Raspberry Pi 4B vs Jetson Nano

Raspberry Pi 4 vs NVIDIA Jetson Nano comparison, also including benchmarks of the Raspberry Pi 3B and Raspberry Pi 3B+. Video covers hardware specifications, Octane browser benchmark, USB and micro SD card speed tests, KZbin and local video playback, and Kdenlive performance and render test.
You can learn more about the Raspberry Pi 4 Model B on the Raspberry Pi Foundation website here: www.raspberryp...
You can also find information on the Jetson Nano here:
developer.nvid...
My previous NVIDIA Jetson Nano review & demo is here:
• NVIDIA Jetson Nano
My previous Raspberry Pi 4 Model B review & demo is here:
• Raspberry Pi 4 Model B
And I have video called “Jetson Nano Vision Recognition Neural Network Demo” here: • Jetson Nano: Vision Re...
More videos on single board computers and other computing topics can be found on the ExplainingComputers channel: / explainingcomputers
You may also like my other channel, ExplainingTheFuture, at: / explainingthefuture
#RaspberryPi4 #JetsonNano #ExplainingComputers

[@g3ne51s]

I tried the Jetson Nano and the RBP 4B + with tensorflow and the RBP 4B +. RBP 4+ processed 2.4FPS using a type of SSD Mobilnet lite training while the Jetson Nano delivered 22FPS on the same neural networks. The Jetson can run DeepStream with Coffe2 networks at 45FPS and the RBP 4+ only at 1.8FPS for the same model. RaspBerry PI is excellent for everyone, except analytical videos and DeepLearning or Autonomous Robot.

[@ExplainingComputers]

Thanks for sharing your data. :)

[@hhhgggds]

Better info summed up in few sentences than whole video. Ty and gj guy, sry video guy :)

[@ralphjohnticsay8771]

what do you recommend for machine learning?

[@syedsyamilsyedosmanabadi3852]

I was testing tensor flow for object recognition on rpi 4. Now it all makes sense for the fps thanks to you. Thanks bro. Will definitely consider jetson nano after this.

[@amirarjmand5295]

@@ralphjohnticsay8771 absolutely Jetson

[@jjk-9]

The pi4 struggles with youtube because they disabled hardware acceleration as it was breaking other parts of Chromium. Hopefully it's just a driver issue and it will soon be fixed and enabled.

[@ExplainingComputers]

Indeed. In some ways it is a shame that the Pi 4 was released before this software fix, as a board that should be able to play 1080p KZbin not being able to play it is not doing the Pi 4 any favours.

[@jjk-9]

@@ExplainingComputers I've just been checking the raspberry pi forums and they have added a fix that you can update but the -disable-gpu flag is still enabled afterwards so you have to remove that manually and youtube should run better. www.raspberrypi.org/forums/viewtopic.php?f=63&t=244585

[@mergemechanism]

older versions of raspbian as used on the 3b and 3b+ also have wonky video decoding support out of the box, which doesn't help

[@multiflexi]

Just install Play with MPV, compile MPV with RasPi decoding support and play it in an external player. I do that on my desktop too because Chrome does not support NVDEC or any other hardware decoding on Linux. Should play 4K just fine.

[@SonicMastr500s]

Another thing, you CAN enable it. It's simply a line in the chromium configuration file. It's just that after you start playing a video, the entire browser freezes if you click anywhere. It's not exactly ideal

[@randyhall5554]

Dang, I'm just impressed by how much faster the Pi 4 is over the Pi 3+! Thanks for doing this!

[@sethrd999]

Its based off of a totally different ARM core design, A72 was the 2019 ( mid to late ) high end in SBC's before the announcement of A76, ( snapdragon, RK3588 ).

[@adymode]

I never expected them to get 4 A72s on the board as well as USB3 and DDR4 - for 35 quid. Released 9 months early too, with big memory options, dual hdmi, gl3/vulcan, h264 encoder... its unanticipated.

[@randyhall5554]

It's Christmas in July!

[@ravagingwolverine]

@@adymode The quad-core A72 was a little bit of a surprise considering the price, as you say. At the same time, it's not totally unexpected as the RK3399 has been out there for a couple of years. It looks like the A72 cores on a 28nm process is the compromise between price and performance. We've seen such examples as the RK3399, and the old mid-range Snapdragon 650 and 652, and now this. And the A72 being a few generations old already probably helps with the price as it's not top end of the line anymore.

[@mbirth]

@@randyhall5554 Judging by the availability of the 4GB model, Christmas might be delayed until August at least.

[@lamarmottedesinternets5134]

There is only one tech KZbinr around here not talking about ryzen 3rd Gen right now and it's you

[@ExplainingComputers]

I've always rebelled against the trend. :) And next week -- a PC Linux video! :)

[@lamarmottedesinternets5134]

@@ExplainingComputers haha I've seen that, can't wait to check it out

[@iangabrielalcantara7756]

@@ExplainingComputers Nice

[@sethrd999]

Chances are your going to get more Ryzen / Navi tech since it looks like Navi is coming to Arm SoC's, it was going to happen and should shake up the sillyness that is binary blob drivers.

[@ciarfah]

rockapartie Wendell at Level1Techs is quite good

[@RetroSegaDev]

I get the impression there is an increasing need for an official 64 bit version of Raspbian especially with the move to 4GB.

[@ExplainingComputers]

I agree.

