This is how you destroy Raspberry Pi
HTML-код
- Опубликовано: 15 июн 2024
- LattePanda's Mu is the latest entrant in the 'Pi Killer' battle, but it has a trick up its sleeve.
LattePanda sent me the Mu and carrier board for review, so I'm marking this video as having a 'product placement'-however, they had no input into the video's contents, and have paid nothing for me to talk about their product.
Some of the things I mentioned in this video:
- LattePanda Mu: www.lattepanda.com/lattepanda-mu
- My SBC Reviews GitHub repo: github.com/geerlingguy/sbc-re...
- Radxa Rock 5C: radxa.com/products/rock5/5c/
- Radxa CM5: radxa.com/products/computer-o...
- Turing RK1: turingpi.com/product/turing-rk1/
- Raspberry Pi 5: www.raspberrypi.com/products/...
- ETA PRIME's video on the Mu: • The New LattePanda MU ...
Also, my blog post to go with this video: www.jeffgeerling.com/blog/202...
Support me on Patreon: / geerlingguy
Sponsor me on GitHub: github.com/sponsors/geerlingguy
Merch: redshirtjeff.com
2nd Channel: / @geerlingengineering
3rd Channel: / @level2jeff
Contents:
00:00 - Hardware is (not) the answer
01:17 - Conjoined triangles of success
02:32 - x86 (not Arm)
03:52 - Where's the support?
06:24 - LattePanda Mu
08:20 - A lot of promise (is it enough?) 
Наука
Moore's law isn't dead, it just has a new definition: The number of Pis you accumulate doubles roughly every 2 years
Is that including or excluding the ones in your desk drawer?
My last Pi was a 4GB Pi 4. I looked at a Pi 5, but with HDMI adapters, USB extensions (too close together), M.2 HAT, clock battery, fan and case, I decided to buy a N5105 system. It cost more, but I now run pfSense, Windows 11, Ubuntu all in one box, with plenty of storage and RAM.
is that with or without accounting for the price for a new Pi?
Not really... there seem to be less in stock every year
@@lumeronswiftless is Moore’s 🤓
Great video -- which I watched on an N100 system running Linux. :)
😂
Haha, wonderful!
I have considered making my N100 mini PC my home workbench computer, mostly because the one I've been using there has a fan and is annoying after a few years!
I think we need a closer look at that system :)
So did I watch this on. The n100 with debian is my daily driver. I love the silence. No fan ❤
We would expect nothing less Mr Computers!
"Or put it in your drawer and hope to someday" That felt like a personal attack, Jeff.
I'm in this picture and I don't like it
Mine finally left that drawer and is running full duty as an AdGuard proxy, wireguard VPN and full-time stats visualiser. Nothing glamorous, but I can run a few ssh windows and keep tcpdump scrolling by and, bmon and htop doing their thing.
I feel called out
"Until next time I'm Jeff Geerling". Drops the next video, still Jeff Geerling. False advertising.
I've always wondered why he doesn't say it the other way around 😂
Can you legally punish your own children by forcing them to build your kernels without pay on underpowered SBCs?
Yes
Why else would you have children? lol
It's not punishment, it's education
“C’mere kiddo, I’ve got an Ethernet driver problem for you to figure out. It’ll be educational!” 😂 I’ve got a whole pile of “educational opportunities” in my basement but I want my kid to LIKE computers and not despise them-which is why her projects get Raspberry Pi’s.
The Geneva Convention would say no
At $185 for the LattePanda MU, forget it. I bought 3 GMKtec NucBox G3 Mini PC's with the N100, 8GB of RAM each, 256GB M.2 NVMe, 2.5GB Ethernet/Wifi 6E.. for $120 each. And the RAM/SSD is upgradable. And there is an extra M.2 slot to run another SSD or external GPU. And you get a Windows 11 Pro license. And you can get that price all day on Amazon and have them the next day.
well the MU is not a home pc, is more like a maker unit, the idea is for you to have all sorts of I/O to make custom devices on your own. If you just want a box with a browser, IDE, and some low-end games, then don't look at the Latte, or any sbc tbh
@@ezforsaken I'm not using them as HomePC's. I'm using them as server nodes in a cluster. They have functionality beyond being used as low end desktop PC's. Outside of a few slim use cases for things like embedded engineering that would require a powerful but efficent processor like the n100 in a tiny package, I'm not seeing a lot of advantages to paying $185 bare minimum for a LattePanda. Most projects are going to be fine with the much cheaper Compute Module 4 anyways as far as embedded goes.
