Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

This site may earn commission on affiliate links.
You've hit the key point here is that by the time Tesla enters the profitable larger markets Waymo and others will already be there crushing margins. It will get bloody quickly (great for consumers) but bad for margins. 2025 Waymo starts receiving bulk deliveries of the chinese made for waymo robotaxi.

Go look at Google and GM's margins on these ventures. They are crushing alright . . . just not in the way you were thinking. Billions in losses for each.
 
It's a decent compromise which allows the more cab forward design of the CT, which is more important in my opinion.

I strongly agree that the design of the CT with cab forward is the superior design.

Although the roomy frunk on the F150 lightning is awesome, the design is wrong. Ford will hopefully adjust it in the coming next gen.

Cab forward gives improved visibility and critical benefits in certain types of collisions IMO. Smart design born out of first principles thinking. Others will be smart to copy it.
 
Wow, now the CT frunk is a major problem? 🙄

If Tesla knew what was important, the Cybertruck would have been front/back symmetric. Full 6.5 foot bed and tonneau cover in front to match the one in back. Just roll up the front tonneau if you need to see out to drive.

That would have been truly revolutionary, but without this the company will clearly fall to competitors.

Look at me, I'm an automotive/stonk analyst!
 
Cruise and Waymo are expanding real robo-taxi services.


This is problematic for TSLA investors, as margins have been compressed, many of this forum have changed their tune to focusing more on FSD as the main driver of future earnings growth.

Yes Waymo and Cruise rely on Lidar and HD maps (big nothingburgers) and bigger computers that cost more $$$, but they are beginning to capture the hearts and minds of people who experience this new technology. Unlike sexy EVs, Tesla is not at the forefront and will not be the brand associated with the tech.

Meanwhile Tesla is nowhere close to starting robotaxi services. Even if the software improves 10x, it won't be good enough for robotaxi. It likely needs to get 100x better, so that's going to take a few years. And then on top of that, its going to take time to actually ramp robotaxi services (and work out the kinks) even after the software is a safe driver.

So Cruise and Waymo are going to have another 2-3 years of easy expansion (much like Tesla did in EVs for a decade).

While Tesla may overcome this at some point, I don't see how this could be good for valuation in the next few years. How will investors gain confidence Tesla can actually secure much of this market when they see competitors way out ahead earning robotaxi revenue and growing services rapidly?

Cruise and Waymo are not going to have another “2-3 years of easy expansion”. They are going to burn billions on a model that can not scale. There is a reason Google cut Waymo loose.

No one is ready to deploy robotaxis st scale. Talk of Tesla robotaxis is premature as long as FSD has a single digit take rate.
 
I have won over a few people in my rural Utah community by giving them a ride in my Model Y. They talk about it every time they see me. For most it is still, "They will have to pry the diesel F-150 out of my cold dead hands."
It is remarkable how much more effective it is to let people actually drive the Tesla. It's easy to forget just how stunning it is when compared to an ICE vehicle.
 
Pepsi doesn't haul potato chips, they haul beverages.
Pepsico owns Frito-Lay, who received ~half of the semis delivered.

iu
 
Some more cost savings in HW4 infotainment (on Y at least)- apparently they halved the RAM and local storage compared to HW3 Ryzen

Some of the debate on this is ludicrous. 8GB of RAM? oh the horror. People are concerned about the ability to do visualisations on the navigation map and the UI.
I coded this stuff for a living, for over twenty years. Let me tell you: even on a PC, 8GB is a LUDICROUS amount of memory. Assuming a screen thats roughly 2000x1000 pixels (2 million), and assuming 32 bit color, thats 8m bytes: 8MB (roughly) to have a copy of the screen. 16MB to double-buffer it (which you should). Now assume triple buffering for ridiculous overkill. 24MB. Now lets quadruple that, so we can cache every possible texture that we will put on the s screen with any UI option at any point ever: 96MB.
Now lets cache 10 full screens for the video options for netflix, disney and oh I dunno.... 10 other services? so 100 more full screens is another 800MB. That takes us to 900MB (ish) or roughly one eighth of the RAM available, to have everything run super-smoothly at 60FPS minimum, even forgetting that in the omg-disaster world where we need to page-in some data from the disk, its a flipping SSD so its lightning fast anyway...

I know this is niche, but the idea that its hard to do silky-smooth 60FPS HD graphics with ONLY 8GB of RAM is ludicrous.

Oh and BTW its about 5x easier when you are coding to a known hardware target to which you have exclusive access and no copy of windows to argue with.

I have no idea what hardware FSD needs, but when it comes to infotainment, 8GB is colossal overkill, 16GB is insane. Anybody getting annoyed about that change has zero idea what they are talking about. Tesla probably worked out that their RAM requirements were way lower than anticipated.
 
