Slashdot Mirror
Survey Says: Raspberry Pi Still Rules, But X86 SBCs Have Made Gains (linuxgizmos.com)
DeviceGuru writes: Results from LinuxGizmos.com's annual hacker-friendly single board computer survey are in, and not surprisingly, the Raspberry Pi 3 is the most desired maker SBC by a 4-to-1 margin. In other trends: x86 SBCs and Linux/Arduino hybrids have trended upwards. The site's popular hacker SBC survey polled 1,705 survey respondents and asked for their first, second, and third favorite SBCs from a curated list of 98 community oriented, Linux- and Android-capable boards. Spreadsheets comparing all 98 SBCs' specs and listing their survey vote tallies are available in freely downloadable Google Docs.
Other interesting findings:
Other interesting findings:
- "A Raspberry Pi SBC has won in all four of our annual surveys, but never by such a high margin."
- The second-highest ranked board -- behind the Raspberry Pi 3 -- was the Raspberry Pi Zero W.
- "The Raspberry Pi's success came despite the fact that it offers some of the weakest open source hardware support in terms of open specifications. This, however, matches up with our survey responses about buying criteria, which ranks open source software support and community over open hardware support."
- "Despite the accelerating Raspberry Pi juggernaut, there's still plenty of experimentation going on with new board models, and to a lesser extent, new board projects."
Plus who cares about numbers? The pi and x86 boards are meant for totally different applications.
Only the State obtains its revenue by coercion. - Murray Rothbard
The Pi has strengths in GPIO pins, communicating with an Arduino, Ethernet, low power, etc. While you can install OSMC or RetroPi, the lack of an SATA slot makes for slow loading times and a poor desktop experience. The Broadcom video leaves Amazon Prime choppy, and DRM makes Netflix unusable.
The x86 strengths are in being a reasonable desktop replacement with a small footprint. The chipset makes it compatible with most services via Linux and Windows.
Shut up, AC. If you don't have something productive 2 say, then dont say it
Wow. Don't hold back BeauHD. Has Whipslash been hassling you?
It hits the sweet spot for price/performance/Low hassle.
Faster and more expensive? I might as well buy a cheap tablet.
Faster and cheaper? But lacks library support and a user based chock full of not just FAQ but rarely asked obsuratta that is key thing you needed to understand to get your job done
If your time has any value then there is no computer cheaper than a pi worth the price difference. One can say that almost factually.
THe ones that do compete are the ones offering more features like beagle bone.
Some drink at the fountain of knowledge. Others just gargle.
The Pi is great and all but its woefully underpowered. I've tried a number of different boards, the ODROID has way better specs and in the same price class.
But every other x86 and even ARM boards I've tried are unstable. UDOO, Intel Compute Stick, UP Board all worthless as they crash from overheating within 48h of operation. And on ARM boards I can find little under $200 that has anything better than a Mali 450 GPU which is already nearing a half a decade old.
Custom electronics and digital signage for your business: www.evcircuits.com
Intel just killed it's IoT platform line, so there are going to be fewer x86 options for SBCs.
Anons need not reply. Questions end with a question mark.
On the whole, people don't want open specifications more than they want something that is well-supported. Open specifications are a good thing, don't get me wrong. But given the choice between something that's a huge hassle to get working (and keep working) smoothly that's open and something that just plain works...well, I offer this survey's results as Exhibit A.
For your security, this post has been encrypted with ROT-13, twice.
This kind of devices will be banned in a couple of years or less because, terrorism. The future is a bunch of black boxes. Get over it.
pot, meat kettle
pot, meat kettle
The first post was indeed accurate and informative. The 'story' is just a survey firm marketing there warez.
The article little to do with x86 vs ARM market share. It is basically a comparison of different single board computers (SBC). To bring up x86 is kind of pointless - especially considering that Intel just killed the majority of their boards.
hypocrisy is a funny thing, sneaks right up on you doesn't it.
You said it is not suitable for "any use." The problem with absolutes is that any counter example shows you are wrong... and in this case there are many. I've already seen plenty of research labs and shops using them for automating equipment that isn't reprogrammable, or as an interface translation, or when a small computer is needed in an isolated setup. There are plenty of other suitable SBCs, or even specific devices for some of the interfacing, but often it is invaluable to be able to buy something down the street for $30 and have it working an hour or two later. Getting something $10 cheaper or twice as fast isn't going to be worth it for those projects if it adds a couple hours work.
If you want to continue calling that a toy, for something used by people paid to get things done in a shop selling things for profit, then you are just deperately try to protect your ego from some perceived threat that people can get stuff done with a crappy computer.
If they come with an HDMI output you would expect it to be able to compose a display at 1080p beyond a single stream movie.