@@ezforsaken I agree, this is technically the original point of the SBC. Which is what makes me wonder why this one has barelly any I/O. I mean, just like Op mentioned, for just that might as well get cheap laptops instead.
And for just I2C and UART in this one carrier board, it ain't worth it. Might as well make my own IO expansion card with USB through a micro or through PCIe and it'd still be cheaper.
these are SOMs and are meant for easy embedded device integration. If all you need is a mini PC, just buy a mini PC
@@leozetalol the Intel cpu they are using does not have GPIO since it's a laptop part.
I never understand why reviewers of SBC's keep saying statements X-times faster than RPi. The litmus test is always "How many products from this vendor still get support 6 months down the line?" In most cases a lot of the SBC's are built around SOC's meant for a specific Android version, and good luck getting long term Linux support. To be honest, I don't care how fast SBC's are. They are not about speed. I couldn't have agreed more with you during this video!
"I don't care how fast SBC's are"
lol then why get anything other then the original raspberry pi? As someone that emulates retro I 100% care about processing power.
"good luck getting long term support."
so... it's like everything else in the IT industry?
@@jspringer86 long term support matters for a lot of people though, and whether this is the case or not in rest of the IT industry is irrelevant to that. Even the current Raspberry Pi OS supports the original Raspberry Pi, while you simply don't have this kind of support anywhere else.
@@jspringer86 because it's much more than an emulation station. And while I don't care might be an exagerated statement, I know that if I'm building something based on Linux, going from Rpi 3 -> 4 -> 5 was just a package and firmware update away. And I never had to deal with the worry of sticking with ancient modules, because otherwise I would end up driver less.
Yeah, you do get third-party mainling efforts for both allwinner and rockchip.
The thing about rockchip/allwinner is that it's extremely cheap, so iterating on prototypes and the overall barrier to entry is a lot lot lower.
Not sure about software side, but price wise it is like 8-15$/25$ for 1QTY(not even for 10-100 QTY order is required).
@@jspringer86 You don't always get the on-paper performance with the less-supported boards, it's practically a lie. I try to stick to RockChip stuff as it tends to get mainlined and get good community support for drivers, making it age like wine instead of becoming e-waste.
Intel based SBCs' main advantage for me is the video encoder which makes it perfect for a media server
There are plenty of arm chips with good video encoders (basically every network security camera has to have one), but as mentioned in the video, the problem is software support.
@@nezu_cc good luck getting plex to do 4k HDR full blast transcoding with an arm chip
@@nezu_cc In some cases, I wouldn't necessarily blame the software, because in some cases, even the Pi, they use proprietary firmware blobs that may only work with some versions of Linux. The actual encoding and decoding support has gotten much better in Linux, but if you can't use the company's drivers, then that suddenly makes it a whole lot harder to use the hardware.
why you buy SBC for that, there are so many mini-pcs that can do that
It's not even about "destroying" Raspberry Pi (Jeff needs to stop outsourcing the writing to Redshirt). There is clearly room for a number 2 in the market (an Avis to Rasberry's Hertz, in rental car terms, or, if you like a Lyft to Rasberry's Uber), and it's been very clear what that takes - dedicated support. And a strong number 2 in the market would only make Raspberry better - it would make them work that much harder.
With apologies to Jeff, this has been obvious for years. What's crazy is that as obvious as it is, no manufacturer has taken up the challenge.
I think it's worth thinking deeper about what that would mean - maybe challenge Jeff to do that, because his knowledge on this far outclasses mine, for sure. Would that mean picking the top five or six Linux distros and ensuring they run on new SBCs out of the box? Guarantee five years of support?
What's the kind of minimum viable level of support that would start to separate the sheep from the goats in the non-Pi SBC market?
Another thing I've wondered - is it possible to have a Raspberry clone? Or a near clone? Is that a viable path for a second player to emerge?
Honestly any of these manufacturers could start by waiting to ship a product until (a) it's available more globally, not just through one or two Asia-based resellers who take 1-2 months to ship, and (b) the firmware/main Linux OS is actually somewhat stable, meaning all the top advertised features and interfaces work as advertised.
Just doing that would be beyond any of the current offerings, and once you offend a new customer, good luck getting them to consider you again!
Libre Computer might be the best at that, right now (based on my limited use of their products), but they are a smaller team, and don't have the most compelling hardware (unlike some of the rockchip board vendors).
tbh you can probably just buy an e waste pc in ebay and it'd beat the pi
@@clementpoon120 not in all aspects, mainly efficiency. If you run it in your house as a server it may be important for you, but indeed, this route (e waste PC) is what a lot of people is doing.
