Jetson Nano Review

The Jetson Nano is NVIDIA's latest machine learning board in its Jetson range. It costs just $99 for a full development board with a quad-core Cortex-A57 CPU and a 128 CUDA core GPU. There is also 4GB of RAM and USB 3.0. Here is my full review.
Introduction to Android app development: www.dgitacademy.com
Let Me Explain T-shirt: teespring.com/gary-explains-l...
Twitter: / garyexplains
Instagram: / garyexplains
#garyexplains

[@1MarkKeller]

*GARY!!!* *Good Afternoon Professor!* *Good Afternoon Fellow Classmates!* Great info as usual Professor. I don't know if I will ever use it like I do the digital camera info or the phone reviews/speed tests, but I come here to learn about it all just the same.

[@fatehyabali]

MARK !!!!!!

[@GaryExplains]

MARK!!!

[@1MarkKeller]

@@fatehyabali *FATEHYABALI!*

[@saisrikargollamudi7892]

Thank you for such a complete review.

[@peegee101]

I feel that your videos are improving, love the channel!

[@shikhanshu]

amazingly comprehensive and informative review... this channel is a gem

[@juliannash3590]

Awesome review, looking forward to getting my hands on one!

[@benjaminlittle7916]

Great review, thank you!

[@hankgraham5818]

Thanks Gary! Great Video!

[@taidee]

Healthy pickings from the tree of knowledge, thank you Prof.

[@vapourmile]

Very good presentation. Very informative.

[@VictorMollo]

Great review, even some stuff in there that is a bit buried in NVIDIA's own documentation. Of course, three months after your review the Raspberry Pi 4 came out with up to 4GB RAM and USB 3.0 :-)

[@xKhaozZer0x]

Will this work as a htpc to decode 4k hdr?

[@williamhart4896]

Thanks professor looking better than I thought .

[@eduaulas]

great review! tank's

[@hk2780]

Do you have plan for any project? I am just wondering that battery recommendation about this.

[@shoshinschool2793]

which fan you are using, i want to buy same model, can you mention the specification or model number

[@qkcb]

Yes,it is here！

[@forceindependance3623]

What cooler should I buy for my jetson? Thank you

[@rajpandey22457]

Does it work like raspberry pi, means running rfid card reader, ir sensor from gpio pins? And does it support picroft?

[@pangrac1]

Good for film footage scanners

[@steveb.548]

As Gary explained, things may be "rough around edges" and it might take a little while till 'everything just works' -- but unlike other ARM SOC boards, the Jetson Nano is based on Nvidia's proven Maxwell series GPU cores which have literally man years of Linux driver development time behind them. So, ultimately the Jetson Nano will offer an order of magnitude better Linux desktop support than other ARM SBC toys like the Raspberry Pi. So aside from 4 Gigs of RAM and a much more powerful CPU and GPU, the Jetson Nano gives you full up to date OpenGL and OpenGLES, full Accelerated Video Libraries, and CUDA support (just like Nvidia graphics on a high end Intel PC, lower processing power perhaps, but same basic features). Other ARM single board computers don't even come close.

[@TheNikolaschoulakis]

Please share a URL for that 5V 4pin 40mm fan you have on ? I cant find it anywhere online in the UK.

[@kevinknight997]

Brilliant video

[@MaidLucy]

the real question is: does the open source graphics driver work?

[@selfdestructint]

I can confirm that with a little rearranging of the cables the Jetson Nano works fine with the Raspad, giving you an Ubuntu tablet. It doesn't accommodate the height and port configuration of the Nano, but I will be designing a replacement back and uploading to Thingieverse when I have time to scratch my backside.

[@bahathir_]

Great review. I have RAspberry Pie 3B+ and it is my main computer. I connected to 24" LCD TB and powered it by USB from the TV. Great low power SOC. Just nice to run GNU Emacs and other CLI programs. The Jetson Nano is promising with 4GB RAM, and more powerful CPU, GPU and better USB 3. But it lacks of Blutooth and WIFI it deal breakers for me. I connected my Bluetooth keyboard and access to network through WiFi.. with Raspberry Pie 3B+. I may consider to upgrade to Jetson Nano if NVIDIA adds Bluetooth and WiFi capability in future version of the product. Thank you for great review. .

