Drive Pegasus is a four-chip solution with two SoCs and two GPUs that will provide a 200x performance of Drive Parker solution. Parker started the whole automotive self-driving movement for Nvidia, moving the company from infotainment only provider to a company that thrives to solve the self-driving problem all the way to totally autonomous level 5. Pegasus was announced in October 2017 by Jensen in Munich.
Drive Pegasus is a four-chip solution with two discrete GPUs and two mobile Xavier SoCs. The new automotive platform from Nvidia is expected to have 200 times performance of Parker. This four-chip solution has 16 core custom ARM 64 CPU codenamed Caramel and two post-Volta discrete GPUs, most likely the ones that people call Turing. The overall performance of Drive Pegasus lands at 320 INT8 TOPS integer operation that sit well with neural networks and image recognition. This is something that many call AI.
Orin is a two-chip solution that should perform like Drive Pegasus a solution with two GPUs and two SoCs. We can only speculate that Orin comes with a brand new, much more powerful GPU and a brand new automotive SoC.
Nvidia managed to do something good with the Tegra that failed to succeed in the mobile world, putting it in cars instead. Nvidia always had a problem keeping the power at acceptable levels for the computing mobile industry but it remains to be seen if the recent Uber and Tesla self-driving fatal accident will question the ability of these low power computers to completely take over self-driving and what people call Level 5.
We can expect more details about Orin by October 9 to 11 2018, when Nvidia plans its European GTC in Munich that primarily focuses on the automotive industry.