|
The cool thing about the pie is the GPIO, price and ubiquity TBH. If you want to do x86 things by all means get an x86.
|
#
?
Nov 20, 2015 00:28
|
|
|
|
| # ? Apr 25, 2024 04:48 |
|
|
eightysixed posted:I'm not sure Oracle is supporting ARM currently. I'd look into that first. They do not (publicly). evil_bunnY posted:The cool thing about the pie is the GPIO, price and ubiquity TBH.
|
|
#
?
Nov 20, 2015 01:05
|
|
|
TVarmy posted:Do any SD cards/USB drives even have anything like trim? High end ones can. But they'd also be wasted on a Pi because they're way too fast for the slow usb/sd reader on it.
|
|
#
?
Nov 20, 2015 01:08
|
|
|
Police Automaton posted:The ARM thingies are still unbeatable on the Wattage/processing power ratio tho. I don't know, personally I get good mileage out of my Pi2 for my server tasks (honestly server sounds almost like to big a word, I'd rather call it auxiliary computer or something) you just gotta know the limitations. No, I know what you mean. That's what I originally bought my Pi for. I wasn't looking for like, mega webserver or anything, I just wanted something to serve my USB drives and run usenet/wget downloads. That is the stuff that you are way better off doing Baytrail-M or Baytrail-D for. The Pi2 is much much better at actual CPU performance but it's still behind, and it still doesn't actually touch the overall system performance from factors like having eMMC, mSATA, USB 3.0, real Gig-E on a separate channel, etc. quote:That being said, I also still use my old (bought when they were all the rage, it was quite similar to the ARM rage now) MSI Wind U100 (Atom N270) Netbook (expanded with 2 GB and a bigger heatsink so the fan doesn't really need to work often) to look up stuff quickly or look at PDFs/Texts or write something. (and while Tablets exist, real keyboards are just so much nicer in every way) Works perfectly, and ~17 Watts under load with the screen on is not too shabby either. Later generations improved on the power consumption, especially when idle. For something that runs occasionally only though and is mostly hibernating, +/- 3-4W really do not matter. I used to run my Zotac Zbox with a 2957U (Haswell-ULT vs Baytrail-M) and it had roughly similar power consumption to a Liva. It made a great TV PC until I stepped up to 4K, which the Intel drivers said was "too much bandwidth" at 4K@30, then I moved to a mITX build with a GTX 750 Ti. Belatedly, I discovered that the new versions of Ubuntu (at least) are willing to drive at 4K@30. Even if it's idling all the time, 4W doesn't matter. That's like leaving a compact-fluorescent bulb on for one day a month. At US power rates that's something like $0.50 per month of electricity. If you're shutting it down, it's going to be literally unnoticeable. evol262 posted:The x86 boards also have gpio The Liva doesn't, but if you can name some names that run x86 that do please feel free. My suspicion is that (all things equal) ARM tends to have better hard-realtime performance characteristics than x86 due to architectural reasons. However ARM are usually weaker than x86 counterparts. And running GPIO directly on the processor, on a non-realtime OS like Linux, is not really the best way to do things. Many (if not all) practical tasks are better handled by a hard-realtime daughterboard like a Bus Pirate for I/O, or things like an FPGA for clock-sync'd logic. Paul MaudDib fucked around with this message at 02:52 on Nov 20, 2015 |
|
#
?
Nov 20, 2015 02:50
|
|
It depends on what you need. A particular system really does not have to be the theoretical best as long as it just practically works for your application.
|
Just curious, this is the Pi combo I bought right here: http://www.amazon.com/gp/product/B008XVAVAW?keywords=raspberry%20pi&qid=1447983284&ref_=sr_1_5&s=pc&sr=1-5 And a lot of why it's an easy thing to buy is that it's cheap, it's available, it's all in one kit, no waiting for a sale price, all that stuff. It's practically an impulse buy. $70 and it's on my doorstep in two days. Two hours later and Wordpress or something is on it. So I have to spend two more hours in six months to replace the flash card. Big deal. (Right now mine is running RetroPie and is used as a media streamer and a light file server.) But anyway I am interested in maybe getting something with more horsepower and reliability. What's the next incremental step up? Any complete kits? Something that's in stock right now. I read the word "BayTrail," then I'm looking at this: http://www.amazon.com/gp/product/B00B7I8HZ4?keywords=baytrail&qid=1447984102&ref_=sr_1_1&sr=8-1 Up front cost is $145, but then I scroll down and see this: quote:You'll need to order memory, mSATA solid state drive, and a "Mickey Mouse" power cord for the NUC kit. You might also want to order a WiFi card. These are easy to install with the instructions provided. Then you'll need to install an operating system on the mSATA drive. I used an external USB DVD drive to install Windows. Is that the next step up? Something that ends up costing three times as much? I'm not incredulous, just wondering what the reality is here. edit: tl;dr - I'm a huge baby who demands that people pick my next purchase out for me. doctorfrog fucked around with this message at 03:08 on Nov 20, 2015 |
|
#
?