[@SoussiAfif]

You can get a cheap usb wireless adapter and call it a day

[@bahathir_]

@@SoussiAfif Thanks for the suggestion .

[@theminertom11551]

Great video. I do have a question: As you mentioned, one of the demos showed a simulation of a ML application using 8 cameras as input. You did mention that this was pushing the limit of the board. I would like to know of the hardware configuration of the 8 cameras. Obviously there are not 8 USB interfaces on the board and I am not sure that USB would be able to support the BW of 8 video feeds unless the resolution was really low. Would these cameras be connected via Ethernet? Seems like an elementary question to ask but I'm not finding much data.

[@ajr993]

You can likely stream udp over ethermet. 8 different addresses that stream data to the jetson

[@dalebetterton5255]

DIY Security monitoring, motion detection triggered recording and remote video feed access system?

[@SahaParikshit]

hi sir i have two queries. 1.Do jetson nano gpi0 40pin gpio has same internal connection like rpi 40 pin gpio? 2. i have "Arducam Multi Camera Adapter Module Fully Compatible for Raspberry Pi Model A/B/B+". Can it be useable with this jetson nano? and one out of the box query. i found Google's Coral Dev Board is also there. Intel's Myraid based some SBC is also there. I'm planning to use an SBC for mainly deep learning based computer vision application development. Please suggest me which would be the best as per your long term experience in this field. can this jetson nano sbc be a good cost effective substitute of the tx1 or tx2 or xavier. Or there is significant performance degradation due to cost & size optimization in the newer one.

[@LaughingOrange]

This product should only be used for prototyping, not training. It is slower than the cheaper GT 730 dGPU. Use a proper GPU for training, and then use this board to test it for real world applications.

[@TheJudge064]

Yes, of course. I imagine any machine learning AI will take ages to learn with such an underpowered processor. Thankfully there are some pre trained ai you can get to test out on it.

[@alvaroesquivel6338]

Do somebody know if it is available for Mexico? I wanted to buy one in the Sparkfun web page but it has some countries with shipping restrictions and Mexico was in the list. I do not know If it is just in that page (sparkfun’s) or is a kind or legal restriction.

[@mrrolandlawrence]

i got a feeling this is going to be one hot seller. if xbmc works on it, then thats pretty much most of the other similar priced SBCs done. wish it had emmc 5.1 though instead of micro SD. otherwise its perfect.

[@ChadKenova]

I love my shield tv I bought it not to long after it came out. Ive seen 2nd generation’s go for $130 on sale and thats a much better deal but ive always wanted a pi, tinker board, or something like this. To bad they cut it down from the shields x1, but it does have 1gb more ram. The shield runs a 2ghz cpu with 4 big A57 cores and 4 small A53 one’s and 1ghz and 256 cuda cores on the gpu. But I would like to see this vs the odroid xu4 and the asus tinker board

[@dgb5820]

Great video

[@GaryExplains]

Glad you enjoyed it

[@scottwatschke4192]

Thank you.

[@ingusmant]

Hey Gary I know it's not the same category but how does this chip compare to the 8cx?

[@GaryExplains]

On the GPU side, I would guess the Jetson is more powerful. On the CPU side I would say the 8cx is better. One interesting thing to note is that Qualcomm is pushing AI via its DSP, not via the GPU, so it would be interesting (but hard at this stage) to see how they compared performance wise for AI.

[@qazimashhood]

Guys, I have a question, i want to differentiate weed plants from normal plant using a live video feed! Will this be able to do that, or should i go with something more powerful?

[@qazimashhood]

@Gonçalo Amaro Thank you :)

[@WhyGoThere]

pip install tensorflow-gpu is what you should have done first

[@mattizzle81]

Yeah forget training neural networks on something like this. This is meant for inference, especially where you need low latency. If latency isn't an issue, you are better off streaming images to a remote desktop server over a network. If you need something to have fast response times on video though, this is the kind of tech you need. I have been waiting for something like this to become available. For example I want to do object detection on top of Augmented reality, by plugging a smartphone directly into it and streaming the images over USB. Streaming a video to a server won't work, because by time you encode the video, transfer over wifi, decode, detect an object, the phone has possibly already moved quite a bit. If you don't need that real-time and have a wifi connection, then probably better off with a desktop computer doing the ML.

