Your video production is superb. Your explainer-in-chief skills are spot on. Its going to be an interesting ride following the future arc of the embedded Jetson line of products. They just get more powerful with every iteration.
Thank you for the unreasonably kind words. The Jetson Orins should be a consolidation point where we get to see some really interesting applications from this new starting point. By the way, me thinks that you are writing this to procrastinate getting your own video out! I'm looking forward to see what you can do with it, and find out your initial thoughts. Thanks for watching!
@@sergeykorol6548 That's relatively good. In my currency Nano is 1500 and Orin Nano is 2600 so 40x AI performance at 1.7x the price is currently not bad. However the old Nano used to cost 600 in my currency.
@@sergeykorol6548 well, you could use a sdcard.. but to be fair I paid $100 in the end to get a working 5v 4A psu for my Nano :D (bought a cheap one first that was supposed to work, but it made the system throttle) and the network card was like $30, and the fan $20..So it was more like $250. This cpu on the new jetson is like a Intel I5 2500k, so I guess it could be used as a dialy driver also.
Thank you for the kind words. I found your video to be fantastic! I found it amazing how good the H.264/265 encoding is working on a couple of CPU cores. Also impressive was that it's relatively straightforward to bring your code base from Xavier to Orin. In the middle of the night yet! Thanks for watching!
Hey Jim, great video as always, though I disagree with "Perfect for Developers and Makers Alike" - because the price in the EU will be not under 600€ - that is simply not affordable to any maker 😢Thus, only companies will buy that product for their product development 🙄
I gave my new assistant ChatGPT one job, name the video. You're telling me it failed? Fired! The price is certainly and interesting topic. If you are a maker and have high computational requirements, there aren't a whole lot of low power options available. Gamers line up around the block to buy GPU cards that cost well over $1000. So the question becomes, "Are the gamers better off financially than makers?" Could be. On the other hand, makers spend many hours creating their projects. If you are making a Star Wars like robot for example, you'll have quite the investment. For the computational requirements, the Jetson Orin Nano may be a very good fit. It's not inexpensive, but it depends on the project. Many maker projects can get along with just $2 microcontrollers. The Jetsons aren't those. Thanks for watching!
That CAN header we need to know more about, that would be a huge plus! Still the pricing does not compare with what we would normally expect, in the past as processors became faster the prices would drop not go up.
It seems interesting. The Xavier NX had the same unpopulated header, I don't know how many people actually use it. Also, I don't know if you need to modify the device tree to get it to work correctly. Thanks for watching!
@@JetsonHacks FIRST Robotics which includes over nine thousand high schools which are required to use a NI RoboRIO for competitions has a CAN bus, and it is able to communicate to the motor controllers, and other devices. FIRST allows a secondary processor(has to be Under $500) for the RoboRIO, and the Jetson would be very popular in handling the ability to identify objects and distances to objects visually using the AI function, which the RoboRIO is not able to do.
@@sy2532 A lot of 1st teams have been using the Jetsons, I believe that NVIDIA supports the series. With the EDU discount, the Orin Nano certainly would be a good choice.
You can plug in a CAN board (like the Waveshare SN65HVD230 CAN Board amzn.to/3KlLlNb). Then you have to configure everything: docs.nvidia.com/jetson/archives/r35.3.1/DeveloperGuide/text/HR/ControllerAreaNetworkCan.html
Must say looking forward to getting my teeth into the Jetson Orin for attaching various block cameras to Orin via the CSI-2 MIPI D-PHY interface by developing a V4L2 driver. Tried integration on RPi but our block cams do YUV4:2:2 and the GPU core of Broadcom chips really only do Bayer RGB... but from working with the Jetson TX2 I do expect YUV4:2:2 in the Orin. Off course it will be interesting to see what libcamera (big on the RPi) is like on the Jetson as that really looks like its getting pushed heavily in the open source community.