@@clementpoon120 An e-waste PC probably doesn't have GPIO headers or fit in an embedded application. Old, second hand PCs are great if you just want a cheap desktop or server, but if you want to do almost anything else that you can do with a Pi, they're completely useless.
Back in 2013 I built displays for a interactive exhibition. We had 12 32" TVs and a 42" touchscreen built into a tabletop. The TVs where close to the exhibits showing how they where put together and what they do and point the "very special handheld terminals" - i.E. cheap chinese phones -- to the internal website with even more information. While the Raspberry Pi B was good enough to play the videos and function as an iBeacon, the Touchscreen Tabletop needed a little more oomph for the CSI effect I had in mind. So I bought a Cubietruck and a Radxa Rock and startet setting them up in Kiosk Mode ...
....
....
and ended up buying an Intel NUC with i3.
Since then I hate the A20 SoC in the Cubietruck and the Mali GPU in both with a passion! I won't use anything without Raspberry Pi levels of Linux support ever again, my life is to short to bother with the others.
I used a rpi4 as my primary computer for two or so years, and from my experience, it's software support that matters most for medium-high end SBCs. I saw significant improvement in usability on the pi4 over the two or so years, and I'd hate to have to go through that rough starting phase again.
"Temperatures matter to a degree" Oh you rapscallion you
It’s a heated debate😂
“HOW DO YOU DESTROY A RASPBERRY PI??”
“You forgot to cherish it”
Oh, I misread the title. I thought it was “This is how you destroy A Raspberry Pi”. I was thinking, “What’s so hard about that?”.
Heh, not quite *that* episode!
I wrote mobile software in the early 00's for Qualcomm and later Kyocera Wireless (who bought out Qualcomm's handset division). Back in those days getting ahold of an ARM compiler was a PITA. They were SUPER expensive and single seat licensed - they were basically node locked. So once you installed it somewhere, it was a PITA to get it installed somewhere else. Say you got a new machine or whatever - yeah, you can't run the ARM compiler you payed over $1k of year 2000 dollars for. There were some enterprising groups working GCC ports, but they weren't quite there yet and didn't produce the same optimized code as the ARM compiler (back in those days on those handsets saving even 100 bytes of RAM or NAND storage space was paramount). The landscape is way different now - the ARM port for GCC is very mature and I don't even know if they still sell an ARM compiler since GCC produces pretty well optimized code now (plus we don't have the same resource constraints we did 20 years ago).
In the Acorn world, the reliance on the Norcroft compiler probably contributed to the demise of the company. As C++ became more widely adopted, all Acorn was able to offer was a version of Cfront, and so they failed to deliver competitive libraries and frameworks for their platform. Then again, the company seemed to accumulate and attract people with this mindset that BASIC and assembly language was all anyone ever needed. I think that the compiler on Acorn's Unix was just the Norcroft compiler, rather than something with Unix heritage.
There was also a C compiler from a third party for RISC OS that was reasonable, but probably not great, and this was updated to provide C++ support that, as far as I remember, was ambitious but flawed. Its developer took a job in the States for a well-known corporation and never got round to finishing the product. It does surprise me that support in GCC took a while to mature, but I think that people forget that ARM wasn't this "dead cert" architecture worth spending colossal amounts of time and money on until a while later. That will have left that lucrative, exploitative niche you mention.
@@paul_boddie Very interesting insider information. All I know is we had to write firmware for BREW handsets, and the only commercial compiler I recall being available at the time was actually from ARM. I still have the box in storage somewhere ;). We were trying to cut costs, and I did some research for open source alternatives which led me down the GCC path. They did have a cross compiler for GCC at the time, but the elf size was way larger than what the commercial compiler was spitting out. In some cases it was just too big and it wouldn't fit on the handsets at the time. Other times the code was just so under optimized that it ran like crap compared to the code generated from the commercial compiler. It was definitely a hobbyists toolset at the time. I'm really happy to see it matured. I truly wonder how much Raspberry Pi had to do with that? Plus the resurgence and wide acceptance of *nix OS. It was a great way for a neophyte to get their feet wet without too much worry about ruining things. Oh, you just did rm / -rf? Re-flash and you are good to go ;)
Good to see that you're making content on some more boards for my birthday this year my wife bought me a new SBC it happens to be a x86 board as well from Radxa. The X2l has proven to be an awesome addition to my workforce of tech. So much so I have begun to daily drive it as well. This SOM platform is definitely singing my song and would love to purchase one soon. Thanks for the awesome content as always.
Taking the company public and getting that free market hand involved is probably how you destroy Raspberry Pi
yeah the raspberry is for the people (soviet union music in the background)
It is a for profit company currently (Raspberry Pi Ltd) - the hand of the free market is what build the company up to today.