Custom electronics and digital signage for your business: www.evcircuits.com
Toys can be useful too. That doesn't mean they are suitable for serious use. Can you use it for collecting data in a situation where it's barely under any load? Absolutely. Could you use something a whole lot cheaper to do that? Yes. Can you set it and forget it to do something important 24/7/365 without any real supervision? Having it to deal with a serious amount of data? Fuck no.
That's my problem with the RPI. It's a really powerful little computer, unfortunately it's usefulness is severely hampered by stupid "cost-cutting" which neither really change the cost very much nor makes much sense. Thus, it ends up only being useful for the most trivial tasks and too artificially limited to be really useful for the things it actually has the theoretical performance for --- because of stupid penny-pinching and because "reliably" never was on the checklist for the designer --- because it was designed as a "toy" for kids to learn to program on.
The RPI was never supposed to run for hours on end, it wasn't designed to deal with lots of I/O, it wasn't designed for large stints of data processing -- it wasn't designed for "serious use". And if you know anything about electronics you can tell.
No, I look at the specs and expect them to do what is listed because I understand it's an SBC, not a workstation.
Anons need not reply. Questions end with a question mark.
The 40nm process the Pi is fabbed on is now nine years old.
The Pi 3 is very thermally limited. Overheating and power supply related problems are very common. It's also only a ~30% improvement over the Pi2 while the Pi 2 was more like a 700% improvement over the original Pi. All of this would be very different if the Pi 3 had been fabbed at 28nm.
Yes, it doesn't make sense to try to push the Pi onto a leading edge process like 10nm, where per-transistor costs are going up rather than down and FinFET design gets more complicated. But a move to e.g. 20nm planar would make a huge difference and be enough to keep performance close to other ARM competitors for many years.
The trouble is that Broadcom has largely moved on and it's going to be difficult to get a die shrink (or any improvements to VideoCore) to happen. The Pi 4 will be a long time in coming.
"Physician, heal thyself"
Can you set it and forget it to do something important 24/7/365 without any real supervision? Having it to deal with a serious amount of data? Fuck no.
We have a couple rpis doing machine vision analysis in our QA shop, and they've been chugging at probably 50+% CPU load along for more than a year. When a vender for more serious equipment was taking a week to get back on an issue with broken equipment, we had a temp solution that became permanent.
The only reliability issue we've had is when using them to control some kilns, where even $1k+ controllers have died. It is pretty trivial and minimum cost to once a month or two to spend five minutes throwing in a new RPi with a copied flash card all setup to go.
Could you use something a whole lot cheaper to do that? Yes.
The equipment cost is trivial compared to engineering time. The cost of a RPi is effectively zero to us as long as you're not burning them for heat. The fact we can hand it to whatever engineer is least busy and have something working in a day is a godsend, especially when the PLC programmers are backlogged. We're not getting rid of the PLC guys, we're just focusing them more on where they are needed. Which is how tools and economics work in the real world: you use what is cheapest over all while getting the job done with sufficient quality. We would gladly pay $10+k for some controllers when needed in a pinch, but if the vender says it will take them a week while in house says they can buy a "toy" and get it done by the end of the day with no long term consequences, we go with the "toy" in that case.
Micro USB pushes 2A just fine, which is more than plenty. You're whining for the sake of it. Shut up.
Can you set it and forget it to do something important 24/7/365 without any real supervision?
Why is approximately 7 years the benchmark you're setting?
The deal-killer with the r.pi is the still-unfixed USB bandwidth bug that has plagued the platform since the first generation.
Twitter supports and protects racists - by smearing their critics with the "Hate Speech" label.
pot, meat kettle
Instead of meet. The mind boggles.
Any chance we could get our hands on a couple of your kiln-killed units? Be interesting to see what failed.
20x the CPU performance for less than 4x the price of a Pi.
That's an x86_64 SBC - actually made by AMD. They use more power (6-10W), have much, much, much greater I/O and run x86_64 code. They support AMD-v/VT-x, so running VMs or containers are trivial.
Just depends on the purpose. For many people, a r-pi v3 is more than sufficient to run a household server with email server, pi-hole/DNS, nextcloud, VPN/ssh Jump box and very lite media serving. But, forget about large disk I/O or network I/O on a r-pi. They aren't built for it.
For a little more money, still silent, you can get an APU2 or something similar. 10W max, but usually under 5W. Hook up some disks and with the GigE networking, you have a backup server and all the things a r-pi can do for cheap. Plus, you can have a few KVM virtual machines or probably 20 containers running specific applications, if so inclined. Or use it as a solid router that can actually be maintained (unlike any mainstream, home routers sold).
There is much more than a r-pi in the SBC space. Lots of options for lots of different needs.
