/s on-tile bisection bandwidth and 36 TB/s off-tile aggregate bandwidth. This architecture allows for the tiles to be scaled with 9TB/s links between them. They can also be plugged in and do not require their own server.

“The defining goal of our application is scalability,” Talpes said at the end of the presentation. “We have de-emphasized several mechanisms that you find in typical CPUs, like coherency, virtual memory, and global lookup directories just because these mechanisms do not scale very well when we scale up to a very large system. Instead, we have relied on a very fast and very distributed SRAM storage throughout the mesh. And this is backed by an order of magnitude higher speed of interconnect than what you find in a typical distributed system.”

The inside look into what Tesla’s building behind the scenes continues to prove how and why Tesla is at the forefront in artificial intelligence and neural net training. It gives the Tesla community an added sense of comfort knowing that Tesla will always have scalability and innovative technology at the forefront of everything the automotive company does.

Watch Anastasi In Tech’s Recap of Tesla’s Hot Chip 34 Presentation

Tesla internal email indicates major end-of-quarter push

By Kevin Armstrong
Tesla's Fremont lot is practically full
Tesla's Fremont lot is practically full
Met God in Wilderness/YouTube

Tesla wants everyone on its payroll to pitch in and help with a frantic end-of-quarter push. An internal email leaked to Electrek indicates that employees, even those not tasked with sales and delivery, to assist in getting as many vehicles out the door as possible. The email says there is “a very high volume of vehicles,” and management requests additional support.

This push should come as no surprise after a disappointing second quarter. The company delivered 18 percent fewer vehicles than in the previous reporting period. Tesla delivered 254,695 compared to 310,048 just three months prior. The decline broke a two-year streak in which the company continued to break its delivery record. The drop in numbers was not due to failures at Tesla, but due to COVID-related shutdowns at its Gigafactory in Shanghai, China.

Nevertheless, Tesla is adamant about righting the ship immediately. This quarter, like no other before, has seen significant competition enter the electric vehicle space and challenge the king of the sector. As a result, Tesla would like nothing more than to get back on track with another record-breaking delivery of its highly sought-after vehicles.

According to Electrek, the company-wide email stated: We will be delivering a very high volume of vehicles to eagerly waiting customers during the final days of Q3. To help ensure we can delight as many customers as possible, the delivery team is requesting additional support with key delivery-execution tasks.

For weeks Tesla’s careers page has listed dozens of Seasonal Advisor positions. The postings say these positions are “essential support at the end of the quarter, operating alongside our Sales and Delivery team to execute record-breaking delivery results.” The jobs were expected to last for two to three weeks.

Analysts predict that Tesla will get back on track and deliver more than 310,000 cars this quarter. In fact, one analyst has gone much further than that. Trip Chowdhry with Global Equities Research has publicly stated that he forecasts the carmaker will exceed 500,000 deliveries in the fourth quarter. In addition, he estimates that the time to produce a Tesla has decreased by 10 percent, and he notes other factors like the Semi coming online and increased capacity at factories worldwide.

Tesla is expected to report its quarterly earnings in mid-October.

Tesla is showing suggested destinations in update 2022.28.2

Future Feature
By Lennon Cihak
Suggested destinations in 2022.28.2
Suggested destinations in 2022.28.2
Not A Tesla App

Tesla's rolling out another unique feature with suggested destinations in version 2022.28.2. This comes off the heels of a previous announcement where Tesla was offering alternate routes.

Tesla's new suggested destinations show up under the Recents tab when you tap on the 'Navigate' button.

Tesla will list up to three destinations before displaying your recent destinations below them. Several owners have reported seeing this new feature in Tesla update 2022.28.2, but it's not clear whether it's available to everyone yet.

Tesla's suggestions appear to be a combination of recent destinations and favorites, but it's not clear exactly how Tesla is determining these.

A user by the name of Randall on the Tesla Discord noticed the new feature.

He was offered three suggestions, but after taking a short drive he noticed that the next set of suggestions was completely different.

Tesla may be using AI to determine your suggestions, so things like the number of times frequented, time of day, day of the week, the vehicle's location and more may be taken into account when Tesla is determining your suggestions.

Like everything Tesla, this is the first iteration of this feature and we're likely to see improvements on this feature in the near future.

For example, Randall noticed that although one of his suggestions was one of his favorites, the suggestion just showed the address of the destination and not the name Randall that given the destination.

The placement of suggested destinations could potentially be improved as well, instead of being displayed under Recents. Each of these suggestions likely has a probability score that Tesla determines and Tesla then displays any suggestions that are over a certain threshold.

It could be helpful if Tesla displayed suggestions directly in the middle of the screen when you first entered the car. If Tesla thought you were highly probable of visiting one of these suggestions they could display it more prominently, making it easier for you to enter the destination.

In recent updates, Tesla has focused quite heavily on navigation improvements. Alternate routes is a welcomed feature, as well as a quieter Joe Mode, and a revamped Energy App and time graph for Supercharger popularity. This suite of features is well overdue and much needed.

As the Austin-based automotive company continues to address various bugs and add features, owners are gaining more data for insight into their vehicle so they can better use it.

CEO Elon Musk has stated that any input into the vehicle should be considered (as a possible) error, so by adding yet another feature that the car tries to calculate your destination is a step in the right direction in Musk’s eyes. This is why there is no gear physical shifter in the refreshed Model S and Model X vehicles.

Instead, Tesla replaced them with an “Auto Shift” feature. This function works based on your vehicle’s understanding of which direction it needs to go. For example, if you’re in a grocery store parking lot, the car will see that it’s wedged between a few cars and know that it needs to reverse, or vice-versa. This feature is reportedly coming to all Teslas.

“I think generally, all input is error,” Musk explains at the unveiling of the Model S and Model X refresh. “If you have to do something that the car could’ve done already, that should be taken care of. The software should just do it.”

Upcoming Release

View the release notes for the upcoming version 2022.36.

Tesla News

Confirmed by Elon

Take a look at features that Elon Musk has said will be coming soon.

Subscribe

Subscribe to our weekly newsletter.

Sponsors

Find out how to become a sponsor and have your site listed here.

Although we share official Tesla release notes, we are not affiliated with Tesla Motors. We are Tesla fans and supporters.

Tesla News

Upcoming Release

View the release notes for the upcoming version 2022.36.

Confirmed by Elon

Take a look at features that Elon Musk has said will be coming soon.

Subscribe

Subscribe to our weekly newsletter.