@@TylerDurden-pk5km There is a big difference between a private company under the umbrella of a non-profit organisation and the entire thing being a publicly traded stock.
First they will tell you that nothing changes.
Then the current leadership retires because they made a bunch of money from the IPO, then price increases and reduction in functionality because the stockholders want more revenue, then the inevitable personell cuts to "make the company more lean and efficient" and finally the company being a shell of its former self broken apart sold to the highest bidder so they can make trash under the RPi name.
@@bami2 That is not wrong - but that likely always was the plan. The foundation was used to generate goodwill / good looks at the early phases of the endeavor. Now it is payday for the effort - and investors.
But if the Raspberry PI Foundation has a 51% stake then that’s not so likely to happen
Excellent video! Interesting comparison and this video very well articulates a lot of the concerns I've had about non-Pi SBCs
i got called out with the "put it in a drawer and hope to use it someday"
Well said Jeff. You can pretty much build anything with a RPi w/o building boards because there is so much 3rd party support and product availability. That lets you focus on what you’re building. I am doing that right now with a project. It will be a bit more expensive than I want and complete in
I was all ready to make a comment about "conjoined triangles of success" but Geerling beat me to it. Foiled again!
Related: Triangle of Sadness. Great movie.
The only limitation for the N100 is the 16G of RAM limit. I have one of those cheap microITX N100 mainboards ($129(ish)), in a NAS case running on M.2 drives, Proxmox hosting pfSense, PiHole, a few more smaller VM's and a bunch of containers, and a NAS (passthrough for 4 x 2T SSD's)... power draw hovers at only 27w ! The N100 may be a real contender, a lot of punch for smaller power reqs. I'd like to see what the N100 SoM can do.
I really hope Intel doubles down on the next N100-class chip, too-like, the pressure from Arm (and eventually maybe RISC-V) on the low end could make Intel's low power chips that much better.
I've seen many people report that the N100 can handle at least 32GB of RAM on a single stick.
I have 2 n100 miniPC, both run fine with 32gb ram.
One is ddr5 sodimm 5200, other is ddr4 sodimm 2666
@@nich98 I'll have to pick one up to try... the spec sheet said max 16G sodimm, but I'll be MUCH happier if it can use 32...
okay, but if it's competing with the Pi it really doesn't need more than 16gb
$155 - $180 is the cost of a 4GB Pi 5. A latte panda is cheaper in Australia, at least.
Woah! Is there no official reseller in Australia?
@@JeffGeerling Australia is the land of the expensive. It's why my friend, RUclipsr Tech YES City, does so well selling gaming PCs there I reckon.
@@JeffGeerling There is a retailer called LittleBird, who also run a secondary Pi-centric retailer called PiAustralia, but I presume since our dollar is weak against the USD, it's just plain expensive to get decently-priced Pi's over here.
Core electronics in Sydney. Pi5 8gb $134.50. Aussie dollars. Always in stock. Bruh, how doesn’t everyone know core electronics…
That is my point. The Pi5 came to late and with bad specs and there is no Pi500 nor a Pi6. The mini PCs with server spec grade hardware (64GB, 48 Cores, GPU option) running as full feature blown homeserver - doing all that stuff we did 10+ years ago with our 2-3 Pis - that is the new reality. The ESP32 that recently came alive - that can even boot a mini linux - will also chew on the Pi-Universe. For me - no Pi5 - (I would consider getting a Pi500).
Very slick intro, I read the title and it was like you were talking to me personally!
Some Great observations, Jeff. Thanks for sharing!
That subtle Silicon Valley joke. Well played🎉
Jeff, you are absolutely killing it. Your delivery in this is 2nd to none. You’re a true pro.
I'm glad to see non-RPi videos. I've kinda been tired of RPi of late.
The N100 and any other cheap x86 processors that might come out have been far more interesting to me anymore.
Agree! Even for firmware wizards its nice to have something working out of the box and not have to fiddle around.
Unless fiddeling around is the project, usually you'd rather spend time one the actual project :)
I'm a big fan of the Latte Panda stuff. I used the V1 to run a LABView interface with the on board Leonardo performing the real time functions in a prototype fuel cell project I worked on with a former employer. Excellent value and an incredibly fast way to get up and running.
I'm subbin'! This guy is credible and has great cadence and communications skills for this subject matter. Thanks man!
I feel like you nailed it in the begining on the price/support situation. To pay $149, even without considering a carrier board, I have to be very wedded to a SBC form factor because I can already slap an old PC on the network for that. I can see paying a premium for the form factor if it really matters.