[@Btray05Blogspot]

It happened

[@sugaryhull9688]

@@Btray05Blogspot Not quite. It's in beta

[@kooky216]

maybe you could benchmark a pi vs an older desktop cpu like a pentium for fun :)

[@johnsimon8457]

CPU wise - A *pi 3* is roughly a single core of a pentium 4, but you have four cores, so maybe like a multi socket xeon. (I'd give a molar to have one of these while I was in college instead of some sun ultra 10 hardware) *Pi zero* is roughly a pentium II or III. But the ram and storage are much faster than the spinning disks of the time. *Pi 4*, I don't know .... one does NOT see a 3x increase in CPU performance generation to generation in PC world. Maybe it's in par with Intel core 2 or Nehalem That said, anything interesting on the Pi or other SBC is not going to be happening while running a desktop windowing environment, much less running a web browser. Raspian/noobs is for people who are COMPLETELY new to Linux and need something familiar.

[@bbaker6212]

@@johnsimon8457 I think the Kali Arm and Manjaro Arm guys would disagree with nothing interesting happening with a desktop.

[@twmbarlwmstar]

@@johnsimon8457 Thing is I was very much involved in PCs back then, and as Chris says, it is all about systems. It makes the comparison meaningless. In some ways a Pi4 will be more advanced and in others still woefully lacking. Another problem is the fastest computer is the computer you currently own. And NOOBS is for new purchasers so they can install known to work software- someone coming from Windows won't notice any handholding because there isn't really any. In fact I'd say NOOBS adds confusion there. I've read the datas and the X3 thing is just a splash headline, sometimes it is over 3.6! Other times no improvement. If you sort of average it out, mash it about, and apply a sort of lightweight PC use case scenario X3 isn't correct but it isn't misleading either.

[@ExplainingComputers]

@kooky216 Great idea for a future video -- noted!

[@kjrchannel1480]

@@johnsimon8457 In comparison the socket 7 era AMD 500's could do a lot, but could not handle streaming from their older and handicapped instruction sets. I even think the socket A 462 is more powerful on the right board although Firefox so rudely stopped SSE support. I was surprised to see a RPI 1B+ could stream good on OSMC, although it was fun watching it chug and buffer streaming on Rasbian. That was with 1100 over clock. I regard X86 as a fleet of American triple trailer trucks, and Arm as a bunch of delivery drivers on mopeds that can only carry small items. So would it be fair to say that one could take the Arm core frequency and divide by 2? Well I don't think it would be that easy to judge performance without looking at the hardware and software that goes with a Arm cpu.

[@jamierogers294]

KZbin is lit up with Ryzen 3000 reviews and here I am watching your SBC comparison :) This video is great to show that not all SBCs are built the same. These two obviously have different markets in mind and that's cool.

[@jaimemedina3351]

These cards are for two entirely different applications (I love them both). The PI w/4 GB is a great all-around Linux system that's pretty hard to beat for $100 USD. And while the Nano has it's own Ubuntu build, it's primarily for crunching CUDA in parallel, and developing Machine Learning applications. So if you're looking for a great Tensor Flow dev system the Nano is the way to go. Love this channel by the way!

[@RocketLR]

Hello Im from 2020 and let me tell you that the Rpi 4 YT playback performance is still horrible... and that is with the CPU OC to 2.0Ghz and GPU to 700Mhz..

[@SergiuszRoszczyk]

It's a shame that Raspbian is still 32-bit. I understand compatibility and so on but it also means it runs older Thumb instruction set instead of much cleaner, faster and optimized 64-bit ARM set. I hope Pi Foundation will eventually switch to 64-bit. I think Pi4 should get much better result on a 64-bit system. Just for fun I ran Octane on my iPhone XR and got around 40k points.

[@primeroyal7434]

my pc is 29411

[@jonathanlange1339]

I dont belive you that you got 40k. I only got 16.4k with my Oneplus 6 and this is a high end Smartphone from last year.

[@SergiuszRoszczyk]

Jonathan Lange I don't force to belive me. Type "octane 2.0 iphone xr" in Google and look for example to Apple Insider. They also got 40k, while Samsung got 23k. Javascript processing is just well optimized on A12 chip.

[@Chris.Brisson]

It is ironic that inexpensive SBCs marketed for education of software programming are themselves lacking with regard to software optimization.

[@JB52520]

True, but it's a very long way from learning Scratch and basic GPIO to implementing hardware video support. The former is literally child's play, while the latter is something I will never be smart enough to do, no matter how much of my life I burn up trying. The Raspberry Pi can start a poor kid on a journey of obtaining technological skill, even if it drops frames for a little while. Turning down the resolution won't hurt a KZbin video's educational value.

[@fwefhwe4232]

Chris Brisson i think the pi foundation did a good job creating quality boards at such price points. its now upto the community devs to create on top of it. the GPUs are a headache on intel / amd too

[@whickervision742]

I'd love to work on gpu driver code. The problem is that it's all locked up by patents and serious prison time if breached NDAs. No thanks. For that reason, VideoCore can be associated with sluggish, crap performance for all i care.

[@paulothink]

​@@whickervision742 "the patents contain no explicit statement of copyright (as is required for copyright protection in a USA filed patent)" found in github.com/hermanhermitage/videocoreiv/wiki/VideoCore-IV---BCM2835-Overview Does that not mean that either the patent is invalid, or it does not regulate any copyright for this code? And therefore it can be treated as open source / free-to-use&modify type of thing? I wasn't aware of any patent for this graphics driver. I know vivante gpu seems to be the only one at the moment with fully open patent, or royalty free, but that's on the hardware side. @ExplainingComputers what do you think? I would love if someone could hook the full stack of functions of OpenGL4, Vulkan, OpenGLES3.1 and Vulkan Mobile to this driver. Really, I'd be able to teach so many students how to deploy to such a platform which would be great for them, having their first project deployed in a standalone mini computer they can tinker with.