[@WilliamDye-willdye]

Yep. That's why I have trouble justifying the purchase of this board, or the Google Coral TPU beta board. I'm sure I could eventually find a use case for the on-board extra horsepower, even if it's just to prettify the user interface with simulated liquids flowing over the GUI window frames. Unfortunately, everything we have in the development pipeline at my job (designing scientific instruments) was designed to allow all the hard work to be done offline. I just don't have a strong use case yet to justify an exploration project.

[@aabb-zz9uw]

Training should be done on PC. Otherwise a much more expensive product would be required(the one for self driving cars)

[@TheDeelunatic]

I like how the Tegra tk1, with it's 198 cuda cores, is never mentioned. It came before the tx1 but was troublesome to use for much due to the way the gpio was set up. At least with this board there is more potential for using it for a raspi replacement.

[@GaryExplains]

You like that it was never mentioned?

[@TheDeelunatic]

@@GaryExplains I mean it in the most sarcastic of ways of course.

[@TheBauwssss]

I'm thinking cheap-ish, dedicated, fast, low power consumption OpenVPN server. My Raspberry Pi model 3B+ only does about 7 mbit/sec throughput with a 2048 bit certificate.

[@reaperforever8478]

Training the module for a line following robot would be really cool , no need to code for it ever again

[@aabb-zz9uw]

The product isn't available yet and is export controlled-only available to Level 1 allied countries. Cannot be sent to Taiwan ,Macau and Israel among others. It is also low-performance as mobile AI chips used by handphones are around 4 times faster. Thus, it is good for education(junior college level) in Level 1 allied countries but not for product integration.

[@rickjason215]

At first I get excited about new products and then I try to figure what I would do with it. The answer is play with it for about 15 minutes and then put it in one of my bags of old and useless tech, that for some reason I just can’t throw out.

[@melgibson6331]

ISrael already copied it through its control of microsoft, intel, google and the NSA and sold the tech to China I'd wager.....Why is Taiwan censored? They are more into democracy and independence than the jUSA.

[@aabb-zz9uw]

@@melgibson6331 This is the export control list which is about the same as the list of nations of which the nationals require special permit for entering a US overseas base. Israel is a 2nd level ally and not considered by US and Anzac to be as trustworthy as EU,Korea and Japan(example:Israel sold technologies like Lavi and Python missiles to China) Israelis need visa to visit US for travel. As for Taiwan, there have been past defections of fighter pilots to China and chinese infiltration as well as many taiwanese companies relying on china. Taiwanese companies sent over trillion USD to China during their better days. And an attempted cross strati cooperation for semicons was thwarted by US policy. Taiwan makes semiconductors for China(Loongson,Sunway,Ingenics) and Russia(Elbrus) which are not allowed to import semicon manufacturing equipment due to export control. Sunway and Elbrus chips are used for larger computers such as supercomputers which could be used by chinese and russian military which is a very very dangerous thing,.

[@aabb-zz9uw]

@@melgibson6331 And Israel has no industry infra to make semiconductors. Even without export controls, Israel's economy is too small for that. Only Korea, a G3 superpower, can do that,

[@magno5157]

Is the GPU driver for Jetson Nano only available on Ubuntu?

[@sayemprodhanananta144]

I guess that's a generic driver

[@CarlValentine]

I really want to give Windows on ARM a try. I hope it will be up and running on this due to the increased performance over the raspberry pi

[@amirabudubai2279]

Windows on ARM looks DOA. It was only meant to be a compatibility layer for a transition to ARM, but it looks like developers/hardware companies aren't willing to make the move.

[@pruthuchauhan2159]

Did you try RetroPie/RetroArch on the Jetson Nano ?

[@GaryExplains]

No. I wasn't aware that there were versions for the Jetson Nano.

[@thekakan]

For anyone interested, here's the source code: github.com/garyexplains/examples/blob/master/threadtesttool.c :edit: Also, I don't think that many Pi HATs will be converted to support the Jetson. The reason? No one wants/needs it. If you have a Jetson, you must be doing something better than using a HAT to control some relays over a remote. I love the RasPi as well as the Jetson, but the use cases of both of these SBCs are really different.