Hi, very interesting video ! I want to build a fixed wing that will need to fly to at least 60 km/h maybe a little more . Object recognition will be used so I need a good realtime performance . Do you think the Orin Nano or even NX would be a good option ? On the real time demo you showed , the video wasn't very impressing. Thank you
There are very many object detection models and techniques that you can run on the Jetson. The sample in the video is not the only one. For example, here's an Orin Nano running object detection @ 50fps on a different model: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-np78wgXmYIM.html Your use case is challenging, but it will certainly depend on how you are planning to detect objects. You have the model itself which needs to be trained to find objects of interest. The Jetson is for inferencing, so it will depend on the complexity and layering of your model as to how fast objects are detected. Unless you are planning to display video on your fixed wing while it is flying, you will not need to do the video display and GUI (rectangles, labels) for object detection. That's a major source of time there. You can ask specific questions on the NVIDIA Jetson Orin forums, where a large group of developers and NVIDIA engineers share their experience.Thanks for watching!
Does the voltage input effect the performance in anyway? I'm using one for a project where it has to run headless off battery power but I only have BECs that deliver 5-12.5V. It runs off 12.5V, but I cant find any documentation if lower voltage forces the system to the lower power mode. Any tips?
Excellent content for Jetson newbies like me. I bought my Jetson Orin Nano Devkit, and spent lot of times trying to flash it using several methods: SD card image, SDK Manager with different combination of host and JetPack versions without success. Can you make a video on flashing Jetson Orin Nano or provide the link to the solution here? Thank you.
Hi, great video by the way. I am planning to buy jetson dev kit for my robotic car and in a dilemma right now as to which one to buy. I plan to implement various deep learning projects, so would like to have something powerful and current (jetson nano dev kit very weak and Orin AGX dev kit way too powerful and expensive). Initially I was considering AGX Xavier dev kits, but those cost around $1.3K in amazon these days. Would you say Orin nano can be considered as replacement for Xavier AGX in terms of performance? Do you have any other suggestions for dev kits around $1K? I would like to ask one more opinion: Which is a better investment, buying cheap jetson dev kits (
Thank you for the kind words. The most popular Jetson for the R/C robotic car set is the Jetson Xavier NX, which has just been replaced by this Jetson Orin Nano in the lineup. There's a full ROS stack, and is popular with the F1/10 group. I don't know how big your robotic vehicle is, but if it is scale this is a good balance of compute versus power. The Orin Nano is a capable machine. However, it does only have 8GB of memory which is small in relationship to the AGX Xavier. It depends a lot on how you intend to use your Jetson, and how many simultaneous vision and machine learning models you are using simultaneously. For example, here's a Xavier NX on a 1/10 scale vehicle: ru-vid.comWzQlmUjh9Nc?feature=share&t=596 at DIY Robocar in Oakland, CA. This is indoors at a small track, outdoors the car reaches must higher speed. It has lidar and cameras for sensors. GPS is added for outdoor use. Personally, I would buy the Orin Nano and try to outgrow it. There's a lot to learn beyond just the compute side of these types of vehicles. There are tasks like path planning, sensor fusion, model building and so on. I don't get the feeling that having a lot of compute power on your laptop will help you get over the learning curve when first starting out. These types of projects are definitely the old racers creed, "Speed costs. How fast can you afford to go?" A lot of the the cost is in the amount of knowledge that you need to acquire. For example in the DIY Robocar races, the fastest vehicles are the one first shown in the link and another vehicle later on in the video which costs about 1/5 as much. But it's two real smart guys who know a lot about the space. The video doesn't do it justice, in person the things are scary fast. The F1/10 group is mostly in the university circuit. They're racing this week in London at the ICRA conference. Here's an example: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-BWRscx68Dl0.html These cars use Jetson Xavier NX, as the Orin Nano are just now becoming available. New vehicles will be using the Orin Nanos. Hope this helps.
