Slashdot Mirror
Raspberry Pi 3 Is a Nice Upgrade, But Alternatives Exist With Faster Performance (phoronix.com)
An anonymous reader writes: With the Raspberry Pi 3 now available, benchmarks have been done comparing the Raspberry Pi 3 to other ARM SBCs. The Raspberry Pi 3 was found to be a faster upgrade compared to the Raspberry Pi 2, but the ODROID-C2 is a much faster alternative. For only $5 more than the Raspberry Pi 3, it includes twice the amount of RAM, Gigabit Ethernet, and a faster SoC. The ODROID-C2 also has HDMI 2.0 and superior Ethernet while the Raspberry Pi 3 has an advantage of 802.11n WiFi. The ODROID-C2 also has a heatsink for ensuring the SoC doesn't get as toasty as the Raspberry Pi 3.
The Pine64 is nice as well.
But standardizing on one or two model also has its perks. Cheap cases and other peripherals, easy to find software, an abundance of tutorials.
Its starting to sound a bit like the history of the IBM PC.
Someone replaced my news with an ad for a Raspberry Pi competitor.
This story is at least two years old -- everyone knows that the reason to go with the Pi is the community support, not the benchmarks.
I find that so many other "competing" boards somehow miss the point. Keeping the Pi at or below $35 keeps it in the realm of disposable. 35 bucks can be put into a robot. $35 can be put into some dumb home automation system. 35 bucks can be sent to school with the kid's project. $35 can be risked in some project that might not survive.
But much more than that means that people start to ration. They don't buy multiples, they don't put it into risky situations, and they don't leave them behind as the brains of some project. To a large extent that makes any competitor that doesn't do the above a cheap crappy desktop. In that case I will just use my laptop/desktop.
The other factor is that there is generally a gradient of embedded systems. Most people are throwing Arduinos into things willy-nilly. This is because they are easy, very cheap, very low power, and really simple. There are a few more capable Arduino like boards which can do more but at a certain point people need something more capable. The raspberry pi is quite ready to step into that breach. It can run basic OpenCV, it can power things like touch screens, it can convert text to speech or the other way around. There are lots of things it can do. And it can do all of these while not rapidly draining a battery and it is fairly small.
But for most projects if the horsepower of a Pi is not enough, it is not probable that a small increase in horsepower is enough. Double or triple is not that big of a leap if you are talking about some Genetic Algorithm that needs to run in real time or some crazy complex image recognition or whatever. Thus a board that is a bit better is not really filling the gap for most people unless you buy something very expensive such as the new nVidia board but that is so far beyond the price range of the Pi as to not really be comparable.
What most people do when their Pi runs out of power is to offload the task to a desktop or laptop over some sort of data connection. Transmitting video in near real time or sensor data is not that huge a task and then you have a pile of power and might even be able to drop back to the Arduino under remote control.
Then there is the whole thing around the Pi becoming a bit of a standard. There are wonderful Python libraries well tuned for the Pi GPIO and whatnot. How well do they work with the Better-than-a-pi-board-2000? I don't know and I don't want to screw around with them for a day. Basically the Pi is pretty much going to be first in line for any ports such as ROS.
So if someone wants to compete with the Pi they need to understand that this is not a desktop war where some extra memory or a few more Hz is going to win my heart. A great example of a competitor that caught my attention is the C.H.I.P. for $9. It pretty much meets all the above requirements in spades. Now the completeness of the Pi with BLE and whatnot is pretty attractive but in many cases I could use the horsepower of the CHIP and the rest would be wasted. Thus I can see a future where I have 5-10 CHIPS in my toolbox, and 3-4 Pis. I don't see a future for a $60+ board in my toolkit. Literally the next step beyond a pi will be something that is effectively a small desktop, even if it needs to be an embedded system.
Computers like the Raspberry PI are great for start-up companies who can't afford to go to a manufacturer and have their own custom electronics made
Hi, hardware startup guy here. I'm going to come out and say that tautologically, if you know what you're doing, then doing custom hardware yourself is actually not nearly as difficult as you think.
First board design is cheap and easy. I'm using Eagle which has a free version if you want to get started, and a reasonably cheap version for smaller stuff.
I respectfully disagree. It might look simple if this is the first hardware project you've done, but wait till the bugs pop up, and *everything* has bugs. For my personal stuff I use the Geda toolchain, which I recommend to anyone getting their feet wet. I've designed and manufactured many devices for my personal use using Geda, but I would not recommend anyone designing their own boards - design daughterboards if you must. The reason is because you may lack the skills in figuring out what went wrong, when something eventually goes wrong (and it will!).
Small example for hardware difficulty: we've discovered problems in some of the units we manufactured last year, very inconsistent, very unpredictable and even the troublesome product would work perfectly *most* of the time. After a few months of long and hard examination of the troublesome devices (of which around 2 million, maybe, were manufactured and already sold), we finally pinned it down to a single capacitor on the device, which did not consistently withstand the washing process. The prototypes all worked perfectly (different washing process for small quantities).
This is something that will kill a small company in a year, even once you discover the problem and have a fix.
Another example (different device) involved static - even after hooking up the lightning generator and running repeated tests, it was still uncertain if it was indeed the static that was causing the problems that customers reported. The EE in charge, IIRC, added more static protection and called it a day, after more than a month of examination, testing and debugging.
In many ways, hardware is somewhat easy, especially if you have good quality 'scopes, signal generators, laboratories, etc; in other ways, it is very very difficult indeed, and there is no replacement for skilled and experienced EEs for debugging the hardware problems, which *will* arise after the prototypes all work perfectly. The best compromise is to pick a cheap module (RPi, *duino, etc) and design a daughterboard for the module that has the exact chips you need for a particular problem. That way, you can do the majority of your software testing and hardware testing from a PC connected to daughterboard, and switch to whichever SoC module suits you best.
I'm a minority race. Save your vitriol for white people.