[@andrewwhitfield5480]

It's time to emulate. Here's your new Wii or Gamecube!

[@zandatsu07]

I hope someone test this Jetson nano for ps4 emulator in Windows

[@DarkOracleOfDeath]

Actually the Nintendo Switch is also based around the Tegra CPU, so maybe you could port Switch games on this device...

[@bradfairbairn2870]

Don't you mean here's your first generation Wii that plays GameCube Games and had a really a really sad release on VC. for almost every for ever Nintendo 64 game. "The darkness, the lack of and HDMI port.... Rip the money printer. That red Wii served me well.

[@andrewwhitfield5480]

@@bradfairbairn2870 No. You should keep up or keep quiet.

[@tasoulman]

What about the emmc 16gb storage ? can I use it to install the OS or do I have to install a sd-card ?

[@GaryExplains]

The module on the development kit doesn't include the 16GB of storage, so you have to use the SD card.

[@tasoulman]

@@GaryExplains thnx

[@lashlarue59]

Does this device also have tensor cores?

[@Face2FaceHardware]

Michael Williams No. it’s just a single Maxwell SMM. No FP16 specialist here, as half precision is done on the CUDA cores.

[@eliaskpst]

Can I use this, for a Minecraft Server?

[@photoputer7797]

This is the only product of nvidia I would like buy . when down to outlet

[@Phynix72]

Can it run on Android ? What about G-test against raspberry Pi ?

[@DSTechMedia]

I would say it surely can. The Nvidia X1 SoC is the same one used in the Nvidia Shield TV console. However, the Android image needs to be from Nvidia because of the custom SoC drivers/firmware. If anyone has managed to hack/build their own you might try searching XDA developers website.

[@AndreiNeacsu]

To me, the Jetson looks like a stripped-down Shield (from unsold inventory) that's more of a novely that's aimed at collectors of SBC, rather than someone designing a real product. Maxwell is 5 years old. As for me, I'm more of a AMD Ryzen Embedded V1605B (twice the price) sort of person. For small projects (lightweight, little power) I use anything from Atmel 32U4, to Raspberry Pi Zero W. For serious things, like in a car, I prefer a quad-core Ryzen CPU with SMT and 8 Vega GPU cores and 32GB of RAM. If I were an Nvidia king of person, I would use the Shield as a development platform for serious things. For playing around with Pi-likes, there are many clones with 8 cores, more RAM, etc.

[@aabb-zz9uw]

Performance is only 1/4 of mobile chips like Kirin 980

[@kristinafraxx5808]

Gary!!! Greetings from yasnaya polyana!!

[@GaryExplains]

Greetings!!! :-)

[@1MarkKeller]

*BIGBY!*

[@RetroBerner]

Nice

[@adamaleksander5226]

Do you think I can run Blender 2.8 on it?

[@dutchdykefinger]

i like that question :) now i want to know aswell... all i know is that 2.79 does run on ARMv7 but i'm honestly i'm not too hopeful for eevee to get ported.

[@TheNikolaschoulakis]

I have run into the same issue with desktop sharing, it freezes. I hope nvidia fixes it soon but how can I VNC remotely into the jetson nano? Can you please make a step by step guide on how to actually connect and test the raspberry pi cam you used ? Can you also show how to run real-time object detection tensorflow models like mobilnet ssd over the camera?

[@GaryExplains]

The only way I could get VNC working was to run vino manually. You need to make sure you set encryption to false and prompt to false before you run it. 'gsettings set org.gnome.Vino require-encryption false' and 'gsettings set org.gnome.Vino prompt-enabled false' then run /usr/lib/vino/vino-server

[@TheNikolaschoulakis]

@@GaryExplains where can i find these settings to change encryption to false please? Is it a config file in some directory?

[@GaryExplains]

Just type those two commands at the shell prompt and then run vino.

[@TheNikolaschoulakis]

@Gary Explains can u provide a link for that fan u r using please? Is it a 5v 4 pin 40mm fan or is it 12v?

[@BuildTimeMC]

Can this do gaming?