Nov 20, 2015 02:55
|
|
|
doctorfrog posted:Just curious, this is the Pi combo I bought right here: http://www.amazon.com/gp/product/B008XVAVAW?keywords=raspberry%20pi&qid=1447983284&ref_=sr_1_5&s=pc&sr=1-5 This is what you're looking for: http://www.newegg.com/Product/Product.aspx?Item=N82E16856501011 32GB onboard eMMC instead of 8GB SD card, mSATA socket, non-lovely power adapter included, case comes standard. That's the model with a mSATA socket and the (marginally) upgraded processor over the original. You don't need a "kit", that's the poo poo that's supposed to come in the box when you order a product. I guess you're on the hook for the HDMI cable though, if you somehow don't own one. gently caress, down to $70 AR. Guess I'm getting a couple more. Paul MaudDib fucked around with this message at 03:09 on Nov 20, 2015 |
|
#
?
Nov 20, 2015 02:59
|
|
|
Paul MaudDib posted:This is what you're looking for: Cool. Eyeballin' that for X-mas.
|
|
#
?
Nov 20, 2015 03:08
|
|
|
Those NUCs are pretty cool. If you're going to setup a low power permanent home server for misc tasks, you may as well go balls-out and get an i5 Skylake version for $350. It's $200 more, but power consumption is the same, and long term it will last you much, much longer. Also you have access to 16GB of ram (potentially, depending on your budget) I bought a 10w cedarview (the one before bay trail) atom "nettop" thing a while back (2012?) for $130 and I liked it but quickly ran out of capacity trying to run VMs in Virtual Box on it. Also plugging in external hard drives via USB I quickly realized was not a good idea. It's now living life as a pfSense (firewall) box. When Haswell came out (2013) I picked up an i5, tossed 16GB ram at it, and it runs a bunch of VMs, one Windows 10 VM for windows apps, one Windows 2012 R2 vm for file sharing, and I spin up various linux VMs for various projects/needs. With the way modern hypervisors work you can stick 8-10 VMs on 16GB memory due to dynamic memory. It sits in a low-noise 4U case with a couple of striped drives that's capable of holding 15 3.5" drives for future expansion. I'm expecting the system to last me quite a while and idles at under 30w in the closet without ventilation. With an i5 NUC you get all of the computing power, all of the form factor, and still get access to fast storage, especially as performance/size/price of SSDs continues to improve. 1 TB SSDs are in the $320 range and will probably drop under $200 after Christmas so you might not need to deal with external USB hard drives.
|
|
#
?
Nov 20, 2015 03:17
|
|
|
Paul MaudDib posted:This is what you're looking for: I really have no need for this as my Pi is my media center and I have two spare Core2Duo laptops available.....but I want one.
|
|
#
?
Nov 20, 2015 03:19
|
|
|
doctorfrog posted:Cool. Eyeballin' that for X-mas. That's still a great deal and it won't stay good forever - the rebate expires at the end of the month and it's entirely possible it goes out-of-stock before that. But they drop down into the ~$80 range pretty regularly, you can get a deal like that fairly regularly. The point I'm going for is that the Rpi's price is artificially low because it doesn't include a bunch of stuff you need to run it. Even if it does, it's unreliable because it's a bare-minimum system. I have been running my OG Liva for 11 months now in exactly the same use-cases I used my Pi in, and it's been rock-solid over the same timeframe that the Pi burned out the better part of a dozen SD cards. Even without the Liva X's mSATA drive, the OG Liva performs an order of magnitude better, it's usually only like another $30 more expensive than a Pi/2, it includes a bunch of poo poo including wireless/USB 3/gig-e out of the box. But you absolutely cannot discount the labor in all of this. I have sunk countless hours of labor into rebuilding burned-out Pi SD cards and I can't tell you how much that it's 1000% worth the extra $30 to do it a single time and be done with it.
|
|
#
?