I hope the support for this gets really good, I've been looking for a good n100 to build a NAS and this one seems promising, though I'll wait a couple more months
Agreed. Raspberry Pi is, by far, the most solidly supported of the SBCs. More than any other they come closest to "just working". Orange Pi can do "ok" though the official images are a bit weirdly hacked and they're stuck on kernel 5.10 short of doing a bunch of patching and fiddling to get newer kernels working. (Note: I only have a OPi 5 Plus - Maybe Armbian does better out of the box with other OPi boards). Another I've found that does quite well is StarFive VisionFive2. Ubuntu has a port specifically for VF2, good directions for how to get it working, and uses a kernel based on current stock Ubuntu (patched as needed for VF2). VisionFive2 performance does - Not formally tested - Seem to be somewhere around Raspberry Pi 3/3+ level. Drawback is VF2 costs more than a Raspberry Pi 5... But its the non-x86 option I have that I'd recommend as most 'complete'/'usable' for someone not wanting to go Raspberry Pi.
Milk-V also has the Mars, and there are a couple other interesting RISC-V boards I'm eyeing that should launch later this year or 2025 that could be interesting.
Still, comes back to support-Milk-V, StarFive, and other RISC-V vendors are still young and proving themselves, but the RISC-V community has done a pretty awesome job so far, getting much further with distro support than Arm had at this stage in the game back when Arm was newer.
It's not just "that things will work" with a RaspberryPI, it's that your one Pi can be repurposed over time. I have Pi Zero-W's that were home cameras, but are now home internet radios, and other Pi's that have been repurposed. Other manufacturers drop support for their products when they realise that their product just isn't that big a success. They are desperate to take market share from the Raspberry Pi's, but don't want the expense or commitment of the long term support.
The Pi's are a truly reusable components, and that makes them great value.
Your value assessment is spot-on
Hey Geerling, could you please make a video/video series on cameras? More specifically what to get for photography, videography, and both, as well as the basics of operating them for the above mentioned purposes? It would also be nice if you have your recommendations. Also, please include what you would recommend for filming/photographing eclipse.
I read "destroy Raspberry Pi" and envisioned seeing some magic smoke. Closest to this was at 3:22.
Wasn't disappointed, still a very informative and entertaining video. :)
very well said. I remember fighting a Banana Pi, trying to get it to boot off the SPI flash only to find out it only boots from bus 0 which is not exposed anywhere. It would have been nice to have documentation that said so.
I think I hear rumblings of Red Shirt Jeff!
The support really is what got me to suck it up and just buy the Pi5. I knew there was more powerful boards out there, but no matter how painful it might have been to be an early adopter I knew it would smooth out relatively quickly. Some of the boards I was looking at had been out for quite a while and the first things I found when looking them up were people still trying to get past issues that had been there since the board's launch.
I don't agree with some of the decisions the RPi Foundation have made lately (and are going to make) and I'm not likely to buy another Pi5 unless this one breaks, but you can't argue with massive amount of support you'll get from either the foundation or the community if you decide to buy one. I'd love to see a company wedge itself in that gap they've made though. Wedge themselves in and pry it open, it'd benefit everyone.
As a casual hobbyist, support matters. I have a 2 Pi 0's and 2 Pi 4's performing different functions (pump monitor, web server, file server, dashboard), and all of them have been running for years (still regret getting rid of 2 Pi 3's for next to nothing). Hardware is reliable, cheap, and the broad support base keeps me attached to these.
Thanks for the video, did Lattepanda incorporate a fix for the NVME problem ?
The N100 (N200? N97?) have turned up as the brain in quite a few handheld, or Ultra Mobile PCs lately. They come so close to what I want, yet always manage to get something "not quite right".
This SOM looks good, for example - until you have to mount it on another board. That puts a limit on your layout options, right out of the gate.
I could've told you that.
In fact, I did, in one of your other "pi killer" videos.
It really does come down to community and support. As I pointed out in those comments, in one of your other videos. Glad you paid attention!
I love this channel! Thanks for the video.
Love your work. I feel your passion. Thanks 🙏
If I remember right, a lot of Chromebooks run on the n100
there are a bunch of cheapo NUC-alikes that are pretty cheap too.
In some cases it's even older than it. There are still many new chromebooks using N4020, a 5 year old celeron chip
@@sihamhamda47 The N5005 and 5010 seem to be taking over, at least in my local market. I saw one a while ago with an i3 1305U as well, a 5-core CPU!
@@DigitalJedi Aren't those rebrands of inferior older CPU's? N100 with DDR5 seems the best Minipc under 200 dollar
@erikkarsies4851 The N5005 and N5010 are Jasper Lake parts. They're little 10nm dies. N4000 series is Gemini Lake Refresh, which are the rebadged versions of older Gemini Lake.