[@dutchdykefinger]

yes, the r-pi already could. snes, mega drive, game boy, ms-dos you name it. you could probably add n64, ps1 and dreamcast to this computer. you probably weren't aiming at emulation though, in which case the the answer is: it's an ARM and it doensn't run your vanilla windows most mainstream games are only written for x86, so no, not really in that sense.

[@andersonfrans]

I wonder what would be the perfomance if they using Turing. Maybe TFLOPS?

[@GaryExplains]

The Jetson TX1 used a 1024-GFLOP Maxwell GPU with 256 CUDA cores. The TX2 offers 1.3 TFLOPs using a 256-core Pascal GPU, and the top-of-the-range Jetson AGX Xavier breaks 10 TFLOPs with its 512-core Nvidia Volta GPU.

[@andersonfrans]

Oh, so they make even the Volta version too? Wow. And the price sir?

[@husmen]

I wish it was at least Pascal, that would make it the perfect young brother of the TX2

[@AndreiNeacsu]

@@GaryExplains Easy there with the 10 TFLOPs claim! It's actually 1.3 TFLOPs in the 30W configuration. Don't confuse floating point operations with integer operations + Nvidia-marketing "tensor operations".

[@GaryExplains]

@@AndreiNeacsu From the AGX FAQ: What's the peak performance of Jetson AGX Xavier? Jetson AGX Xavier is capable of more than 30 TOPS (trillion operations per second) for deep learning and computer vision tasks. The 512-core Volta GPU with support for Tensor Cores and mixed-precision compute is capable of up to 11 TFLOPS FP16 or 22 TOPS INT8 compute. Jetson AGX Xavier's dual NVDLA engines are capable of 5 TOPS INT8 or 2.5 TFLOPS FP16 performance each.

[@BaliHouseMafia]

Can you put android onto it?

[@richardrichard9567]

Would this work with TensorFlow?

[@TheNikolaschoulakis]

Richard Richard yes it does

[@crosero]

Does it have Bluetooth?

[@piff57paff]

I like your videos, yet your assumption about how students learn programming at university does not reflect the reality at my university (KIT, Germany). We were taught parallel computing from the 3rd semester onward, if you took the courses. You could even a major in parallel computing at latest in 2009 probably even way before that, but I never checked how long this major existed before I took it. ;) We were taught everything from parallel algorithms on clusters, GPUs, multi-core CPUs, down to parallelism at the level of the CPU. One is not forced to take those courses, but it is rather easy to get exposure to those concepts at my university and I would be surprised if other universities would lag significantly behind in that regard.

[@ADR69]

So how much Python do you need to know to at least dabble in ai?

[@XadaTen]

Just start dabbling adn stop asking questions. You WILL learn it by doing it. You'll never learn it by talking about it.

[@TheJudge064]

Yeah, that's what I did and now I'm a software dev on the side. If you have a good question, I'm sure anyone on forums will be more than happy to help you out

[@Brot1984]

Don‘t know if you have any programming experience at all. If you have it is quite easy too learn it just by doing it. You can directly start with the ai stuff. For example with some tensorflow keras tutorials. If not maybe do some python tutorials first to grasp the basic concepts of programming before you start with the ai stuff. In general python is one of the easy too learn programming languages and at least as long as you are focused on high level programming and using libraries for the low level stuff it is a really powerful one.

[@SimonRichardMasters]

To learn, you must do too. To repeat, do more than two.

[@TheJudge064]

@@SimonRichardMasters well said.

[@leledumbo]

They have X1 successors already, but why do they keep on seeling things utilizing this SoC? Don't get me wrong, it's not a bad SoC, particularly in a form factor bigger than smartphone. I just keep remembering how flawed A57 design was.

[@amirabudubai2279]

The successors aren't better. Maxwell my be quite old, but the newer designs won't beat it in this setup; Pascal, Volta, and Turing don't have better performance/watt, so you won't see any improvement from a newer Cuda Core. In fact, almost the exact same Cuda Core was used up until Turing which added float/int pipe-lining. Same for the A57; they can't upgrade because their is no simple upgrade. If they did a full redesign and kept the same power target, you would be lucky to see a 2-3% improvement. The X2 just increased the power consumption, but then they discontinued it because that moved it into competing with the highly competitive 15W laptop processors. TL:DR. The X1 is actually better than the X2 and their is no way to beat the X1 until mass production of 7nm revs up.