Nov 20, 2015 03:21
|
|
|
eightysixed posted:Can't tell if that's directed at me or not. But any drive is going to fail. When it fails, it's obviously no longer useful. I have some old WD Greens running that are way past their EOL timetable. I'm running ticking time bombs. Just because they haven't yet, doesn't mean they're not going to. Same applies to SSD, SD, your regular flash drives, etc. Same with anything/everything else electronic. Sure, it's directed at you. SSDs really, really do not fail on an automatic timetable. They are not spinning disks and they do not have bearing wear, arm failures, platters that get scraped, etc. If you treat them well, they will last until the flash cells start failing. I have just started replacing all my first-gen SSDs from the 2011 timeframe. So far I have had zero total failures - I am replacing them because I am tired of dealing with 120GB system disks. They are built to identify flash cells that are near their write limits (with a given safety margin) and move the data to safer cells. That's a huge part of what the flash controller does. Can errors occur? Yes, totally, but it's much more predictable than HDDs. Electrical failure modes reveal themselves right away, and there's no mechanical failure modes. If you keep them exposed to electricity once every while, they really just don't fail that much. The one exception is hard-power-offs during heavy writes. Most SSDs don't handle that well, except the Intel drives. Now, write limits? No, that's not an issue. You do need to leave some un-allocated space so the flash controller can move worn cells around without write amplification. Ideally you should avoid highly random writes, like heavy 4kb-sized paging or small un-cached random database writes. You definitely should not race out to buy used PCIe SSDs because I guarantee some corporation has been writing them to death in their database servers. But for a new SSD, or a SSD that was used in a piece of consumer corporate equipment like a desktop or laptop, you really can't kill it. Even under extremely heavy load you are talking 10 years of wear or more, and the usual lifespan of the flash memory is 20+ years. Everyone knows hardware fails sometime, but SSDs really don't wear out within timeframes like HDDs do. The only two caveats are that you should leave some unpartitioned space to allow the SSD to do wear levelling, and that you should install enough memory and then disable swapping/paging. 99.9% of the time swapping isn't a problem but if you do it while the disk is full and it ends up being a lot of continuous, small, random swaps (ie your working set is really just too big for memory and you are heavily thrashing your pagefile) it could trigger enough write amplification that you could have a measurable amount of SSD wear. This is mostly just me being paranoid though. In comparison though an SD card has a controller but it's your retard brother's retard brother. It's mostly just designed to write sequential blocks, and it doesn't perform more than a trivial amount of wear levelling at best (the SD spec does not require any at all). It's designed for writing a bunch of 20 MB photos one after the other and then being fully erased, not for hosting a running Linux operating system. Forget speed, TRIM would help with system stability, at least a little bit... Paul MaudDib fucked around with this message at 01:55 on Nov 21, 2015 |
|
#
?
Nov 20, 2015 03:41
|
|
|
Bovril Delight posted:I really have no need for this as my Pi is my media center and I have two spare Core2Duo laptops available.....but I want one. Ugh, ordering one as well to gently caress around with.
|
|
#
?
Nov 20, 2015 03:50
|
|
|
Muad'dib you should start a new thread, man. Save some future effort with one good effortpost, cuz people need to know.
|
|
#
?
Nov 20, 2015 03:56
|
|
|
That really is cheap. Too bad these things cost around 160-180 euros where I live, if you get them at all. They don't really seem to be in the market here in germany which frankly is quite odd. Also yes, start a thread about these x86 architecture mini PCs, they're quite interesting.
|
|
#
?
Nov 20, 2015 05:17
|
|
|
Paul MaudDib posted:The Liva doesn't, but if you can name some names that run x86 that do please feel free. Minnowboard did/does. So does the gizmo. Gizmo's relatively pricey, and the minnow is a little older, but it's a thing. There are breakouts for Edison, too. I agree about the rtos, but I think qnx for x86 is still free...
|
|
#
?