N100 is of course better than any of these. We were not discussed which is the best, but which we see in the wild most often.
It's incredibly important to understand that raw power and low price are not everything! With proper support you get tested and working devices out of the box without any additional tinkering needed. While this might not be always what you are looking for it's really easy to underestimate the impact of support when dealing with SBC class devices.
Excellent video ! 🎉 bravo ! More like this please ❤
This was one of the hardware products I immediately ordered when I got the announcement email. Having an Intel x64 SOM is just too cool of a prospect for DIY projects.
Congratulations Jeff for the analysis. Recalling that the Raspbarry pi project was created to promote teaching in basic Computer Science in schools, allowing the democratization of knowledge.
It turns out that several manufacturers began to fight a battle for performance and focused on other objectives.
"It's hard but it's not complicated": I'm in love with the thought and experience behind the statement here... Industry and academic experience on my part concurs.
Thanks, Jeff, another entertaining and educational video. You have said what we have all thought for the last few years; when we buy any computer (or anything), there is an assumption that it just works. Raspberry Pi seems to be the only player in the market that makes that a priority. And who really has time to fix a product that you just purchased? Simple, yet profound thinking.
I would LOVE to see how this cluster will perform.
I think this review nailed it. Absolute performance is not primary, cost is meaningful (especially for home / school projects) but comprehensive and accessible support means that projects by the 'unskilled' and 'naiivly adventurous' just work. My night time nature cam powered by a Zero would be the perfect example.
PI support is a major plus in my book too. However I2S input is a hardware feature that has no support in current versions of the official OS, I would guess there are other unfixed issues that can be fixed with older versions, but I kind of like to work with the latest version.
Lattepanda is great and all but question is if it does not have GPIO why not use generic N100 miniPC. I think it is same as running homelab on RPI. It is great, but offers no advantage compared to any small miniPC with good power consumption. Most of the time those miniPCs are cheaper and have same power consumption.
That's actually quite legit point.
the carrier board has GPIO, but yeah if size isn't a constraint, you can get cards that expose boatloads of GPIO that plug into a PCIe socket for peanuts in the right places.
Generic minipcs and used Dells have always been an option if you want a cheap pc, but sometimes people actually do want a SBC or a compute module to build into a robot or handheld machine or something and for x86 LattePanda is one of the few doing this
@@SussyBaka-nx4ge Yeah, but for those sorts of uses, who cares whether it's x86 or ARM anyway? I mean, sure, the N100 is cool, and it does offer pretty good price/performance/power ratios for certain types of applications, but the fact that it's x86 is pretty much completely irrelevant in this application, IMHO...
(Really, the only place nowadays where x86 potentially matters at all is if you're building a (Windows) desktop system for running arbitrary applications, in which case other mini-PC options are probably going to give you better bang for your buck anyway.)
@@foogod4237 Very few ARM boards work with mainline Linux and very few distros ship images for anything other than RasPi and maybe cloud as there's no real generic platform on ARM, so you're stuck with whatever ancient build of Debian or AOSP got shipped with the board.
For x86, it's just a PC and Windows or any Linux distro will just work.
Jeff, have you tried or covered the VMWare ESXi ARM Fling on Raspberry Pi? I'm testing it out this weekend on my Pi 4 to boot Windows 10.
The Pi is low power and does not need a fan.
So many people review a "Raspberry PI killer" that is similar price and faster, however it requires a fan and/or sucks up more wattage which makes it non-comparable for me.
I would posit this: All of these are exceptionally good deals. What I would challenge any of you to try is to make a simple PCB, no components, and send your gerber files out to a board house. You'd be hard pressed to get one board made under $100. That is without the cost of the components and manufacturing. Here is the REAL...REAL kicker. The cost of software development is HUGE. The fact that the pis are open source gives you opportunities you'd have NEVER had before. Support is underrated.
I absolutely agree with you, but I think we need to be careful when we consider power efficiency.
You're absolutely right in your video, efficiency is computational power over energy consumption, but we also have to consider those values in relation to each specific user scenario, because each scenario have different needs.
I'll give you an example, if you need a home server for hosting services, make backups and so on, all the power an N100 can give you is completely useless, an RPi4 is more than enough, for someone RPi5 may also be a little overpowered; on the other hand that's a user case where uptime is usually 24/7 and absolute power consumption does really matter (in some places more than others, think about EU vs US) and there's a significant difference in power consumption between RPi4/5 and N100.