@@JetsonHacks Thanks for the quick response. This is exactly what I was looking for. I am planning to build upon the F1tenth platform. For your information, i am quite familiar with the robotics field. Recently i graduated with a Masters degree in Robotics with more focus on software side which covered ros, path planning, controls, perception and DL. I am a mechanical guy and likes working on hardware. So was planning to build the robot car and try to implement projects which I haven't covered before. Right now in the process of shortlisting projects and selecting hardware for the car. My plan is to work of couple of projects which may last upto 1-2 years. So I don't want compute power to be the reason for not able to try the project. The 8gb ram about orin nano , worries me a bit. Thanks and have a great day!
@@jeffinkachappilly9708 Difficult question. Depends on which peripherals you want to attach, budget, how much headache you're willing to put up with when upgrading. ConnectTech has a large selection of carrier boards in different configurations, and are usually the professional go to. However, on new releases most of the third party boards lag a little, as they have to provide an updated board support package for the release. If you want the easiest route, you can use the NVIDIA dev kit use the Orin Nano until you outgrow it, and then replace it with an Orin NX module. If you want to skip the Orin Nano part, companies like Waveshare: amzn.to/3C7EdPA and Seeed Studios make dev kit board clones. The dev board design is open source and available on the NVIDIA site, so you don't have to worry too much about the engineering. Some of the signals on the dev kit are relatively high speed, so there is some engineering involved when laying out boards with those features. You can buy them populated or unpopulated with the Jetson module.
Hi, I am a software developer myself and I don't have experience with AI kits like these at all ( I just use my macbook or windows PC for development ). What are these things specifically great for? What kind of projects? And why should I buy one if I own a system with a Ryzen 5950X + RTX 4090 GPU? Thanks for the video.
You are welcome. There are a couple of major use cases for the device. As the name suggests, these are development kits. People use them to integrate Jetson modules into products. Probably the best known is the Nintendo Switch video handheld game device which is based on an earlier generation of Jetson chip. The Jetson provides a large amount of computing with several key advantages. The most important is low power consumption, followed by size and weight. Whereas your system consumes ~750W+ of power, the Orin Nano consumes 7-15W. This makes the Jetson well suited towards a variety of applications such as robotics and smart edge devices. Most of the smart edge devices are computer vision applications. You can usually think about it as devices that run on batteries or remotely. Devices that have a limited power, but need large amounts of compute power. For example a smart "mirror" in a car which does object detection. NVIDIA offers a combination of two Jetson modules on one PCB for Advanced Driver-Assitance Systems (ADAS). ADAS is the fancy name for what we think of as self driving cars. Also, it's a single board computer which means that everything that you need on the system is on the carrier board. Various 3rd parties offer carrier boards in different configurations. Some people will use Jetsons as replacements for Raspberry Pis when they need more compute power in the same physical footprint. There are currently over 1M Jetson developers, so it's difficult to summarize all the uses cases. You can take a look here: resources.nvidia.com/l/en-us-jetson-success-stories for some Jetson success stories. With all that said, there are reasons to buy the Jetson! However, it's not a replacement for the desktop that you are using, nor does it do anything beyond what the desktop does beyond breaking out hardware GPIO pins and such. Thanks for watching!
@@JetsonHacks Wow, thanks a lot for the detailed information! Highly appreciated! I will definitely do more research using the info you provided. I might find a use case for a project of my own since I think it is a great high performance, low-power chip! One more thing: where can I get those Nvidia cups? They look amazing!
You are welcome. I do not know the answer to your Coral question. Please ask on the official NVIDIA Jetson Orin Nano forum where a large group of developers and NVIDIA engineers share their experience. Thanks for watching!
Hi, appologies for this might be a silly question, what is the difference between the Jetson Orin Nano development kit and the Jetson nano edge device in terms of form factor, interfaces, ports, connectors, performance, power consumption etc.