[@leledumbo]

@@amirabudubai2279 From wikipedia, X2 is stated to have the similar TDP, as well as Xavier but this one can have TDP up up to twice the lowest configuration. I don't know how better Denver or Carmel cores are compared to A7x series, but I'm sure A57 is really that bad (thermal raises like rocket) because I used to have one (well, nvidia somehow nails the thermal better, but still no better than A7x). A7x is supposed to be backward compatible with A57 anyway, so I don't see any reason why this is no simple upgrade. After all, the smartphone markets have made it smoothly up to current A76. From overall SoC performance side, phoronix once done a comparison between TX1 and TX2, which are powered by respective SoC without T prefix, and TX2 gives 30-100% performance boost (Max-P mode, around half of it in Max-Q) for around 10W power consumption max (the same as TX1), even better the thermals are cooler than TX1. Anyway, I can't find any reference X2 being discontinued, in fact some rumors said new Nintendo Switch should be powered by it. Can you give some pointers? They only don't list it in their Tegra page, but it lists discontinued devices so I think it's the page that's out of date. The only "make sense" reason for me is pricing. This Jetson Nano is priced waaaaay below even TX1 (at launch). For $99, this is the cheapest SBC Nvidia ever released. So, using 2015 SoC sounds like a deal for the given price. After all, its performance is still above many cheap ARM SBC even some x86 ones.

[@amirabudubai2279]

@@leledumbo TDP does not equal power consumption. TDP usually works in a class system to make things easier on system integrators; it is cheaper to put a 15W TDP rated cooler on a 5W SoC than to create 10 different heatsinks. I can't find any any good benchmarks on the X2 since it seems to have only been used in one developer board since it came out 3 years ago, however you seem to have found some misinformation about it. According to Nvidia( www.nvidia.com/en-us/autonomous-machines/embedded-systems/ ) , it still uses the same A57 as the X1. 100% improvement in graphics sounds about right since it has the double the CUDA cores(and the die size to match), but no way it is in the same power class with 2 high power cores(in addition to the same cores in the X1) and twice the CUDA cores.

[@leledumbo]

@@amirabudubai2279 I know TDP doesn't equal power consumption, I just state it to see what thermal performance it's designed for. Check phoronix for the benchmarks, that's the first google result I get. I'll try to find another one, usually notebookcheck also has one. I didn't say that it doesn't use A57, it still uses A57 + additional custom cores (Denver in case of X2). Why I compare to A7x series goes back to my initial question that despite A57 is so bad, why they keep using it knowing that A7x is a lot better? There could be another money playing around, as using newer cores mean that they have to have another deal with ARM.

[@JayRKID]

Will this be better than the rasp pi for retropi?

[@GaryExplains]

Is retropi available for it?

[@ihaveanothereye]

I ordered one to run emulation tests on it. I will be trying all consoles up to and including gamecube...

[@JayRKID]

ihaveanothereye awesome

[@ihaveanothereye]

@@JayRKID unfortunately mine doesn't ship until April 19th. So I won't have results for a bit

[@OrcsMustDie-tl3dj]

Can it run Crysis?

[@muhammedadiljahfar5512]

Petition to have Mark kellers comment to always be pinned by prof.gary...great video as usual

[@GaryExplains]

Granted (for this video at least).

[@vedantgawade9812]

But can it run Crysis ?

[@Yetus]

Vedant Gawade actually yes but not crisis 3

[@vedantgawade9812]

@@Yetus . Worth it at this price point.

[@trialnterror]

How does this compare to processing power to something like a new i7 ?

[@GaryExplains]

Which i7, there are dozens of models and variations.

[@trialnterror]

@@GaryExplains that’s why I said the newest i7 ? Pick one and how would it compare?

[@GaryExplains]

@@trialnterror There is a huge difference in performance between the U, Y, K, etc versions of the i7. That is why you can't just pick one.

[@trialnterror]

@@GaryExplains is there a benchmark on these then? That’s where I’m going, Howe does this compare to.....