Some of the debate on this is ludicrous. 8GB of RAM? oh the horror. People are concerned about the ability to do visualisations on the navigation map and the UI.
I coded this stuff for a living, for over twenty years. Let me tell you: even on a PC, 8GB is a LUDICROUS amount of memory. Assuming a screen thats roughly 2000x1000 pixels (2 million), and assuming 32 bit color, thats 8m bytes: 8MB (roughly) to have a copy of the screen. 16MB to double-buffer it (which you should). Now assume triple buffering for ridiculous overkill. 24MB. Now lets quadruple that, so we can cache every possible texture that we will put on the s screen with any UI option at any point ever: 96MB.
Now lets cache 10 full screens for the video options for netflix, disney and oh I dunno.... 10 other services? so 100 more full screens is another 800MB. That takes us to 900MB (ish) or roughly one eighth of the RAM available, to have everything run super-smoothly at 60FPS minimum, even forgetting that in the omg-disaster world where we need to page-in some data from the disk, its a flipping SSD so its lightning fast anyway...

I know this is niche, but the idea that its hard to do silky-smooth 60FPS HD graphics with ONLY 8GB of RAM is ludicrous.

Oh and BTW its about 5x easier when you are coding to a known hardware target to which you have exclusive access and no copy of windows to argue with.

I have no idea what hardware FSD needs, but when it comes to infotainment, 8GB is colossal overkill, 16GB is insane. Anybody getting annoyed about that change has zero idea what they are talking about. Tesla probably worked out that their RAM requirements were way lower than anticipated.



...you seem to be conflating video memory with system memory? In this case, since there's no discrete GPU on non-S/X cars, the same memory has to be shared for both things, while your math is only considering display buffer size not actually running any software or anything else.

That said- I agree 8GB is fine to run the infotainment OS. Most concerns were around how it'd do trying to run the more modern video games folks thought would be coming to these cars and so far only are available on the S/X.... there's plenty of games that have 16GB as their recommended, or in some cases minimum required memory.

The fact they've halved the memory AND the storage (some games would use the entire HW4 NVME storage for a single game these days- some won't even fit at all with only 128GB of storage) suggests anybody waiting for that is gonna be waiting a long time.


As I wrote- if they have no plans to do steam/local higher-end gaming after all on the non-S/X then this is a reasonable cost savings.... some sort of cloud-based gaming service would also "solve" this to some extent since most of the need horsepower would be on the remote end.
 
Last edited:
...you seem to be conflating video memory with system memory? In this case, since there's no discrete GPU on non-S/X cars, the same memory has to be shared for both things, while your math is only considering display buffer size not actually running any software or anything else.

That said- I agree 8MB is fine to run the infotainment OS. Most concerns were around how it'd do trying to run the more modern video games folks thought would be coming to these cars and so far only are available on the S/X.... there's plenty of games that have 16GB as their recommended, or in some cases minimum required memory.
You know the APU of the infotainment system has like an old old Vega GPU with 11CU. Any games that require 16gb of ram is already a mismatch for the GPU of this chip. 5+ year old games can work decently on this chip but only at low resolutions (720p, low settings), which means 8gb of ram will be far from being saturated. I mean last gen high end dedicated GPU like a GTX 3070 only has 8gb of ram.
 
Last edited:
Sure, but the video being discussed, and the information from it, was all about the Semis that Pepsi is using, not the ones that Frito-Lay is using.

Pepsi(co) isn't using the Frito Lay semis to haul Doritos and such?

Considering the paint jobs I always figured they were double-dipping.

That's the soda news that'll be hard to wash down.

(we could keep chipping away at this...) ;)
 
...you seem to be conflating video memory with system memory? In this case, since there's no discrete GPU on non-S/X cars, the same memory has to be shared for both things, while your math is only considering display buffer size not actually running any software or anything else.

That said- I agree 8GB is fine to run the infotainment OS. Most concerns were around how it'd do trying to run the more modern video games folks thought would be coming to these cars and so far only are available on the S/X.... there's plenty of games that have 16GB as their recommended, or in some cases minimum required memory.

The fact they've halved the memory AND the storage (some games would use the entire HW4 NVME storage for a single game these days- some won't even fit at all with only 128GB of storage) suggests anybody waiting for that is gonna be waiting a long time.


As I wrote- if they have no plans to do steam/local higher-end gaming after all on the non-S/X then this is a reasonable cost savings.... some sort of cloud-based gaming service would also "solve" this to some extent since most of the need horsepower would be on the remote end.
Yup, I know it will be RAM and not VRAM, but it makes no difference. Adding more RAM doesnt reduce the latency from RAM to screen. In fact when you accept its all RAM, then its even more true, because in some setups the VRAM gets shadowed in system memory, but in this case its all in the same place.
The laptop I'm typing this on is all system RAM. Intel likely uses some marketing BS to pretend otherwise, but its definitely not dedicated :D.

And yes, playing Battlefield 2042 on a HW4 car with 8GB RAM will not be ideal, but this also depends on screen size. Even the model S screen is not that big compared to high end gaming monitors.
But frankly who the hell is playing a super-hi-end twitch shooter in their car anyway? :D

My argument is not that 16GB cant be justified for gaming. Its the idea that this will be in ANY WAY noticeable outside of playing modern games. To me, this makes way more sense as a cost saving than lumbar support or USS removal. Gaming is way down the list of features in a car.