Nov 20, 2015 06:05
|
|
|
Paul MaudDib posted:99.9% of the time swapping isn't a problem but if you do it while the disk is full and it ends up being a lot of continuous, small, random swaps (ie your working set is really just too big for memory and you are heavily thrashing your pagefile) it could trigger enough write amplification that you could have a measurable amount of SSD wear. This is mostly just me being paranoid though. * My friend actually payed someone to try and fix the laptop before bringing it to me - they wanted to replace the (perfectly fine) 500gb harddrive based on that windows error, but crystal disk info showed the real culprit. I ended up replacing the msata drive with something big enough to be the system drive itself, with the hdd now being used as a storage drive - it goes much faster than that useless cache setup.
|
|
#
?
Nov 21, 2015 01:33
|
|
|
Hadlock posted:Those NUCs are pretty cool. If you're going to setup a low power permanent home server for misc tasks, you may as well go balls-out and get an i5 Skylake version for $350. It's $200 more, but power consumption is the same, and long term it will last you much, much longer. Also you have access to 16GB of ram (potentially, depending on your budget) What NUC are you talking about in particular? I've really been thinking that I should do the same thing, and virtualize a couple machines on a big SSD. It's usually not like I'm running 100% load on every single machine of mine all the time. Tried Hyper-V Server 2012r2 on my Athlon 5350 and it ran like total poo poo, orders of magnitude worse than it should have. I have a weird boner for commodity builds. I have this thing for the Build-A-Blade BB-ITX84 rackmount, which puts 8x thin-ITX boards in 4U of space. Lately I've been thinking that something like a bunch of Liva Xs in a cluster with mSATA or some other fast-local storage would also be pretty nice. I suppose that's what Knight's Landing is about though.
|
|
#
?
Nov 21, 2015 01:44
|
|
|
You'd probably have more luck with a different hypervisor. Hyper-V performance on AMD is inexplicably terrible sometimes. But the NUCs perform OK for a virt lab. Then again, almost anything performs OK for a light usage virt lab these days, including Atoms. You'd really just get killed with the lack of expansion if you ran up against shared storage limits (or worse, tried to use gluster or vsan to share fast local storage). More NICs are better in many ways. Intel's putting much more money into making hypervisors smarter about scheduling interrupts and such than into silicon improvements. Phi is to take away some of GPGPU's steam, if possible. evol262 fucked around with this message at 01:54 on Nov 21, 2015 |
|
#
?
Nov 21, 2015 01:50
|
|
|
Here is the i5 Broadwell NUC - http://amzn.to/1S9ZRDH Here is the i7 Broadwell NUC - http://amzn.to/1S9ZWHx Please note that they're Broadwell and not the new Skylake I didn't mention it but I'm actually running WS2012 Hyper-V Server. VMWare is probably a better option as that will give you more widely applicable work experience with a bare metal hypervisor. But Hyper-V works loving fantastic on Intel Haswell-class stuff. I would guess that it runs great on Broadwell/Skylake too. Vsphere/ESXi are both free to use for non-commercial use (as is Hyper-V, obviously). If you can wait for the Skylake NUCs to come out, probably next month or in January, they will have 16 and 32GB SODIMM support, which will be pricey at launch, but at least you'll have that option. Otherwise you're stuck with 8GB SODIMMs due to different memory types in Broadwell and the newer Skylake. I have 24GB in my VM lab box but I'm only using 12 of 20 avalible memory, and I know three of those VMs are grossly over-provisioned (one runs my Visual Studio dev environment with 4+ GB RAM). And also yeah you might be able to get away with an older gen CPU in your NUC, CPU isn't super important anymore, just access to gobs of memory is. Intel i5 and higher both have vt-d and vt-i which allows you to do some cool stuff like funnel storage spaces (virtual disks) between VMs without latency issues. I ended up with this 4U case (Rosewill RSV-L4500), it's about $100 but it has 4x 120mm fans running at low RPM and is under 40dB at 3ft. The last thing you want is a hairdryer that lives in the closet and makes so much noise you power it down most of the time because you can't stand the noise. Hopefully the last server case I'll need to buy in the next 15 years. http://www.newegg.com/Product/Product.aspx?Item=N82E16811147164 Hadlock fucked around with this message at 02:18 on Nov 21, 2015 |
|
#
?