The spreadsheet misses the Ci20 from Imagination - 1GB ram 2xMIPS.
However, that board in my experience is not stable. It hangs after a while under high load (such as building GCC). I have no idea why. Nevertheless, it is the only dev board with more than one hardware thread and with 1 GB of RAM (and Linux), so it's the only game in town.
Intel has just discontinued all its x86 devs boards.
Note finally all the "x86" boards here are x86_64, not i686.
I recently managed to snag an original Minnowboard, with an actual 32-bit Intel Atom on.
I've been looking for some time for a POWER or SPARC dev board (LEON) would do.
Best I've found are one or two niche and high-end SPARC or LEON boards, both about 2k USD each. No sign of POWER.
I was surprised by the Raspberry Pi 3 stability. I constructed a light on/off switch to switch on light in the apartment via internet when I am on holiday. It can run without reloading for long time. I do not know for how long as the system never has got any issues.
The only reliability issue we've had is when using them to control some kilns, where even $1k+ controllers have died. It is pretty trivial and minimum cost to once a month or two to spend five minutes throwing in a new RPi with a copied flash card all setup to go.
So far at least.
The equipment cost is trivial compared to engineering time. The cost of a RPi is effectively zero to us as long as you're not burning them for heat. The fact we can hand it to whatever engineer is least busy and have something working in a day is a godsend, especially when the PLC programmers are backlogged. We're not getting rid of the PLC guys, we're just focusing them more on where they are needed. Which is how tools and economics work in the real world: you use what is cheapest over all while getting the job done with sufficient quality.
No need to lecture me on the nature of economics. There is a question you're avoiding though, the one that usually gets overlooked until it's too late; What does it cost if the cheapskate solution fails?
If the controller failing because you're making it and its power-supply do things it was never designed to do doesn't result in anything being lost which is more valuable than cost of setting up a better solution, by all means go a head. It's a toy, and you're apparently not making it do anything particularly important anyway.
If, however, you're putting something which represents a greater value than the cost of $BETTER_SOLUTION at risk, such as e.g materials and/or significant amounts of time because the process got interrupted or went haywire, or getting corrupted data because of voltage problems, you're a fool.
If they come with an HDMI output you would expect it to be able to compose a display at 1080p beyond a single stream movie.
compose a display?
Honestly I've never checked. I only ever use the HDMI port for debugging. Generally they run headless for me and on the rare occasion I've built a gadget with a screen, then I use the official screen. The graphics (which I freely admit were pointless animations that I did for fun though amazingly actually increased the usability of the device) ran smoothly.
So, works for me (tm), but I've never tried using it as a desktop or laptop replacement.
SJW n. One who posts facts.
Micro USB pushes 2A just fine, which is more than plenty.
The problem isn't really the amps, but the volts. The RPI really needs it's 5V to work and the voltage tends to drop quite significantly in the cable, even if you manage to find a "psu" which actually is up to spec.
One would think that the very fact that most distributions actually have a visual warning for low voltage built in would suffice as a caution that there might be dragons that way, but apparently that's not enough for everyone.
You're whining for the sake of it. Shut up.
Look, the ignorant telling the bringer of complicated stuff to shut up. Well, that never happened before. Luckily, I don't need your permission to speak, sunshine.
And four of the top ten. Well done RPF! (Raspberry Pi Foundation.)
I believe that one of the most important factors is the same thing that makes the Arduino so popular amongst embedded controllers: Community. Both have vibrant active communities where newcomers can share ideas and get help. Both provide support for those getting started.
With the Pi Zero and Zero W we have a ridiculously inexpensive platform that runs a full blown OS. True, it is not up to snuff for replacing your desktop and costs do add up adding peripherals, but there are still some applications where the cost is significant.
FWIW, I have two employed. The first drives a retired monitor to provide a 'fireplace display' for SWMBO's office (https://github.com/HankB/pi-video-player) This one actually required some programming and runs on a Pi Zero. The other application is a Pi 3 B running an MPD server to play music on my home stereo. It would probably run on a Zero W (which was not available when I built this) but I'd have to rig up something for audio output since the Zeros do not have an audio jack.
These are very handy little devices.
First you said "any use" and now it's "serious use". Just say you hate the Raspberry Pi and be done with it, you are rambling like an Intel CEO trying to put down a competing product.
That doesn't we should be happy with poor thermal design and crashes. If you give a board certain features, then using them for any period of time should not cause them to overheat and crash.
You can't expect certain things from an SBC but not crash every 48h because there is some load is one of them.
Custom electronics and digital signage for your business: www.evcircuits.com
There's a difference between use and use. There's also something called context.
How about you just admitting you can't read for shit and get the fuck off the Internet, sunshine?