TSLA Market Action: 2018 Investor Roundtable

[RoadsterPlease]

277 incoming, 280 close?

 

[Tslynk67]

277 incoming, 280 close?

We need one of our legendary last 10 minutes of madness to make that... Seen it before though, many times!

 

[MP3Mike]

I don't think there's any "integrated GPU" on the Nvidia board. The Parker chips are 2+4 superscalar ARM cores, for execution, they are not vector CPUs.

The Parker chips contain a Pascal GPU: Introducing Parker, NVIDIA’s Newest SOC for Autonomous Vehicles | NVIDIA Blog

A new 256-core Pascal GPU in Parker delivers the performance needed to run advanced deep learning inference algorithms for self-driving capabilities. And it offers the raw graphics performance and features to power multiple high-resolution displays, such as cockpit instrument displays and in-vehicle infotainment panels.

The "discrete GPU" Elon refers to is the "Pascal chip": 12 TFLOPs monster with 256 hardware cores.

While it appears that the GP102 has 3328 CUDA cores.

 

[SpaceCash]

HOD close looks inevitable and S3XY as ffffffff

 

[Cherry Wine]

HOD?

EDIT: Oh, "high of day".

 

[MikeQ]

(There's also the question of whether the Tesla AI chip uses separate RAM modules - a possible alternate design would be for the RAM to be integrated into the AI chip itself, as a sort of very fast transistor based SRAM. This would have a number of other advantages as well, such as close proximity of NN 'weight' data with the functional units representing 'neuron' nodes.)

HBM and interposer.

 

[Fact Checking]

The Parker chips contain a Pascal GPU: Introducing Parker, NVIDIA’s Newest SOC for Autonomous Vehicles | NVIDIA Blog

Very good point - thanks for the correction - fortunately I could still edit my post which was really wrong about that. Please see my updated comment - it now includes the image of the Parker chip you linked to, which shows the integrated GPU.

This explains Elon's point about the integrated GPUs!

 

[Unpilot]

let me put on my Space Cash interpreter..H-High O-of D-Day

 

[Cherry Wine]

By the way, congrats to whomever in here bought those 1/19/19 $300 calls at $18.20. Trading at $23.13 as I write this. I only had the cash for a single $360 call...hoping to sell to close on a nice ER bump.

 

[Tslynk67]

So popular mechanics has a new article titled "In defence of Elon Musk". I spent around half an hour reading it and my perspective has changed on the significance of some of the more erratic parts of Elons last few months. Even though I always supported him and understood he was under immense pressure from many sides, I was somewhat concerned about what was going on with the pedo tweets and the lack of a formalised approach to "funding secured ". Honestly, the last few months now feels like a footnote in an incredible story that is about to unfold. Nothing more. Please read if you have the time.

In Defense of Elon Musk

For those that didn't do so yet, take the time to read the above article, it will top-off what has been a great day for $TSLA.

 

[Chop110]

While I am happy the stock is bouncing back up, I went all in at 315 and did not have funds to take advantage of the massive dip. I'll be doing better once we are back over 300 just before earnings.

 

[humbaba]

the last minute selling push didn't even take a full dollar off the daily high. That was a good trading day.

 

[Nocturnal]

I didn’t read it initially or even click on it, but did, and regret it. Murdoch seems like such an odd duck in the board. Any chance the shorts were planning a hostile takeover/buyout, and he’s part of it? I know he was appointed last year or so... just thinkin’. I don’t know much about him, other than FOX. Anyhow, prob not right place for this post. Sorry.

Murdoch has been quoted as one of Musk's friends.

 

[Fact Checking]

While it appears that the GP102 has 3328 CUDA cores.

Indeed, so the distribution of CUDA cores is 3,328+256+256.

So most of the processing power is in the discrete graphics chip - but there's another ~20% in the two Parker chips as well.

 

[Tslynk67]

**** me, look at Netflix...

 

[mickificki]

For those that didn't do so yet, take the time to read the above article, it will top-off what has been a great day for $TSLA.

That was a phenomenal article. So damn good. Almost as good as a banana.

 

[MikeQ]

Edit: I think you are right!

This is the HW 2.5 layout of the Nvidia board (AutoPilot ECU):

The "discrete GPU" Elon refers to is the "Pascal chip": 12 TFLOPs monster with 256 hardware cores.

Note the four black squares above the Pascal chip: those are the GPU RAM chips, 2 GB each, 8 GB total. They are close to the Pascal chip for maximum performance.

Edit/Correction:

According to @MP2Mike the Parker chips include a 256-core Pascal integrated GPU as well:

The Parker chips also include 2+4 superscalar ARM cores, for execution.

But there's also a second board in the Tesla HW 2.5 computer module, an Intel board (the MCU), reportedly with an Gordon Peak and Apollo Lake SoC's: I have not found any tear-down images - the board's more interesting side is sealed inside the computing module.

Intel Apollo Lake Atom CPUs do come with an integrated GPU: but its performance is not comparable to that of the Pascal chip - it's maybe 2-5% of its performance. So I doubt it's used to process video data or do any significant AutoPilot work - it might be used for smoother infotainment graphics though.

I think all of AutoPilot runs on the Nvidia board: the neural nets execute in the discrete Pascal GPU (GP106 based), and the two Parker chip integrated GPUs, processing input from all 8 cameras, the sonars and the radar in that single chip, and the actual high level vehicle control logic (which uses the continuous stream of NN output) executes on the ARM cores of the two Parker chips.

The Intel board is used for visualization and all the apps and infotainment processing. It has an integrated GPU for smooth graphics: a HD Graphics 500 or 505 integrated GPU.

You really don't want the infotainment system having anything to do with the autonomy features. Especially if third party apps are ever to be supported on the infotainment system. Can you imagine what would happen to Tesla stock if an Elon tweet crashed the infotainment system and caused cars on autopilot to crash?

On the other hand, Elon could use this as an Order 66 to take out any enemies....

 

[Tslynk67]

Indeed, so the distribution of CUDA cores is 3,328+256+256.

So most of the processing power is in the discrete graphics chip - but there's another ~20% in the two Parker chips as well.

You say there's an Atom processor in the mix, I hope it's not the C2xxx series - seems to be an issue with those...?

 

[MP3Mike]

Indeed, so the distribution of CUDA cores is 3,328+256+256.

So most of the processing power is in the discrete graphics chip - but there's another ~20% in the two Parker chips as well.

Something to keep in mind is that EAP has to run on the AP2 hardware as well which only has one Parker chip. So there is only ~7.5% more power available in the Parker(s) for use by EAP. (My guess is that they are using the second Parker, or parts of it, in the AP2.5 hardware to enable the dashcam feature.)

 

[Fact Checking]

While it appears that the GP102 has 3328 CUDA cores.

BTW., if they are really using a GP102-equivalent Pascal chip with 3,328 CUDA cores, then that's about 85% of the available GPU computing power (the two Parker chips have 256/256 CUDA cores) - so I think the discrete Pascal GPU is probably processing most of the cameras. The 7% CUDA cores available on the Parker side might be able to process down-scaled images only, with reduced size neural nets.

I.e. exactly as you originally suggested.