Not quite sure what you're asking. You can look up the specifications for each on for comparison of details. In general, the two are the same form factor. The power requirements are slightly different, the Nano is 5V and the Orin Nano is 9-20V. The Orin Nano is ~20 to 80X faster in compute depending on the task. Thanks for watching!
Hi sir, I couldn't listen the video voice but i watched it. First of all, i apricate what you share, what you do. Thanks a lot. I want to ask you that: Is there any way that i can find your benchmark scripts? To test my own nvidia product. I am waiting for you response sir. Best regards.
Versions of the demo are available from NVIDIA. See: developer.nvidia.com/blog/bringing-generative-ai-to-life-with-jetson/ and: www.jetson-ai-lab.com Thanks for watching!
It's more of a replacement for the Xavier NX than a Nano replacement. The Nano is still being offered. The Orin Nano is ~ 80x faster than the original Nano, so that might be a factor to consider. The original Jetson Nano is now $149, up from the pre-pandemic $100. The Xavier NX was $399, so it's still a jump. However, everything is a little more expensive now. The Xavier NX has not been available for quite some time due to parts shortages. Just being able to be able to buy Dev Kits again feels like a big step forward. Thanks for watching!
@@leonpano Same form factor. The main differences on the carrier board are 22 pin vs 15 pin CSI camera connector, USB-C vs micro-USB connector and the addition of another Key-M slot on the underside of the board. The Orin Nano module replaces the Xavier NX module. Thanks for watching!
Greetings, Jim! Thank you for the interesting video! I have also seen the announcement. Apparently, Orin Nano Dev kit will be available mid-April. A couple of quick questions, if I may: 1) Does it come with a pre-installed operating system? Or we have to install it ourselves using an SD card? 2) I might have missed it, but is there an audio jack (in case if one decides to experiment with speech recognition)? 3) Can it be used as a substitute for a desktop? I was looking at Edge 2 (by Khadas), which is prices at ~$370 in "Pro" version. It's cheaper than Orin Nano, but not too cheap. However, it's the lack of connectivity that really bothers me. Edge 2 doesn't have Gigabit Ethernet, no M.2 slot for SSD, only two USB ports. It is not an easy choice at this price point, but I do not want to "save" money by buying something with a limited utility. How would you approach this dilemma?
The Orin Nano is like the Jetson Nano and Xavier NX. At a minimum you need to have a SD card to boot from, NVIDIA provides a SD card image. As an alternative, you can use the SDK Manager if you want to boot from NVMe SSD or a USB drive. I don't know enough about the Edge 2 to make a recommendation. If your intention is to just use the Orin as a desktop replacement, you should consider buying an inexpensive laptop instead. Thanks for watching!
@@JetsonHacks Greetings, Jim! Thank you very much for answering. I do appreciate that you always find time to interact with your viewers and answer questions. The reason I have asked about SD card is that "Seeed" offers same Orin Nano dev kit, but packaged as a mini-PC. I am referring to reComputer J3011. This model does come with a pre-installed JetPack 5.1.
@@leibaleibovich5806 Right. It uses aOrin Nano production module. That module does not have a SD card. The NVIDIA Developer Kit has a SD card built into the module. The intent for the production model is that it will boot from NVMe.
Would you recommend this "Ultimate Orin Nano," for a beginner project to create a duel brush-less motor controller, with a car alternator as the power supply?
I'm not sure I understand the question. Do you mean to use the Orin Nano as a brushless motor controller (typically called a BLDC)? Or as a device to control a brushless motor controller (like a VESC or an O-Drive)? Is the alternator running from a car or other motor?
I believe that's correct. The RTX 3050 runs in 35-80W, the Jetson Orin Nano in 7-15W. The memory is different, GPDDR5 vs LPDDR5. The Jetson is a little different of course in that it has CPU cores and the rest of the system on the same chip. Also, the memory is shared between the CPU and GPU (unified memory) . The 3050 has to send it back and forth from GPU memory to main main memory. Thanks for watching!