It'd be neat if you tried the cognatek mRLP. It can be configured to use a 6p8e 1370PRE and fits on a 3.5" SBC. I think its super neat for tiny builds but it is pretty pricey...
conjoined triangle of sbc value HAHAHAHAHAHA
love the SV ref!
For device production you also want to be sure that you can get hold of said SoM / SoC in quantities without having to wait eons and then the line is discontinued. This is why our product based on the RPi 3B+ is still selling and is still cheap and easy to maintain. Even the nVidia Jetson SoM's are really expensive, not that available in bulk and the docs are not that great really. Good thing nVidia's forum has some great mods or I'd ditch them in a heartbeat!
How's the power draw on the Mu? We usually build around ARM boards because embedded SoC power draw is a major concern, and debating whether it's worth looking at x86 (cpu-intensive applications) when it may exceed the power supply in our remote terminal use cases.
Found myself atm best bang for the buck for my specific use case (diy-nas) was Odroid m1. Had xu4 cloudshell but needed more hard-drives. That's where m1 nvme shines over alternatives. The hardware is also more than capable to handle gigabit networking and smaller software raid.
I'm intrigued by that little x86 board. I'd like to build something like a Zynthian (Pi-based synthesizer/sequencer workstation) with an X86 CPU to cope with some VSTs that I like a lot.
"Cheesy value triangle" sounds like it should be Kraft marketing material :D
Spot on, Jeff.👍
RPi was my 1st Linux computer. Now, I have every model, except for 5.
ROCK Pi 4b's impressive specs was a disaster, due to old kernels & lack of support. Radxa? Never EVER again.
I found the same with other SMBs, until I purchased the Latte Panda 3 Delta for my planned Linux laser controller.
But, the software I needed only ran on Windows. Fortunately, the Panda 3 Delta's x86 didn't mind, which saved the day. Mission accomplished.
BTW, it runs better than my Beelink miniPC, which refuses an SSD upgrade and is stuck @ 500GB..
The easy road for the support corner is to make your thing fully compatible with existing Pi accessories like carrier boards and etc, just like any successful Arduino clones all have the same "hat" connectors so you can just about slap on anyone else's "hat" and move along.
I have been thinking the same here. I built the little pi Nas in previous videos. I accidentally order the wrong top for the Rock 3c and not the Pi. So I ordered a Rock 3c and a Pi Hat. Good thing these are pretty inexpensive. The Pi build was simple. The Rock 3C I have struggled with it and the documentation or forums are just not the volume that the pi has. I like the Radax Product line; I just hope more get involved or linux just gets better with these SBC's.
You and me both! Radxa has some great hardware, I just think they spread the software side too thin with so many new products
I'm so glad you guys get cheap Pis, but here in Australia, they're very hard to justify when it's $100+ to get one up and running.
The Dalek Poster...Epic...
It seems you can only purchase the thing as a $190 kit -- at least, when you click the 'buy' link on the Latte Panda site, that's what you get with no obvious way to purchase just the compute module -- that's still a fair comparison to the PI5 as the latter includes gpio, hdmi, usb etc.. I would tend to use this as a small project/appliance machine similar to the Intel NUC or the old ASUS eepc compact machines.
They offer it for sale separately, too (just the module for $139), but you have to search on the store pages to get it. I'm guessing they push everyone to the kit since very few people would have any other way of using the Mu without the official carrierboard.
As they were laying there side by side, it seemed to me that the pin layout is the same as the Turing RK1, so am I right to assume that it at least physically fits into the Turing Pi2 cluster board? The next question then becomes: what is needed to make it actually work?
8:05 Are you going to do a Windows Kernel compilation video ?
I'm kidding, but there's that Windows2003 leaked kernel everyone has
Haha!
The Radxa X2L seems like more of a Pi killer. At $49 it’s the cheaper than a Pi 5 except x86 so compatible with so much more, actually takes USB PD for power instead of nonstandard 5V5A, and has built in M.2 support without needing some extra hat.
I am really hopeful here, the N100 is a very nice little spot, though it might be a dead end as well. There's a real space to make an awesome 'upper range' board out of this thing.
i've been looking at the N305 cpu to replace my low-power opteron x3216 NAS/server with. Intel's E-cores are fine for general compute so all they need to do is put a whole bunch of them on a chip and engineer them to work efficiently together, and that's exactly what the N100, N300, and N305 are.
I would LOVE to see 64gb or even 32gb ram versions of something like the lattepanda mu so I could toss a bunch of them on a multicarrier board that has 10g networking for each board. - they would make FANTASTIC low cost proxmox blade servers for homelabbing
The hard part with x86 soc is the short hop to micro atx or itx. Depends on the project, ofc, but decision making gets muddy.