[@PoeLemic]

Thanks for contacting the source (17:50) about the future of the Nano. I'm very interested, and I hope you'll keep us up-to-date on all the nuances of these little boards. Soon, say next generation or so, I think many things that used to be on bigger motherboards will be able to be handled by little devices like this. Even now, they're almost there for most stuff for almost anyone.

[@deldia]

Mr Barnatt is the master of the dramatic pause. Let’s take...a closer look.

[@Flash136]

And I hope to talk to you again...very soon.

[@mtubeluck2044]

@@Flash136 LOL agreed!!

[@NewAgeDIY]

Songs will be sung one Sunday morning about Explaining Computers! A good amount of time spent on testing two SBC’s today. It appears that they are both winnings. I would love to see the 4X4 Pi being used as your main everyday computer and see how it performs. Run it for a week and let us know how you made out. The Raspberry Pi 4B with 4MB of RAM a Raspberry Pi keyboard, mouse and power supply combo is as close as one can get to a complete computer package. They definitely have come a long way to make the Pi a good affordable replacement computer for everyone!

[@ExplainingComputers]

Hi Dale. As you say, both of these boards could be used as a main PC, so a "Pi 4 Week" and "Jetson Nano" week have to be on the cards -- maybe when they both have improved software. And in two weeks time here I hope to have cracked cooling for the Pi 4.

[@NewAgeDIY]

/ yes! Great idea, looking forward to that project!

[@alvallac2171]

Explaining, not expanding.

[@fwefhwe4232]

dont forget the power savings. only 15W max compared to about 200W on a typical desktop. the power savings alone should recover the cost of pi4 in about 2 years. nost office / school / casual browsing work can be done with this.

[@Peter1986C]

@@fwefhwe4232 If power management is doing its job, an average modern PC should not be drawing 200w.

[@adryncharn1910]

After i overclocked my Rpi4 to 2.0 Ghz, its octane score is ~10151, and it plays youtube very well. It also plays browser games without lag as long as my internet can keep up

[@johnintheuk00]

Can we have a comparison video between Mr Scissors and Stanley the knife please??

[@Administrator_O-5]

Chris did you do the RPi 4B firmware update? They released it a couple of days ago. It reduces the temp by 3C-5C & increases speed slightly, especially with USB read & writes.

[@ExplainingComputers]

I have installed the update, but it was not available when I made this video.

[@Administrator_O-5]

@@ExplainingComputers ah, the unfortunate downside of videos. BTW, congratulations on the article in the next edition of MagPi!!! (The RPi 4 B edition at that!

[@kenalex0353]

Your previous videos prompted me to purchase both of these boards in my quest to grasp the basics of the SBC and Linux..

[@redpillsatori3020]

Ken Alexander ..any first impressions on using both?

[@kenalex0353]

@@redpillsatori3020 For me it's all about the Linux learning curve....I can cook book my way through things and have done so on both the PI and Jetson. I'm astounded by the power of these small devices and can see many applications....I'm getting my parts and pieces together to build the JetBot project.....I'm not one to be able to pass judgement on either as I am so new to Linux but the PI seems much more "new user" friendly the Jetson is more intimidating has less avenues open to a beginner...

[@ericturner7744]

I want a tiny computer in the living room behind the TV and a Raspberry Pi 4 would be ideal if they sorted out the issues of playing KZbin 1080p videos in a web browser, which is the main reason that I wont purchase one and it can be fixed as my old £30 Amazon Firestick handles 1080p in a web browser without issues, so should a Pi (and other small boards) in 2019.

[@ExplainingComputers]

You express matters perfectly. Given that (via VLC) the Pi 3B+ could play 1080p KZbin very well (hardware accelerated), they really need to sort this out for the Pi 4.

[@alvallac2171]

*won't

[@fullmetaljacket7]

You should show us your single board computer collection!

[@ExplainingComputers]

Now that would make a good video. Idea noted!

[@goeiecool9999]

There are patched version of chromium floating around that attempt to add h264 acceleration support with VA-API. I would like to see if these work on the raspberries and the Jetson nano.

[@Gentleman1337]

Great video as usual, keep up the good work!

[@philliptoone]

It was just a few days ago that I did a search for "Jetson Nano vs Raspberry Pi 4" and got nothing. Now this appears in my recommended list. Thank you.

[@AnimalFacts]

I indeed found your performance useful

[@whothefoxcares]

WOW! Christopher is overclocked and loaded!!

[@PixelShade]

to be fair, since the nano uses normal sized hdmi and DP connectors. It makes little to no difference which board you choose in terms of price. at least here in Sweden. the RPi 4 is basically 65$, and each micro hdmi dongle/cable cost around 10-20$ so, you'll end up with 75-95$ for the Raspberry Pi 4 4GB, similar to the nano. those connectors were such a bad decision by the pi foundation.

[@joohop]

Imagine Showing This Technology To A Computer Expert In The 1970's. ?

[@ExplainingComputers]

What a thought! They would be staggered.