@@theabyss5647 If the laptop suits your needs, that is an alternative worth considering. If you are using the Jetson Orin for the intended uses, such as robotics/smart cameras/drones and so on, a laptop is poor choice. Thanks for watching!
Hi dear, help me. I have a jetson orin nano developer kit and I can't get it to output the signal via HDMI, it has a displayport port and I am using an adapter to display the image on an HDMI monitor. I already recorded the operating system on an SD memory, I already inserted it but when I turn on the monitor there is no signal, is there any other step I should do? Maybe connect a jumper on some of its pins. Thanks
You should check that the DVI adapter is fit all the way into the connector on the Jetson. There should be no further modifications/jumpers that you need assuming that the board is properly powered on. If you have further questions, please ask them on the official NVIDIA Jetson Orin Nano forum, where a large group of developers and NVIDIA engineers share their experience. Thanks for watching!
It certainly costs some money. However, large groups of gamers spend 2x or more than this for high end graphics cards. So I would think there's certainly a group of makers who could afford this device. Thanks for taking the time to comment, and thanks for watching!
You can use a USB camera, like the Logitech C920 (amzn.to/3N6SCB9) or you can use CSI cameras like the ones from Raspberry Pi: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-EuRXAUU61yM.html Thanks for watching!
@@Bwajster Kinda. The Jetson Orin Nano is a module which you mount on a carrier board. The development kit is one kind of carrier board. However, as the name implies, it is a development board. Typically people will use the development board to develop products with commercial carrier boards with cases, power supplies, batteries and so on to meet their application needs.
@@JetsonHacks are you sure, because raspberry pi camera v2 has only 15 pins. and NVIDIA Jetson Orin Nano has 2 MIPI CSI 2 ports (22 pins each). So I guess I have to use a CSI cable to change from 15pins of v2 camera to 22pins of Jetson?
It is my understanding that Nemo is for training large datasets, with a cloud based option. You then take the resulting model and use it for inferencing on a Jetson. Typically people use Riva on the Jetson to run the model. Thanks for watching!
The price is of concern. The Orin is $2000 on Amazon, with much better performance. If the "baby" is already $500, as opposed to the original $200 initially announced, it is showing nVidia is unwilling to cater towards the rapid prototyping scene or the lightweight/low-cost ML-at-edge device category.
I understand your comment about lightweight/low cost. I don't understand the rapid prototyping scene comment. With the amount of software available on the Jetsons (along with all of the machine learning tools in the NVIDIA ecosystem) you should be able to rapidly prototype pretty much anything within reason. Thank you for taking the time to comment, and thanks for watching!
@@JetsonHacks I bought the old 4GB one for $99 last year, and this one if I order the Orin Nano 8GB it is $625 with tax. (25% tax where I live in Europe). For camparsion a new M1 Mac mini is $680 here. I think its a bit too expensive for me, but I would buy instant at $300.
@@AndrewTSq I understand your concern. The M1 Macs are great machines, well worth buying if that fits your use case. Unfortunately they aren't really practical for a lot of ways that people use Jetsons. Different type of application. BTW, I tell the Ferrari people the same thing. If it cost $300, I would buy a Ferrari in an instant! Thanks for watching!
@@JetsonHacks ah sorry, no mac mini for me just had it as an example of price :) The good part is that this Jetson has psu, wifi card, and a fan included, aswell as the nvme ssd port. I will buy one for work, but I want one at home too.
@@AndrewTSq I think calling it 'Nano' confuses folks. It feels like to me more of a Xavier NX Dev Kit replacement. If you look at the price difference ($399 vs $499) of each at introduction, it's a 25% price increase. But at the same time it goes from unobtanium to an actually shipping product. Still, it's not inexpensive. Also, NVIDIA sold the Xavier NX at cost (it was the same price as the production module at the time), this time through it's a little more traditional product pricing. This change will probably make their 3rd party manufacturers a little happier.This product is something the early adopters will most likely enjoy first.