You absolutely nailed the support aspect. If I’m building some embedded system, I have zero interest in cutting my teeth on integrating the weird pitfalls from an unknown sbc provider. Raspberry pi dominates because someone else has already done that for you.
I think the biggest kicker for support is the size of the user base. The more people are posting on forums the more chance someone else has already come across the problem you have and posted a solution to it and thats where raspberry pi has the advantage. You don't just have to make hardware that's better and cheaper it has to be so much better and cheaper to upset the market so much that it can come to dominate in order for the support to eaqual what you find with raspberry pi. I think it would take another pandemic scale chip shortage where some how a competitor has a significant advantage that they can continue to supply where r-pi can't. During covid all chip manufactueres were eaqually hit so it was hard to get anything. Unless raspberry pi just withers and dies I think that's the only way they'll be surpassed
The issue I had with the Pi 3, is that the cpu clock was used to derive spi bus clock.
When the cpu clock is throttled due to heat, so is the spi. I didn't try, but I believe the i2c and uart are also throttled.
I heard a lot of people had trouble trying to use the console uart due to this fact.
I don't know if this is still an issue on newer versions.
1:43 plus the other key part: software; from Canonical with Ubuntu desktop (yes the RPi4 runs the full desktop, although a bit sluggish), raspbian OS, debian and a couple of dozens Linux variants for ARM without issue to distros with niche uses like media servers, routers, audiophile-grade music players, adware remover, arcade machines and hundreds of video and I/O real-world applications, the Pi got you covered.
I think its fair to point out that there are a lot of great value options in the ~$150-$250 price range for single board computers or thin client desktops. My current 8GB Pi 5 setup with a 1TB nvme drive costs about $200. Now, providing an expandable platform like the Latte Panda Mu in that price range is impressive. I feel as though there is sufficient pressure in the market for more x86 SBCs like that to come out in the near future, which may give Raspberry Pi a run for their money.
The problem with the Mu is price and size. 🕵 LattePanda Mu Micro x86 Compute Module Evaluation Kit = LattePanda Mu Compute Module (8GB RAM and 64GB eMMC 👀) + Lite Carrier Board (smallest possible carrier board) + Active Cooler = $190, 146mm×102mm. There is a quite smaller and much cheaper N100 mini PC. For example Firebat T8 Pro Plus (16GB RAM and 512GB SSD 👀) = $130, 89.4mm x 89.4mm x 43.5mm.
After watching this I ended up buying three minisforum UN100D. They seem very good value. I know they are not SOC but someone pointed out something the other day to me, thats worth considering... How well will the emmc wear? It may be groundless but it got me wondering.
oh goooooooood the pm triangle
I'm thinking of trying out an Atomic Pi for a handheld Morrowind. I think the boards are discontinued but still available online new old stock around $50.
You should do a few projects with this board. Try making a NAS, a small media server or even throw a hypervisor on it and see what it can do.
5:30 Can you use SteamOS to get around the support issues? I heard you can install it on basically anything.
Valve affirmed that SteamOS 3.0 will continue to be freely available, with the intention of allowing other hardware developers to take advantage of it and build similar handheld computing devices like the Deck.
LattePanda: "Bye bye! 🤫👉😉"
My brain is so rotten it's ready to fertilize the fields for the next harvest
This is exactly what I've been telling people for years. If you are a maker, and you need a mini PC for a hobby project, pi is almost irreplaceable, because of the incredible software support and the huge community (I learned to program as a hobby with PI). If you only need a system for streaming movies, then the other options are just fine, but for almost everything else, PI is often the best solution.
'If you wish to make a Raspberry Pi from scratch, you must first invent the universe.'
the key is to implement a base interface between hardware and software, then be able to use the peculiarities of that hardware as traction grows. it comes down to modifying the architecture of linux. we need to separate the hardware and software interfacing in linux. this is different from a hardware provider supplying a device driver especially because it's an sbc, but even for modular x86 desktop pcs, we need a separation between hardware and software.
Is that a PCMCIA USB adapter I see in the drawer at 6:19?
the CM family of boards needs to have extra pin placements for standardization, and just basically have them be expansion ports, where they are effectively just PCIe, so that the carrier board can do all of the heavy lifting, but leave the basic 2 expansion ports for the basic connectivity we already have, but improve the features like maybe now it supports 4k 120hz instead of only 4k 60hz, even if the carrier board doesnt, it wont break support
kinda like how USB does it
Jeff, there are new Odroid H4 series SBCs available, they seem nice...
That's a great proxmox micro cluster