[@GaryExplains]

The Jetson has 4 Cortex-A57 cores. Processors using the A57 were first released in 2014. So 6 years ago. So if you compare it to a 2020 i7, the of course the i7 is faster.

[@andrewwhitfield5480]

$99 dollar miracle.

[@illustriouschin]

The modular design doesn't make sense for a drone because you'll need to plug it into something for i/o and power anyway.

[@jobicek]

The idea is that in a finished product, you can have an application specific board which will be lighter and cheaper compared to a full-on development board. Development boards are supposed to enable development, they're meant to be universal, giving you full (and preferably convenient) access to the chip in question, not optimized for a concrete application. In an actual drone, you probably wouldn't be using USB cables to connect components or a barrel jack to connect battery. You certainly wouldn't put on connectors that you don't need. And using a module can simplify the design of a board and make it cheaper. The downside is that it's more restrictive; the module is what it is. You have to accommodate it. Not to mention it's probably going to be heavier and use up more space. One of the many engineering decisions to make. This approach can be taken even with custom modules, not just off-the-shelf parts.

[@sineltor]

Eclipse probably doesn't run because this board "only" has 4 gigs of ram. Eclipse is a monster.

[@aabb-zz9uw]

Normally people will cross compile on pc then deploy on the board.

[@Sitharii]

lol !!! I have the exact same nVidia "synapse"-wallpaper😃 on my monitor !!! As a longtime nVidia fan and customer i have to say that i really love this wallpaper 😃 !!

[@illustriouschin]

Thanks for sharing I'm so happy you enjoyed it! 😂😂😂😂😂

[@JOELwindows7]

1:27 *My urge to get Jetson fell to 0* 2:35 Yep *it's real. My urge fell into -1* 5:07 oh, *my urge to get Jetson rises into +1* 8:26 perhaps you should get it. *+2*

[@GaryExplains]

Why?

[@JOELwindows7]

@@GaryExplains maybe I can use it, to scan something. If I don't scan something, maybe it will be useful to play some games. Not at max graphics obviously.

[@codebulletin]

@@JOELwindows7 Sir can i output 4k video from this to a monitor pls reply sir

[@JOELwindows7]

@@codebulletin I haven't yet an idea about this. If it is possible, maybe about 30 Hz or 60 Hz.

[@madmax43v3r]

Haha! Eclipse is broken on 18.04 Ubuntu across the board! It's a madness.

Haha that’s not Ubuntu 18 lol

[@madmax43v3r]

@ It is, nano runs on 18.04

[@Zenas521]

Terminator 800 in your pocket.

[@jastirria]

I thought the tk1 was first

[@tech-vp5xe]

Yup jetson TK1 not the TX1

[@teleringdataspesialisten1869]

But can it run Windows ARM + Game? ;D

[@Yash9278]

I think it can some.

[@manu7irl]

klipper is python based could run virtually on any platform

[@maximvdw]

Kinda sad it does not support OpenCL as well

[@geordonworley5618]

Are you sure it doesn't? I expect it to work, but I guess I will see when I get it.

[@maximvdw]

So far haven't been able to. Drivers won't find it as an available CL platform. Think they prefer you to use CUDA

[@harrytsang1501]

Linus: You'd need a viable arm-based development platform and cheaper deployment to get arm in servers Nvidia: the first half is done

[@243263843]

Can someone tell me if this can run with android?

[@TheDenkoken]

Yes it can but if your idea is to put Android better buy a nvdia shield is the same hardware yes it is 100 us more but come with controller memory an Android is all ready install

[@i077]

This board has about half the GPU core of a Shield, 73% of the CPU frequency, and disabled low power cluster. Unless you are really desperate, the shield with its bundle controller is 100% better than this and with OS preloaded no less.

[@timsievers2067]

But, can it run goldeneye?

[@GaryExplains]

The game or the weaponised satellite system? 😂

[@timsievers2067]

Gary Explains, the satellite weapons system may be doable, if it can do machine learning. ETA prime has his hands on one now, so he will most likely give it a shot!

[@fatehyabali]

Sir we are waiting patiently for Huawei P series Speed test G😃

[@wangkevin3431]

Just look at the mate 20

[@illustriouschin]

Have fun with sending your credit card info to the Chinese government.