@@emreucar1888 Please ask for help on the official NVIDIA Jetson Orin Nano forums, where a large group of developers and NVIDIA engineers share their experience: forums.developer.nvidia.com/c/agx-autonomous-machines/jetson-embedded-systems/jetson-orin-nano/632
Yes, you need to provide a micro-sd card. I recommend a 64GB to start with. If you are planning on using more storage, I would certainly consider adding NVMe SSD for even better performance. Thanks for watching!
If the SSD is M.2 Key M, it also has to be PCIe (some of them are SATA). Once you have it installed, you can use the SSD as either additional storage, in which case you will need to format the SSD. Or, you can use the NVIDIA flashing tools to set up the Orin Nano to boot from the SSD. When the JetPack 5.1.1 software is released, I'll go over that process. Thanks for watching!
It's not clear what kind of issues you are having from your description. A good place to ask this question is on the official NVIDIA Jetson Forums, where a large group of developers and NVIDIA engineers share their experience: forums.developer.nvidia.com/c/agx-autonomous-machines/jetson-embedded-systems/jetson-agx-xavier/75
After 7 Years of Nintendo with the old Tegra (which is still in Google's Best Media and game player all these years) Do you think Nintendo will upgrade the Switch with a Tegra Orin 8GB Nano? They don't even have to call it anything fancy. Just Next Gen Switch. Or Something like that. Put an upscaler in the dock, make it just a Tad Bigger for those of us with adult hands, and that's basically it. Everything else can be a drop in replacement. This is Literally Better than the gaming PC I was gifted, with is an AMD FX-6300 with an Nvidia 1060 with 8GB of system memory. All Nintendo needs is an upscaler for 1080P and 4K, allow Netflix, Disney Plus, Hulu, and RU-vid, and they have a block buster again for the next 5 years until they use the Full Nvidia Orin Nano 32GB.
If it's coming out, it better be soon if it's for the holiday rush of 2023! There's been a lot of speculation as to if they're even keeping the Tegra processor. I think one of the issues for people "with adult hands" is that the phone game market is pretty stiff competition. Most adults buy phones "anyway" or need an excuse to upgrade. New phone models come out every year, so they're multiple generations away from the Switch by now. It'll be interesting to see what happens. Thanks for watching!
The Orin Nano is Still ahead of All the SOC GPU hardware. It Should even outclass AMD's integrated GPU's. AMD's latest 15 watt APU is just on the bubble of a 1060, with the full Orin Basically a 2060/3050ti.@@JetsonHacks
I am not sure I understand the question. You can program the GPU directly using CUDA supported programming language and tools. The Jetson Generative AI lab (www.jetson-ai-lab.com) has many GPU enabled Docker images specifically for Jetson. Also there is the NVIDIA jetson-containers Github repository ( github.com/dusty-nv/jetson-containers) which holds GPU enabled Docker containers for most of the popular computer vision and machine learning collections. Thanks for watching!
That's one way to look at it, though it's missing the point. People building embedded system for which the Jetsons are intended (robots, smart cameras, battery powered devices like hand held gaming devices) don't have a power budget that can accommodate a RTX card. The Jetson Orin Nano runs in 15W, the full dev kit 36W. A mobile RTX card starts at 115W. An entry level RTX desktop is ~ 275W. The mobile and desktop need the rest of the system of course, CPU, memory, drives and all that. The Jetson would draw ~ 3.5A per hour, a laptop with the RTX would probably draw ~27A. Apples and oranges comparison. If your intent is to play games and you can plug into the wall frequently or have a very large battery, the RTX is a good choice. If you're building a hand held gaming device, for example the Nintendo Switch which uses the Jetson Tegra chip, you would use the Jetson development kits to develop it. Likewise for any number of other devices that have low power consumption requirements such as autonomous vehicles. Thanks for watching!