Nov 21, 2015 02:12
|
|
|
Fuzz1111 posted:No you are probably right. The only SSD that I've ever seen get worn enough to trigger windows generic "hard drive failing" warning* was an msata 24gb SSD being used as a cache drive (paired with a 500gb mechanical drive) in a friends laptop - it was only 2 years old but as a cache drive I'm guessing it saw a usage pattern similar to what you are describing. What I've done so far is to run a SSD as the main drive and then I pop out the CD drives and run a 2.5" HDD in the bay for bulk storage. So far I have had zero eMMC or SATA drives fail on me, even the crappy 2010-vintage Toshiba SSD that I got as a corporate refurb Analogously, I recommend running an eMMC or mSATA drive for boot, with 10% unallocated and no swap partition, and then running USB 3.0/SATA/etc for a spinning-rust drive for bulk storage. Again, if you don't run the drive full or you don't thrash the pagefile heavily I think you would be fine, but my philosophy is to play it safe. There's a principle in programming that when something goes wrong you fail loudly and abort, and similarly I would rather have a program abort because it's out of memory than to have my drives just silently thrash and wear out. Obviously a lot of laptops don't have CD/DVD bays but you can get nice Thinkpad workstations real cheap nowadays since corporate users are done with them. My GF and I both have Thinkpad W510's - I paid $480 for mine in Jan '14 and I paid $228 for hers in April '15, both with 900p screens, (weak but) discrete graphics, etc. It's not like there have been many substantial improvements since the original Core i-series release anyway, for most practical purposes.
|
|
#
?
Nov 21, 2015 02:19
|
|
|
Paul MaudDib posted:The Liva doesn't, but if you can name some names that run x86 that do please feel free.  quote:My suspicion is that (all things equal) ARM tends to have better hard-realtime performance characteristics than x86 due to architectural reasons. However ARM are usually weaker than x86 counterparts. And running GPIO directly on the processor, on a non-realtime OS like Linux, is not really the best way to do things. Many (if not all) practical tasks are better handled by a hard-realtime daughterboard like a Bus Pirate for I/O, or things like an FPGA for clock-sync'd logic.
|
|
#
?
Nov 21, 2015 06:17
|
|
|
My Rhythmic Crotch posted:The original PC standard included the parallel port, which is GPIO. Cool bro, but how is the IBM XT standard from 1980 relevant to single-board computers in 2015? Nobody here is running an IBM 5150 PC, you have to make the case for why that standard matters if you want it to stick. quote:I'm dying to know why think ARM has better realtime performance than x86   Highly-microcoded architectures are inherently inferior to RISC architectures for hard-realtime applications because the execution time is longer and less predictable. Yes, an x86 processor is faster in the long term, but in terms of millisecond-level interrupt handling, how do you know what your CISC processor is going to do behind the scenes with out-of-order superscalar execution and all that jazz? CISC processors (like x86) can't really guarantee hard latency times nearly as well as a RISC system. And poo poo like an FPGA with a hardware-defined clock signal than either is infinitely more reliable because you forego the hassle of clock/edge sync. Paul MaudDib fucked around with this message at 07:03 on Nov 21, 2015 |
|
#
?
Nov 21, 2015 06:48
|
|
|
x86-64 is mostly RISC with a CISC compatibility wrapper, if I understand it correctly ARM is a totally clean-sheet design, though, and was designed from the start to throttle it's clock speed and sleep cores not in use. Intel has added those features, but there's still some (but not much) cruft left over having to support old 16 and 32 bit applications. OSX, iOS, Windows, Unix etc aren't real-time OSes so real-time performance isn't really a consideration unless you're building an ultra-low latency robot or GPS sattelite, NASA space probe, etc. The Raspberry Pi foundation does provide a RISC OS which is a realtime OS, but I don't think it comes with a TCP/IP stack and/or wifi drivers.
|
|
#
?
Nov 21, 2015 07:04
|
|
|
Yeah I'd definitely agree there's a divergence into "poo poo that runs on an ASIC/FGPA in transistor-logic" and "poo poo that runs on tens/hundreds of microseconds on a non-RTOS like Linux". For most stuff the difference doesn't matter, but when you're running things that involve the term "hard-realtime" x86 is at an architectural disadvantage. Hard realtime is all about the question "can I guarantee poo poo gets done before the next data cycle and/or interrupt" and when your processor architecture is getting fancy the answer is usually "no". It's very much a "worst case scenario" problem and average instruction throughput doesn't matter if you can't guarantee the worst case. Paul MaudDib fucked around with this message at 07:18 on Nov 21, 2015 |
|
#
?