[@joohop]

HA HA ha haaa YES Earthling Keep Up The Good Work

[@cdl0]

@@ExplainingComputers As a voice from the past, I would say we would be very excited and happy rather than staggered. By the second half of the 1970s there were a variety of development boards available, which were in many ways similar to modern SBCs. For example, if I remember correctly, there was one popular model with an eight-bit Motorola 6800 cpu, about 512 bytes RAM and maybe 1 kB ROM, which was about the size of an A5 sheet of paper or a small table mat. The clock speed was about 1 MHz. People were really quite ingenious in the way they made use of every bit of RAM, and connecting up stuff to ports and interfaces on the boards. The previous generation from the end of the 1960s would have been a large suitcase-sized minicomputer weighing several tens of kg. So, we were well aware of Moore's law and its implications, and fully understood that an incredible technological revolution was in progress. I remember conversations with people at the time trying to explain this, and how exciting the future would be.

[@cdl0]

@Kevin Lee On the earliest computers that I used, we had punched paper tape, and punched or marked (with a soft pencil) cards. During the 1970s I did also see a few larger systems with magnetic tape and disc packs, but these were much less common. The discs were loaded in a machine resembling a large top-loading washing machine. By the second half of the 1970s, is became quite common to use compact cassettes for small systems. Nevertheless, we were aware of the possibilities that the future would likely bring, inspired by the physicist Richard Feynman who said that there is plenty of room at the bottom [APS, Caltech,1959].

[@Mmmm_tea]

interestingly RaspberryPi people said they had doubled the sdcards performance, seems they weren't exaggerating it is almost exactly doubled ! but at those speeds It runs think I would rather run the pi4 using usb3 flash drive if it would support booting

[@ExplainingComputers]

Yes, when we (again) have USB booting on a Pi, it will be great!

[@vidthreenorth4007]

As a photographer I can say that the Micro HDMI has become known as a mistake. The better cameras are starting to leave them behind. I have already broken a Micro HDMI connector on a camera. I would rather have a R Pi 4 series with a single full size HDMI or Mini HDMI, or perhaps a stacked pair of HDMI connectors.

[@ExplainingComputers]

Totally agreed.

[@GateKommand]

I have been subbed to you for around a year now and consider myself very much a lay person where computing is concerned. I find your videos both very interesting an extremely informative, thank you for making them!

[@MichelMorinMontreal]

(But like you, I am impatiently waiting for proposals for cases that will allow adequate cooling to this little marvel.)

[@twmbarlwmstar]

I have too many computers now- or rather I always have had too many but now the small mountain has reached my wife’s attention and might I consider building a few more before buying any more. The Speakers of the Interweb recon the chip is good for at least 2Ghz, but nobody has worked out how yet. Soon as they do and we have ‘skies the limit’ then I will buy and build an oil cooled system. At it’s simplest that is the cheapest type of aquarium, a pump and the Pi4.

[@ashwinmohan4503]

Another awesome vid Chris. Thanks for that! I cant imagine why the Jetson Nano with its dedicated GPU utilizes CPU for decoding video, just doesnt make sense. Maybe some codec issue? I have an Nvidia 1080Ti and when running ANY videos on VLC player, cpu usage drops below 5%.. So, Nvidia CAN decode and has access to the needed codecs. Anyways, I agree with you that the Pi 4B is the better deal! Hope they make the Nano a compelling buy, cos I see some potential in it. Like I was thinking of building a retro gaming portable machine, for which the Jetson would be perfect if the GPU is utilized properly. Nobody makes it clearer than you Chris, thanks again and keem 'em coming :-)

[@edgarmondragon4708]

Because it is intended for Edge Computing/AI and not as an Android Media Player?

[@krukhlis]

That's not NVidia fault or problem. VLC uses its own codecs and these codecs for Jetson Nano are not using CUDA cores for video decoding. Also, I'm just wondering if EC has compiled custom version of VLC, because officially there was no 64 bits version of VLC for Linux on Arm.

[@macchambers3013]

As a user of both platforms, I can say that the RP developer experience is dramatically better and I am looking to hardware accelerators like those from Coral over the Nvidia dev products. It's not about performance, it's about ease of development and agility, and in that case, Nvidia requires a huge amount of buy-in with time and libs. All I really need is something that can accelerate tensor-flow lite or similar nets, for what it's worth.

[@noumanfaheem1928]

hi can you guide among RP 4 , coral dev board and jetson nano which is best in terms of ML inference and ease of use?

[@xmchughs]

@@noumanfaheem1928 I'm not sure what would work best for your particular application, but do you need opengl acceleration or are you working with multiple platforms? My thought above was from the pretty closed developer experience I had with Nvidia and how much specific knowledge I needed to develop with that platform. If you want to use just the jetson products for many things, it's still a really good platform. If you're focused on ML and want to use different libraries like pytorch too, nvidia might be the way to go. I hope that helps.

[@xmchughs]

I would also buy a couple in case one has an issue, like the one I had, where it just didnt boot.

[@noumanfaheem1928]

Thanks for the reply actually I am new to this field and still learning so i dont really know what open gl is but mys situation is I have an inception v3 based tensorflow model already developed for TB detection and I want to deploy it on hardware using real time xray images hope that clarifies my situation

[@QlueDuPlessis]

Banggood has the Jetson Nano for half that price. I think I should buy now.

[@ExplainingComputers]

That sounds like a very good deal. And it is a very good board.

[@jjk-9]

I can only see one for £113 on Banggood as I just had to look and maybe buy one. Shame but I'm not paying that much.

[@raelynnbaranowski5102]

the NVIDIA board has a better GPU?! *shock*

[@largewallofbeans9812]

I mean, why wouldn't it? It was made by NVIDIA, who also makes graphics cards.

