Tesla FSD Chip - Revolutionary or Over Hyped? ??‍♂️

Video quality	The size	Download

Информация о Tesla FSD Chip - Revolutionary or Over Hyped? ??‍♂️

Название	:	Tesla FSD Chip - Revolutionary or Over Hyped? ??‍♂️
Продолжительность	:	6.29
Дата публикации	:
Просмотров	:	61 rb

How do you guys feel about a fully autonomous vehicle? I’m a huge tech nerd and still think it will take some SERIOUS getting used to!
Comment from : Electrified

Yet again on 10/11/2022
Comment from : Wicked Fifth

No One Does What ELON DOES go TESLA TSLA!!
Comment from : Florencio Vela

Tesla do NOT have the hardware required for SAE Level 5 It is impossible for a Tesla out on the streets today to reach Level 5 It will NOT happen
Comment from : Xanthopteryx

Would love if you did a comparison Tesla FSD chip Vs Intel's MobilEye : ) Include Tesla, Nvidia and Mobileye maybe? Should be a very interesting video
Comment from : Pete H

New subscriber! Thanks for providing valuable information The video answered my question
Comment from : Laval P Olaya

It is overrated
Comment from : Gus Jackson

Great information All substance and data
Comment from : --

Another informative video from this channel opening my eyes to something I didn't know before Honestly The last time I heard about chips being used for self driving cars was 2016 when codename NX speculation was at fever pitch I can remember Supra metallic dave did a video outlining why he thought the rumor about NX being the very thing it turned out to be when revealed as the switch The davester was saying that Nvidia's focus was on making chips for highend pc graphics cards chips for self driving cars and that there was zero evidence at the time that Nvidia had any interest in getting back into video games after their contributions to the original Xbox and PS3 were outshines by AMD and what they could provide brbrDave was obviously wrong as we know that Nvidia are the wind beneath the switch's wings And based on this video it's clear that they won't be the wind beneath Tesla's wings now that they're doing what apple have done and making their own chips specifically for FSD applications I have nothing but respect for that
Comment from : unstoppableExodia

seriously, how much more biased can this be? no valid comparison with competition at all
Comment from : H C

Hm, so they sacrificed flexibilty and are now stuck with the chosen neural network implementation automatically excluding future innovations in this field
Comment from : 81FXB

It would also be nice to know about the two chip redundancy, and how these signals are amplified to control the car through what type of stepping motors, solenoids, etc Just sayin
Comment from : Jim Larsen

Heinrich Peter Maria Radojewski Schäfer Leverkusen

Oops 3:20 no no no Lider is a sensorbrDid you say neural network chip architecture? brbrAnd: until Tesla has complete voice control ofbrall information and commands, it's only halfbrthe story Who wants to get lost in thebrmonitor's menu maze?
Comment from : Heinrich Peter Maria Radojewski Schäfer Leverkusen

Their cars as of today cannot be fully self driving SAE 5 They cannot even handle snowy weather
Comment from : Xanthopteryx

Excellent, informative and original Thank you New subscriber now!
Comment from : Deb Banerji

Thanks for the video!! :)
Comment from : kraken1122

3:19 LiDAR isn't a type of neural network architecture 🤔
Comment from : David Metcalfe

It takes balls, the size of Elon Musk's, to decide that you can do better than the bestand actually do it!
Comment from : Faustin

To answer your question: it is impossible to tell from this, but 100 watts for a CPU is A LOT A quad-core, Intel Core i7 with 8 threads running at 42 GHz uses less power than that, and it includes built-in 4K UHD video on die The proof will be in the results The fastes nVidia chip can do more than 144 TOPS, but this number is misleading in that absolute operations per second may not be the determining factor in computational capability It all depends on the SOFTWARE
Comment from : SalvadorMonella

in house designed not built
Comment from : Mr Picky

Range is everything and as such power consumption is primordial NVIDIA simply didn't understand this or didn't care because they want to sell to ICE car producers
Comment from : Matt Casters

4:09 Tesla FSD compared with other chips
Comment from : Doug Grinbergs

All those big corporations have become complacent in innovationbr(Nvidia, apple, ford) while still raking in the Dough This must stop! Thank you ELON!
Comment from : Mr Ramirez

Make their own ship is the best way
Comment from : fred hernandez

Even elon said on the presentation that nvidea is a great company but they make a very generic chip instead of optimized for image processing nural-network they just optimized it i think custom chip would be a thing in the future for other companies also
Comment from : Serak Shiferaw

Interesting information with a typically bad reading/writing style Please think about pace and edit your scripts to be easier to listen to
Comment from : Justice Vanpool

What made it amazing is this is ASIC The code is in the silicon
Comment from : Colin Maharaj

