Tesla DOJO Exa-Scale Lossy AI Network using the Tesla Transport Protocol over Ethernet TTPoE
Tesla DOJO Exa-Scale Lossy AI Network using the Tesla Transport Protocol over Ethernet TTPoE
Tesla-Answer is a dedicated community forum for Tesla owners and enthusiasts. Our platform connects experienced Tesla drivers who share practical knowledge, troubleshooting tips, and maintenance insights to help you maintain and upgrade your Tesla.
Tesla DOJO Exa-Scale Lossy AI Network using the Tesla Transport Protocol over Ethernet TTPoE
Say what now?!
Tesla DOJO Exa-Scale Lossy AI Network using the Tesla Transport Protocol over Ethernet TTPoE
Thanks, much appreciated! 🙏
If you didn't get that: Tesla DOJO Exa-Scale Lossy AI Network using the Tesla Transport Protocol over Ethernet TTPoE
Well the future is about SDV. We already know that CAN bus cannot support SDV, and neither does the TCP.
The field is white open for a new standard, expect Toyota and VW/Rivian coming up with alternatives
The only words I understood here are “network” and “transport protocol over Ethernet”. WTH
This can save Tesla millions of $ by not using high end interconnect solutions. Think of how Google in early days built big table and custom DB to not give money to Oracle, and not get locked in proprietary/costly solutions. This can play out the same way, non-nvidia chip makers and software products can implement this protocol and develop an ecosystem around networking solutions for AI.
Can I get an ‘explain like I’m 5’ translation?
I understood some of these words.
This kind of news is neat, but kinda useless because we don't know if it will actually result in hands-free FSD. Let's hope it will.
Tesla intelligent transport infrastructure enabling sentience. TITIES
let me understand one thing
so Elon bought Nvidia chips for X.ia and Tesla
but now tesla decides to introduce its DOJO chip, which before this was previously ignored by tesla with complete silence.
what are you playing?
From what I gathered from past AI days, they’ll use Nvidia chips while they are developing their own DOJO chips.
Dojo's approach is similar to what Cerebras is doing.
It was a hedge during COVID when chips were in short supply. They've been talking about it for years. It has taken a similar amount of time to release as it takes other chip manufacturers. Its speciality is pumping gigs of video through, instead of just the plain txt that LLMs require.
tesla decides to introduce its DOJO chip, which before this was previously ignored by tesla with complete silence.
Dojo was first mentioned by Musk in 2019 during Tesla's "Autonomy Investor Day", then revealed at Tesla's AI Day 2021. Hard to call public meetings with live streaming (then multiple Tweets) "complete silence".
Article was a bit technical. Can somebody verify that I understood correctly.
Tesla has replaced TCP with their own protocol, TTPoE, which transfers work from software to hardware. Does faster network benefit so much that it's positive effect despite needing more hardware? This is currently in use in part of their data centers?
So they're basically describing the first gen EFA network interface AWS had in 2018, and trying to pass it off as a new idea.