[@zxkim8136]

A great set of tests for both boards and they performed pretty good respectivly. Both boards are new and like you say software support is limited at the moment. Fantastic vlog Chris as always mate 😁😁😁 Kim 😁😁😁 it would be interesting to visit both boards in say 3 months and re-run the tests....😀😀😀

[@mattizzle81]

I have both, and I have two Jetson Nanos. Honestly at this point I am disappointed. These little computers are not for everyone who is into mobile projects. After fiddling with these for a bit, they are collecting dust and I have a hard time justifying keeping them. A good high end Smartphone does everything these can and more, in a MUCH more convenient package. I can do tensorflow lite object detection, image classification, all the ML stuff that you can do on these and more on Android. I don't need to fiddle with power supplied and dongles galore to get things like a screen and portable power supply. Smartphones have all that built in. I don't get the hype over these. Price, ok yes, but who doesn't have a phone? Robotics with the IO pins, ok. However if you're not into robotics or trying to make some cheap server, I don't see the advantage. Even when the price is taken into account, it's not worth the hassle and disadvantages of these vs just programming a project for Android. For a TV Box, I use an NVIDIA Shield Android box. It actually has hardware acceleration, where the Jetson Nano despite being the same hardware does not have the same performance for video. I keep trying to think of a justification of why to use these over an Android device for any use case, and robotics is the only thing I can think of.

[@VIPINSAINI_20]

Jetson Nano Cost around 130 USD by Nvidia India.😒😒😒😒😒😒😒 Not Happy😑.

[@yashveerthakur7265]

@Mai Mariarti that's GST .

[@stuartclubb4302]

@Mai Mariarti Nah, India charged him $31 in tax for being in India. He should be grateful he didn't want to buy a Harley, when India put a 100% tax on motorcycles (that was reduced to 50% when Trump got involved). All that "free stuff" in other countries is expensive ;-)

[@VIPINSAINI_20]

@Mai Mariarti Yes Just Being in India.

[@VIPINSAINI_20]

@@yashveerthakur7265 it's not gst.

[@VIPINSAINI_20]

@@stuartclubb4302 You are 100% Right. It's just being in India.😧

[@KingJellyfishII]

Did you add a good heatsink to the pi4? Although it would make the test slightly unfair I'd say that it's a performance benchmark not an out-of-the-box-how-does-it-perform benchmark EDIT: Nevermind I should really watch the whole video before commenting....

[@trixniisama]

You talk about 4k h265 decoding at the start of the video and don't even try that... Everybody knows it'll play fine at 1080p !!

[@johncnorris]

Feature rich SBCs in 2019!

[@Skookumshroom]

I nearly bought a Pi4b today. After reading the comments here, I will wait until everyone is celebrating watching KZbin in full screen. Thank you EC 🙂

[@ExplainingComputers]

You may find interesting my cooling video on the Pi 4 next Sunday . . .

[@wzr3293]

@@ExplainingComputers Yeet it into the freezer

[@iluvrgb]

I knew this was going to happen this video is exciting

[@gaurichauhan6455]

raspberry pi 4b will be the great choice..

[@markdm5415]

I think the lack of 64 bit OS support on the 4b is what is hndering the 4b from winning in all tests. Also worth noting that LivreELEC a common Kodi build, is eliminating support for Nvidia GPUSs

[@strayastray7319]

I am curious about the temperatures of the JetsonNano compared to the Pi 4 during testing. I only heard you mention the Raspberry Pie Temperatures, is this because Jetson Nano doesn't have the same issues with temperatures under load?

[@tberry7348]

Any plans on testing the lan up/down. If I remember right on one of the pi's you showed us how the chipset didn't support full gigabit even though it had a gigabit eathernet port. I would like to see a up/down comparison between 3+, 4, and jetson....

[@rootbeer666]

You all know the true cost of ownership of a Raspberry Pi. It was never $35, you always invested ~$50. With the Pi 4 you will be spending on Type-C adapters or power supplies, those pesky micro-HDMI connectors that nobody else wants to use, a heatsink with a fan, etc. With the 4GB Pi 4 your cost of ownership be $80+, and with Jetson Nano coming down in price below $100 ($95 on Amazon as of the moment), you get the SBC will all standard normal-sized connectors and none of the nonsense of adapters. The cost difference becomes negligible and the Jetson Nano is definitively a better board overall.

[@LuciferGloria]

Amazing video Chris. I recently got the Jetson Nano for 99+4 shipping. And its awesome. Waiting for the Jetson Nano week

[@ExplainingComputers]

I agree with your assessment. The more I use the Jetson Nano, the more impressed I am with it. And if they do open out its target market with some broader CUDA support . . . We are also starting to see some great peripherals, such as those from Geekworm: geekworm.com/collections/nvidia

[@LuciferGloria]

@@ExplainingComputers the peripherals are interesting especially the SSD shield. Can you please do a video about Jetson nano peripherals

[@ExplainingComputers]

We both homed in on the same thing! I will contact Geekworm I think.

[@alvallac2171]

*it's (not possessive)

[@LuciferGloria]

@@alvallac2171 thanks for the grammar lesson

[@lancelotxavier9084]

After 6 months of use. The LINUX distribution on the Jetson is horribly maintained by nVidia. Extremely buggy. Also, many packages are terribly outdated. Highly recommend just using a old PC with a cheap nVidia GPU.

[@nikkytheawesome7556]

Update: I have a Raspberry pi4B 4GB and yt video performance is much smoother. No stutter at all.