🌞
Comment from : ZACHARY ROEBUCK

Definitely over hyped and not ready for prime time I've posted in depth about the AI spatial recognition and discrimination a number of times and won't do that again here, as it is long and complex The problems, in a nutshell, include construction, lane diversions, accident stoppages/slowdowns, fast lights, emergency vehicles, ripple effects, and a whole host of issues nothing less than a supercomputer can process in anything close to real time Yes, it can maybe work on a dedicated route, but historical evidence shows even that is problematic to this day Someday it will work, but because spatial perception of human beings is thousands of times better and faster than current generation AI/hardware, and humans have all kinds of serious (and quite often fatal) accidents, you aren't going to see any company able to field a commercial model FSD car due to extreme liability issues
Comment from : Gypsy JR

New subscriber here Thanks for the reply
Comment from : SevenDeMagnus

Bunch of bull shit never trust a machine with your life let Musk do it personally do it oh he’s got a personal driver that’s human
Comment from : P Zakkly

So cool Hardware 3:-)
Comment from : SevenDeMagnus

You downplayed the reason I think they did thatit was so they can retrofit the new system into older cars
Comment from : CoolRaul

According to the video, you would have broken tons of traffic rules in Germany
Comment from : antisteo

5:13 I thought the HW 30 had to be under 100w and ended up only needing 72 watts
Comment from : Andrew

I get the new chip this Friday for my Model 3 I am very excited that my baby will get a new brain I can fully see her actually talking to me in the future
Comment from : shepherdsknoll8

Thank you for the review of Tesla's processing hardware It helps explain how Tesla has delivered the most innovative driver assistance features available on the market today Unfortunately Tesla has continued to fall short of the capabilities that Elon has promised Fortunately for Tesla they have not tied their capabilities to any industry standards for autonomous vehicles, this allows a great deal of flexibility in defining what exactly is "Full Self Driving" Elon has recently stated FSD is feature complete but that it “doesn’t mean features are working well” Tesla's FSD currently operates at SAE Autonomous level 3 and may have the ability with the current hardware to reach level 4 but it will not be able to act as a robotaxi as Elon has stated last summer I would be curious to know how many people purchased the FSD feature with the idea of using it as a robotaxi someday? The more important question is does it make sense to purchase the FSD feature for what it can do? I should also mention that have worked with autonomous ground vehicles since the second DARPA Grand Challenge and autonomous vehicles in general for the 25 years
Comment from : flantc

NN base AI is hardware and software Software for NN is data A lot of databrOnly Tesla has enough data flow for training NN in few years For other competitors, which has 1000 times less real world data So, if Tesla needs another two years, others need 2000 years This difference made other competitors feel like vision based self driving impossible
Comment from : Pepperidge Cookies

"NVIDIA's chips must support a large number of neural network architectures, like LIDAR"brOMG
Comment from : Bokolo Aranyfa

Tesla 'Frame Shift Drive' chip?
Comment from : terner

Lidar isn't a neural network
Comment from : DELTACX10

They can develop spacex this is piece of cake for them
Comment from : Suwir Wong

Lol you compared Tops and said that beside the optimized chipset for their software nothing is revolutionary That cannot be further from the truth Their hardware redundancy capabilities itself are an order of magnitude better than existing solutions There are many other advantages to their architecture and design as well
Comment from : Rohit Kurup

Nvidia’s chip is a super-parallel polygon processor meaning tasks need to be split into numerous concurrent calculations It becomes a Herculean effort to keep tabs on the calculations Hence the chips become inefficient in both MIPS utilization and power consumption Mobileye is a better comparison to Tesla
Comment from : Greg Helton

Cool video The discord link looks expired
Comment from : welsingetorix

So, NIVIDIA lost Tesla
Comment from : DaB

Anyone that has experience running tensorflow on GPU knows how power hungry GPU are Building ASIC is always about efficiency It's why Google create TPU It's why the industry on the whole are using mixed precision and tensor optimized float docsnvidiacom/deeplearning/sdk/mixed-precision-training/indexhtml brbrbrbrgoogle's article about bfloat16 cloudgooglecom/blog/products/ai-machine-learning/bfloat16-the-secret-to-high-performance-on-cloud-tpus
Comment from : woolfel

What the hell do You mean by overhyped chip? Wrong the chips are manufactured by MICRON NOT SAMSUNG If NVIDEA CHIP was any good they had no need to develop their own chip that used 70 watts not 100 watts Why don’t You mention Jim Kellers name who was the architect of chip and it was not just 1 but a 2nd one for redundancy
Comment from : Butch Bailey

Love that you included the pegasus in the comparison, I felt like it was less biased Subscribed Keep the great content coming!
Comment from : Seth Heerschap

According to your timelines HW4 chip could be complete later this yearbr( 18 month after former chip completed)brElon mentioned it will be 3 times more powerful and faster than the HW3 chip
Comment from : Tomer Hacohen

When Elon introduced the new FSD chip he said there will be a second-generation in about a year and 1/2 Has anyone heard any updates on this I'm going to buy a model S and want to make sure it has the latest and greatest as I plan on keeping it for a long time
Comment from : JAW