Could anyone with good enough technical knowledge provide us with a summary of the relevant data that has been gathered so far from Linux commits, as well as of what it means and what it potentially implies (if this has been done already, could someone then please provide me with a link to the relevant post?) ?
The Big Summary Post I put together should still basically be accurate, though it mixes the Linux reveals with the NVN2 leak.
I personnally hate this kind of thing. It's impossible to pronounce lol Nintendo Twitch, obviously
famiboards.com
The short summary of what's in the Linux data:
- T239 has lost most of the self driving car machinery
- It has an 8 core single cluster CPU
- It has half as much memory bandwidth as Orin
- It has a File Decompression Engine which is "for games" but whose exact function is unclear
My understanding is that a commit to the Linux kernel regarding T239 was first discovered back in september of last year, from which we've learned that Drake has got 8 CPU cores in one cluster.
Since then, some of the more tech savvy users here such as
@oldpuck and
@Thraktor (sorry if I forgot about anyone) have subsequently dug deeper into the matter and discovered some older updates not only to the Linux kernel, but also to a specific Linux distribution called Linux4Tegra, some of which go back to as far as April or May 2022.
One thing I'm specifically wondering about is how long before the final product's commercialisation those kind of Linux updates typically occur.
For instance, I believe the first consumer-oriented, Linux-based device featuring the Tegra X1 was the Nvidia Shield TV, which released on May the 28th, 2015.
Do we know how long before traces of the TX1 SoC started to appear in the Linux kernel and/or a specific Linux distribution?
T210 - the chip ID for TX1 - showed up in July of 2015 in the mainline Linux kernel. That was the same month that Linux 4 Tegra made its first (I believe?) public release. The TX2 data dropped basically the instant the TX2 did. Orin - T234 - showed up in the mainline kernel a few months before Orin was available
Wouldn't that make a good reference point to guess how far we are from the release of a Linux-based product featuring the T239 SoC?
No, but also yes, but also no.
I will explain!
Tegra chips wind up in lots of places, running lots of operating systems. Linux 4 Tegra is a version of Linux that Nvidia provides that runs on those chips for development purposes. Basically it's the OS for Tegra devkits, even if the final product you might be building with tegra runs a different OS.
L4T runs a bunch of Nvidia specific software. But the Linux kernel needs drivers for the chips themselves to even boot up, and standard practice is to push them in the shared Linux code base. That process can take some time, and is subject to approval by people who don't work at Nvidia (namely Linus himself, the original Linux creator).
So the process is that Nvidia develops L4T, releases L4T when they have a Tegra development board, and pushes the Linux drivers upstream into general Linux somewhere in the neighborhood of that release. Sometimes it works out a little before, sometimes it works a little after.
In the case of the TX1, Nvidia released the Shield TV before they released the Jetson TX1 Devkit. Since the Shield TV is a blackbox product for end users, it didn't need a Linux at launch, so there was no L4T release. And the L4T release, when it came, happened about the same time that Nvidia started trying to get its drivers upstream
In the case of Orin, the L4T release was held till the Orin AGX module was out, but apparently, the drivers were in a good position that they started moving them upstream a few months earlier than that.
If Nvidia were mainlining Drake support into Linux, it would tell us that they were likely getting ready to add L4T support for Drake, which would tell us that a Jetson Drake (or whatever) was probably incoming. But that could come
totally independently of the Switch.
Double However: Nvidia actually isn't mainlining the Drake drivers. They've actually gone out of their way to try and hide them. The only reason that T239 wound up in the mainline Linux kernel is
probably because the driver is shared with Orin, and at that point the "official" place for Orin drivers was the mainline linux kernel
Almost all of the information we have about Drake from the Linux kernel is like this - Drake's driver actually isn't in the public repositories but there are "stubs" in the Orin driver, where the Orin driver loads by default, and if it catches running on Drake, it loads up a second, separate module for Drake specifically. Originally Orin and Drake's linux drivers were developed together, and at some point earlier this year they were separated, leaving only some vestigial references.
Triple however: it seems likely from these drops that Drake's design is finalized and possibly even being manufactured. There are several places where Nvidia publishes source code and reference docs, and in general, they get updated the day that Nvidia releases a new product - we saw it with Lovelace. T239 was updated in August, right before the driver got mainlined. So it seems highly likely that Drake is ready to go, and that some product (even if it is just another Shield TV) is imminent - or was planned to be (depending on how you interpret recent news)