[@ExplainingComputers]

Yes, they have made big software improvements since this video was made. :)

[@suomi422]

I have Nano, but I'm not recomending it! It has 4GB ram, yes sure, but 1.3GB is passivelly used by GUI, (rp ia about 200mb), so when you start Keras with TS background, TS will occupy about 2GB and you have only 700MB for deep learning - same as RP3 B+... Even jetpack examples crush sometimes because of ram shortage...

[@MnemonicCarrier]

Can you not do a bare-minimum Linux install without the GUI? Or can you strip down your current install by removing all the GUI packages?

My choice is Rasp. Ever. Jetson for me is Hanna Barbera cartoon :D

[@مقاطعمنوعه-ذ7ش]

Very nice to me with me 👍

[@DavidWilliams-wj4sc]

can you boot from USB 3.0 on the Jetson, unlike the Rpi 4? I'd love to put a m.2 MVME drive on usb 3.0 on one of these.

[@ExplainingComputers]

Sadly not -- both currently lack USB booting. :(

[@RayZXA]

You can boot from USB. I use that and i have no problems. kzbin.info/www/bejne/a2TVg4CoZq6jjdU

[@KaiseruSoze]

This review didn't compare write speeds or LAN speeds. And the RPI 4 fails badly for write speed either locally or across the LAN. The Nano as a server can give a full 1GB download speeds and around 49MBs for uploads. The RPI 4 does fine for download but for an upload it'll run ok for a few seconds and then drops to zero bytes per second and bounces around 0 to 30KB/s until it finishes. Also the bluetooth range on the RPI 4 is really limited to a 2 feet. At least for a bluetooth mouse and keyboard. The Nano with it's add in wifi card does bluetooth mouse and keyboard just fine. (I'm using my Nano as I type with my bluetooth keyboard at about 10 feet from my Nano.

[@Administrator_O-5]

Chris, 1st things 1st. Everyone needs to go to the terminal & type: sudo apt update -y & hit Enter. That will update the repositories from the test to stable Raspbian Buster. Then type: sudo apt full-upgrade -y & hit Enter. This will update everything to the absolute latest stable release of Buster dated July 8th. Then reboot! Now, by adding the overclock lines too the config.txt under: [pi4] over_voltage=4 arm_freq=1750 gpu_freq=600 I was able to get a score of 9005 on Octane. I am using a 4GB, I also have a Pimoroni FanSHIM shop.pimoroni.com/products/fan-shim installed set to auto run at 55C. Everything is running perfectly stable with these settings, but keep in mind I'm using shop.pimoroni.com/products/fan-shim

[@ExplainingComputers]

Good result on the overclock. I will cover cooling options and show results in a video on July 21. I totally agree on keeping the systems updated -- this was done before these tests were run, but inevitably before 8th July, as the video posted on the 7th! :)

[@Administrator_O-5]

@@ExplainingComputers Chris, hope you are doing well my friend! Just an update. I increased the arm_freq=1800 & it still runs completely stable! Again the only cooling I'm using is the Pimoroni FanSHIM & that's it. I'm able to keep it around 58C under load.

[@stapedium]

The core issue is software. The youtube playback on my 1 GB Pi 4 is silky smooth at 1080p when I use LibreElec with the standard KZbin add-on. Hardwired ethernet seems to be another key, kudos for doing this in the video. KZbin video using Raspian Buster with Chromium is painful, especially if you try full screen. I'm not trying to discount the benchmarks in this video. It used the default settings in the most common official distributions. But a little deeper digging (installing LibreElec from NOOBS) can get much better results on the Pi4.

[@LMacNeill]

What an excellent comparison of these two SBCs. Seems like it always comes down to the software support when you're trying to get the best performance out of one of these things. I agree with you that if nVidia gets the drivers working so most (if not all) of the apps on their Linux distro will use the GPU, that will be a gigantic game-changer. That GPU should blow away every other SBC on the market, if it's supported in software correctly.

[@GertBoers]

At 5:20 you say that both OS's are based on Debian. You are somewhat correct ;) as Ubuntu is a derivative of Debian. But the JetPack (or how you write it) is definitely Ubuntu with the Unity DE (or I must be very much mistaken). That also leaves me a bit mixed, because that means it's a very old OS (I think 16.04?), especially compared to the new Raspbian, which is based on Debian Buster, the absolute latest (version 10). Support by Ubuntu for version 16.04 is long overdue, so no security fixes.... Could you run a lsb_release -a on the Jetson, to confirm my suspicion? Rest me to say that I thought this video was a very comprehensive comparison between the boards and it amazed me that the Jetson outperformed the Pi 4B. Well done, sir!

[@viveksark]

Great video. It will be really great to see training performance for a moderately large ML model on both these devices.

[@Adulto357]

128 GPU cores for 99 dollars. What more do you want. Pop 10 of these together and you've got a super computer. Also has built in ai applications. F the pi I'm getting this

[@THERE_IS_NO_DATA]

Bicycle with stick v.s 9th Generation tank

[@richardkelsch3640]

The Jetson Nano requires "GStreamer" to play videos using the GPU. The Raspberry Pi also has OMXPlayer to use its GPU. Both of those were never tried. Your video benchmarks were both CPU only.

[@chefbink61]

As always a great vid with no BS!!! Thanks

[@vasiliynkudryavtsev]

The OpenGL/Vulkan was not tested. Though, it is obvious that Nvidia would dominate, but how much?

[@SamCircuit]

Every other tech youtuber doing rysen zen2 but this channel loved your content subbed!!

[@ExplainingComputers]

Welcome aboard!

[@dominicwoo3372]

Could you please compare Raspberry Pi 4 with SBC equipped with RK3399 CPU? I really hesitate right now which one to buy.

