Look inside Tesla’s onboard Nvidia supercomputer for self-driving

In the past, we have extensively covered the fact that Tesla has been equipping all its vehicles produced since October 2016 with a very powerful computer in order to eventually enable fully self-driving capability.

While Nvidia confirmed that the computer in question is based on its Drive PX2 platform for autonomous driving, they offer several variations of the product and we never knew which one for sure until now.

Nvidia describes the Drive PX 2 as “the world’s first AI supercomputer for self-driving cars”. Its computing power is comparable to about 150 MacBook Pros and the company estimates that one can support a level 4 self-driving system while two would be necessary for a fully self-driving level 5 vehicle, but Tesla is aiming for its software to be efficient enough to run level 5 on one.

But if that doesn’t work out, the company made sure that it is relatively easily accessible so that it can be swapped. CEO Elon Musk’s recent comments make it sounds like Tesla is still not sure if they will need to replace it for level 5.

It is mounted on the passenger side of the dashboard right above the outer panel of the glove box. By removing the panel, it can be accessed and removed by unscrewing a few screws and unplugging the connections.

We previously reported on owners removing those panels to confirm that the supercomputer was there, but Model S owner Kyle Day went a few steps further and dismantled the entire thing to reveal the actual board that Tesla is using.

Nvidia advertises two versions of the Drive PX2: Drive PX2 for Autocruise with a single GPU and single camera/radar input or Drive PX2 for AutoChauffeur which is equipped with 2 GPUs and several more inputs for cameras and radar antennas.

People assumed that they were using the Drive PX2 for AutoChauffeur based on the inputs alone – and Nvidia was even displaying the board at CES earlier this year where company representatives referred to it as the board in Tesla’s vehicles.

But Day’s teardown shows that it is actually a custom board in-between ‘Drive PX2 for Autocruise’ and ‘Drive PX2 for AutoChauffeur’ with a 1 SoC and 1 discrete GPU.

Therefore, it appears to be a custom solution for Tesla and it is not as powerful as Nvidia’s most powerful solution for self-driving, which the chipmaker itself doesn’t believe can enable fully self-driving.

Though again, Tesla kept the door open to upgrade the onboard computer. When first announcing the second generation Autopilot with the new hardware, Musk said that it came down to NVIDIA versus AMD and the former was the better choice at the end.

The company is also reportedly working on its own custom SoC (System on Chip) especially for self-driving cars and looking at Samsung to manufacture it. The chip would likely be designed by famous chip architect Jim Keller who now leads Tesla’s Autopilot hardware team and brought on board several top chip architects at the automaker.

As reported yesterday, Tesla is very much focused on its vision and control software to bring self-driving to market on its current hardware suite, but it will be interesting to see if it will be able to run on the current computer.

Featured Images: all credit to Kyle Day – reprinted with permission and first posted to TMC.