[@ytmohits]

Vino Issue with Unity: bugs.launchpad.net/ubuntu/+source/unity-control-center/+bug/1775423 I use x11vnc (x11vnc -forever &) to fix vino enable flag with Unity or switch to gnome3 environment Eclipse issue: askubuntu.com/questions/1041197/eclipse-doesnt-start (snap stable version not available for arm64 as well) I use Nsight (Nvidia built Eclipse for full development support) from Jetpack or NVIDIA SDK Manager (developer.nvidia.com/embedded/downloads) UI sluggishness: This is kind of due to power mode/limitation. Clocking with official jetson_clocks.sh script increases overall perf and remove any clocks cap. Quake: Not sure what is the issue here. Cheers

[@kray97]

The desktop performance seems a bit sluggish to be honest.

[@ajr993]

I mean it's 99 dollars for a new complete computer. Like have you seen how aluggiah the 200 netbooks are? They suck! This is good relatively speaking.

[@kray97]

@@ajr993 It's sluggish on Ubuntu, not Windows. Generally running Ubuntu well is not that hard.

[@tlkh]

Probably I/O limited due to the microSD card

[@AKASHSOVIS]

Why do you always say 'hechDMI'?? He he Awsome Sir. Thanx

[@GaryExplains]

he he because that is the right way to say it he he

[@mr_beezlebub3985]

The only thing called Jetson that isn't made by Hanna-Barbera

[@AyushBakshi]

Hey Gary! I can't hear you properly. Granted my smartphone's speaker is crap but I don't find others' video this hard to listen. Will be easier for me if you raise the gain a lil in post.

[@GaryExplains]

I do raise the gain in post and I don't have a problem hearing the audio, also nobody else has complained. I would suggest the problem is at your end. Sorry.

[@bschlueter]

3-4 Years? No 5 years old en.wikipedia.org/wiki/Maxwell_(microarchitecture)

[@GaryExplains]

True. But technically the Jetson Nano uses a second generation Maxwell!

[@bschlueter]

@@GaryExplains oh sorry, then you are right 😅

[@mansaha9288]

It should be faster than SD855

[@GaryExplains]

Why?

[@mansaha9288]

Considering its size, it is bulkier than SD855.

[@GaryExplains]

@@mansaha9288 If you are talking GPU performance then yes the GPU in the Jetson Nano is more powerful. But if you are talking CPU then, no.

[@melgibson6331]

Ubuntu was the worst choice if you want privacy developing your new product...ubuntu is joogle light but it still gathers data.

[@steveb.548]

The worst offenders these days in Linux are the damn browsers and apps themselves not the O.S. itself, and unlike Microsoft and Screwgoole apps, at least in Ubuntu it's trivial to eliminate all telemetry and data gathering from the O.S. and it won't come back every ten minutes with an update. The biggest issue that I hate in Linux these days is the lack of application level firewall support in the O.S. and that is true of pretty much ALL distro releases. There have been independent projects to add this support (such as Douane), but they don't get any support because the kernel developers are arrogant and stupid, and don't see an issue with applications "phoning home" as long as their precious stupid kernel doesn't get hacked by yet another dumbass zeroday root exploit (of which there have been hundreds, thanks to the kernel being written in C, which is pretty much the worst computer language on Earth when it comes to security).

[@markissboi3583]

👨‍💻i like apple PIE & cream haven't tried a raspberry pie yet 😳 😜👉Oz 🐨straya mate! cheers

[@jhon-cg4rg]

nintendo switch

[@oladunk9986]

The most expensive LED blinker ever ?

[@preston5947]

*EMULATION!!!!!!*

[@ythaloluigui]

its look like a cheap nintendo swith dev kit

[@PyromancerRift]

Maxwell is 5 years old, it's a scam at that price.

[@GaryExplains]

A scam at $99 LOL

[@PyromancerRift]

@@GaryExplains Nvidia makes 60% magin on all their products.

[@CyberAnalyzer]

please record your screen with a proper software. using your camera to record your screen looks ugly.

[@GaryExplains]

What screen recording software for the Jetson Nano would you recommend? Have you tested it?

[@slappymcphee]

This is laughable. How the hell can you call this a review with such little time. If anything this is a Preview.

[@GaryExplains]

eh?