[@mohammadsalman1455]

Wish, You were my professor in College!.. Love from India!.. Long Live!..

[@ExplainingComputers]

Thanks.

[@SuperHaunts]

I agree with your assessment of using the 'stock' OS for each board for comparison. It looks like that pure numberwise, the Pi4 is better for computing, but Graphics the Jetson is better. So if you are monitoring equipment, the Pi is best, but if you are looking for media work, the Jetson would be for you. (At this current time in development)

[@Giblet535]

Very nice comparison. I especially appreciate the 3B and 3B+ results as that provides a broader basis for the faster boards' numbers. I'm interested in seeing how you cool your 4B. I have an old Asus laptop heat pipe on mine. Don't laugh: it works well.

[@HS-eq3gk]

I am hoping they can make a pi4a. would be super interesting what kind of smaller portable projects could be done with it, or straight up make a pi 4 zero...although i dont see that happening :)

[@Rik.B]

Will you post a short update when and if the Jetson Nano gets FFMpeg integration?

[@ExplainingComputers]

I will indeed -- it will open up so many possibilities.

[@mikeblackjr]

@@ExplainingComputers Would this be an ASIC solution?

[@johnny_123b]

I would point out custom gstreamer libs are already prepared for use

[@AbodyRulez]

@@ExplainingComputers Projection mapping possibility for sure!

[@mtubeluck2044]

I want to have a media center PC running Linux. I want it to work out of the box. I am a Linux newbie. I have a 4K display, but nothing to drive it with. (The best I have right now is an Intel based Win 10 laptop that can display HD). I want to watch Netflix, KZbin, etc. in 4K / 30 FPS. This video and a few other videos have been perfect for me, because I am zeroing in on the Raspberry Pi 4B & Jetson Nano. I will show my ignorance now. (Putting on flame suit). So the lack of FFMpeg integration is the cause of the dropped frames Chris was having in this review? When that gets fixed, will either one of these SBCs be able to do smooth 4K video streaming? Also, both SBCs reviewed here claim to have the ability to display 4K video. I guess 1920 X 1080 is the "basic" streaming standard today. The resolution that most people see when watching KZbin, for example. And if a SBC can't handle that, then it certainly can not handle playback in 4K, right? Is that perhaps why Chris did not show any 4K streaming playback tests in this review? Refering to this video: NVIDIA Jetson Nano Review - Tegra X1 Single Board Computer Apr 9, 2019 ETA PRIME kzbin.info/www/bejne/aHvOgqqrlLF5ac0 At 6:21, the author starting talking about using the Jetson Nano as a Desktop computer. He covered streaming video playback. Even though the Jetson Nano is rated as being able to display 4K at 30 FPS & 60 FPS, in the above video the author says: 6:50 - 07:09 "...but with 1080p the Jetson Nano has been amazing at least with this first release. As for 4k video playback online from KZbin, it is a bit choppy, but that could be fixed down the road. This hardware has more than enough power to do it. So for everyday web browsing, 1080p video watching, this little device has been really awesome." 07:24 - 08:10 "...but for now this has been an awesome little board. As for native video playback, 720, 1080, and 4k 30fps have worked really well here. I see a little bit of stutter in this video. I test it on a lot of single board computers. This is the 30 FPS 4k version of Big Buck Bunny. But you got to keep in mind that it's still really early for the Jetson nano. Performance will improve over time. I also tried out the 60fps 4k version of the same video, and it's really not even where it's worth showing at all. It's very very choppy." So what exactly is holding back both of these SBCs? And in particular, the Jetson Nano? Is it the lack of FFMpeg integration? And that is just a software fix, right? It will not need a firmware update or re-write at all, will it?

[@Дмитрийсергеевич-о9п]

эти платы - НЕ ИСПОЛЬЗУЮТ В КАЧЕСТВЕ ДЕСКТОПОВ!!!!! в качестве десктопов их школота использует только из-за своего тупоумия )) И автор канала ту да же.. тесты все для десктопа.. за исключением системных тестов.. куда мир катится..

[@plaguedoct0r]

Huh, so they finally made something that can emulate n64 and ps1, cool.

[@MrHurricaneFloyd]

N64 emulation has always been a software issue anyhow.

[@miquelvp]

a revisit of this battle would be great. Just to check if there are any of the promised optimization improvements

[@leefoster4133]

Well mate. This has been what I have been sort of thinking of. The GPU performance on the Pi's are well, not where it should be if it's used as a desktop or portable for that matter. What I have thought of now is using the Jetson for the CP (Control Processor), video, keyboard and mouse. While I would have 8 RPI 4's working as worker boards in the overall system. The issue is I don't care about backward compatibility for 32 bit code. Really, don't see the point. I can say running the hdparm test on Rasbian and also on Centos for the same model of Pi. Centos won with 27.87 MB/sec vs 25.93 MB/sec on Raspbian. Not a big deal but faster. So, I have been mucking around with buildroot to come up with a Centos like OS where it's running in 64-bit mode. Once up then it's just a matter of keeping up with the patches as I would have to compile and install them. I have done this before when I worked as an network engineer. Just check for patches to code say every quarter or so. But what I have thought is to build a cluster aware system where I would use something wine to run say Kerbal Space Program and see if it could handle the load. One of the problems you mentioned is overheating. I see this on my laptop now running KSP.

[@leefoster4133]

Just ran across lutris so I may have to go with Debian style. What I may have to do is go through wine and see how hard it would be to make it cluster aware.

[@aurelf3158]

again and again ,let s talk about prices /nano cost 140 euros -come on man !with this kind on money ,shureley can improve the mainboard /pi sticks to his range -35 dollars

[@matthewspencer7058]

Thanks for sharing another great video. Was really surprised with the substantial sd card performance on the RPi 4 to the earlier versions.

[@twmbarlwmstar]

What that shows, like you say very clearly, is just what the USB2 bottleneck was doing for performance. Anyone struggling needs to read up on USB specification. The pi3+ has one US2 lane- FOR EVERYTHING! This made it a pretty crap NAS, which the Pi4 solves. But that bottleneck really harmed the Pi3+ because no amount of HATs/add-in boards were going to get around it. One odd NAS I made would slowly fill its SSD, then off-load that data via USB3 HAT much more quickly to a more fully featured NAS. The best use case I found for that design was around a Pi0 CCTV system, with the Pi3 acting as the server.

[@MVVblog]

Simple and effective explanation

[@rfrancoi]

Did they give a timeline of the FFMPEG integration? Would LOVE to see what it can do with Kodi.

[@ExplainingComputers]

No time line I'm afraid, but I get the impression they understand how significant it would be for opening up wider markets.

[@rfrancoi]

@@ExplainingComputers Thank you.

[@BrainToMush]

LibreElec/Kodi. Solved! Quit crying, Everyone!

[@BostjanCadej]

Where did you learn that RPi4 can decode 4K h264? As far I know it can only decode 4K h265. Please reply. Thanks.

[@ExplainingComputers]

Sorry, I think my graphic was perhaps confusing -- 4K in H265 only.

[@jacktheninja]

I feel like the GPIO speed should be compared.

[@nelsontam7365]

Octane 2 6:08 Data Transfer 7:17 KZbin 10:05

[@lifeonloop9578]

Ty

[@dhruvbose8294]

Really not pushing it to the limit with these applications but ok. A good take away was that the 4b is much better than the 3b plus.

[@capiberra4118]

Got one for Christmas! Woo Hoo!

[@200gb3]

Can Raspberry Pi 4B or JetsonNano use USB connected HDD as the external storage? I am interested in using these as servers with 2tb HDD

[@ExplainingComputers]

Yes, you can do this. But any external 3.5" drive will needs its own power supply.

[@BlakeSandenMedia]

Brilliant.. I really appreciate your illustration in comparison to the Nano.. I love your content. My *Wallet does not* .. Thanks

[@davidmalone9865]

Where and how did you get this system info of CPU usage and temperature to show up in the upper right corner of display at 12:20?

[@ExplainingComputers]

They are panel gadgets -- click the panel (menu bar) on the Pi and you can add them in from a list.

[@davidmalone9865]

@@ExplainingComputers Thanks, you're the best.

[@DavidIFernandezMunoz]

Developing a commercial application with SBC for WiFi RADIUS on site, and with mind set-up on the RPI-4. Then you come in and show that I can also use the GPU for machine learning over DNS on the Jetson. The question is... Why do I watch your videos? Still, great nails...

[@stampydragon2739]

We got a really large version of the pi that uses the CPU and GPU from the ps3slim

[@azbesthu]

Nice comparision. As I read RPi4 may have accelerated chromium build later. It is not enabled currently. So maybe the video decoding is done by gpu but putting pixels into the screen is not yet. On pc linux desktop it is also limited now, but here is a link with some hints about how to check if chromium has hw acceleration enabled: www.linuxuprising.com/2018/08/how-to-enable-hardware-accelerated.html Also it would be interesting to hear about AI specific comparision. Nvidia advertises their board as AI development stuff. RPI had few topic about Google's TensorFlow based AI.

[@mergemechanism]

"So maybe the video decoding is done by gpu but putting pixels into the screen is not yet" for a long while the opposite was true on the rpi3, although the new raspbian did include properly built ffmpeg when i checked out the buster repo a few months back so hopefully this is just a chromium end issue

[@leexgx]

It's typically decode and render that is not enabled, so it's been done purely in software witch even with faster cpu on the pi4 will chug On first gen pi I had a full media center running on my pi1 and could handle 1080p30 perfectly fine as long as it was using the correct file and video format for the gpu to do decode (the cpu failed on that one, the cpu had microcracks on the solder as t would sometimes work when pressed on)

[@theoriginalKland]

I mean.. the Jetson is like $100 more than the Pi, so it better outperform it.

[@antonnym214]

One of the best channels on youtube. I love ALL your videos! So informative and well-explained. I am planning to get a pi4 for robotics. I'm excited because it's a nice step up from my pi 2 and FuzeBASIC should run on it just fine.

[@thusi87]

I got interested in the Jetson Nano recently, and came across a few of your videos! Just wanted to say thanks for the great content, and Nvidia should be providing you a royalty or something for the marketing :D (Your video featuring the Bot with Jetson Nano made me immediately buy it!)

[@lacuevadelvampiro]

he def a 2019 guy.

[@BigManT]

From what I can tell If you want to go into AI, go with the jetson nano If you want more processing power for say a website go with the raspberry pi If you want just a logic board go for an arduino Both jetson and pi have the same connections, just jetson has 2 full sized ports and all 3.0 usb

[@oldcowbb]

you usually run an arduino with either the pi or jetson for low level control from my experience, so it's not really fair to put them in the same category

[@ProjectPenguinNetwork]

Did you use any heat sink on the Raspberry Pie 4 for the benchmarks? Since the jettson has a quite big heat sink

[@ExplainingComputers]

I did for the video tests onwards -- see about 12:38 in the video.

[@alvallac2171]

*Jetson