Nov 21, 2015 07:08
|
|
|
How many of you are building stuff like ECUs for cars though? In such application circumstances you're usually straight in the Realm of ASICs/FPGAs/microcontrollers anyways. That stuff is ARM country (at the very best) and will be for the future for aforementioned reasons. You do not want any added complexity in such applications there, already alone for liability reasons, especially in products where people can get injured or die when they run into some unforeseen problem. Some complex OS with a 2 GB RAM overhead you just will not find there. Also remember that lots of embedded computing in industrial settings still depends on "ancient" technology like 68k (also RISC) and only in the recent years has seen a switch, mostly because the supply situation of spares is becoming very complicated. A company recently started to produce an 100% pin-for-pin-ready-to-drop-into-old-designs compatible 68020 clone, there's enough need for these old CPUs that it is worth it. New designs are expensive and in stuff like commercial aviation.. forget about it. You don't just go and "change" things there.
|
|
#
?
Nov 21, 2015 07:47
|
|
|
Police Automaton posted:How many of you are building stuff like ECUs for cars though? In such application circumstances you're usually straight in the Realm of ASICs/FPGAs/microcontrollers anyways. That stuff is ARM country (at the very best) and will be for the future for aforementioned reasons. You do not want any added complexity in such applications there, already alone for liability reasons, especially in products where people can get injured or die when they run into some unforeseen problem. Some complex OS with a 2 GB RAM overhead you just will not find there. Also remember that lots of embedded computing in industrial settings still depends on "ancient" technology like 68k (also RISC) and only in the recent years has seen a switch, mostly because the supply situation of spares is becoming very complicated. A company recently started to produce an 100% pin-for-pin-ready-to-drop-into-old-designs compatible 68020 clone, there's enough need for these old CPUs that it is worth it. New designs are expensive and in stuff like commercial aviation.. forget about it. You don't just go and "change" things there. This reminds me a bit of those articles that come out now and then announcing that the US Navy has a bunch of nuclear computers running Windows XP (or something else embarrassing sounding). I'm sure that there are plenty of outdated dumb as hell things the military keeps around for tangled reasons, but I'd imagine that certain things you just wait for the funding to replace the whole dang system because who knows what bugs'll crop up.
|
|
#
?
Nov 21, 2015 08:26
|
|
|
One thing to remember for stories like the us government still running XP in military applications is that if you're willing to pony up a million plus a year, Microsoft will continue delivering patches for any operating system back to at least Windows 3.1 + DOS 6.
|
|
#
?
Nov 21, 2015 17:28
|
|
|
Paul MaudDib posted:Cool bro, but how is the IBM XT standard from 1980 relevant to single-board computers in 2015? Nobody here is running an IBM 5150 PC, you have to make the case for why that standard matters if you want it to stick. Though that book appears to be published in 2012, on that page the only reference cited is from 1991. A lot of the points raised would have absolutely made sense in 1991, but now I'm not so sure. Intel's R&D has done some pretty amazing things to x86. I took HPCA last semester and we used only a very dumbed-down model of the pipeline, and I could see how while the out of order capability may make sense as an argument, there's also so much room for optimization that we never touched on. But anyway, it was just funny to see you poo-poo on x86 latency and then suggest a Bus Pirate as a better way to do things. No latency there, no sir!
|
|
#
?
Nov 21, 2015 17:48
|
|
|
Actually I suggested a Bus Pirate as being an easy way of getting GPIO on a machine like a Liva that doesn't have any GPIO by default, not that it was the highest performing interface out there. For demanding hard-realtime applications you probably want an FPGA, as I stated. And generally that a non-RTOS like Linux probably isn't the best idea either. Once you're on an RTOS you're 95% of the way there - processor architecture doesn't really matter next to things like overall performance, but all things being equal I still think there's some small advantages to RISC-style architectures for that type of application. For most Raspberry-Pi style "hook a thing up to I2C/SPI" projects the Bus Pirate does fine though. As somebody said, nobody here is homebrewing GPS receivers. Paul MaudDib fucked around with this message at 23:55 on Nov 21, 2015 |
|
#
?
Nov 21, 2015 23:45
|
|
|
RPIO on Raspian (and I assume, Ubuntu) hooks in to the RAM timing clock, I've had no trouble doing smooth timing for stuff like servos that requires sub-2-milisecond timings. That gives you access to real-time operations even when doing heavy duty stuff like OpenCV (robotic vision) and ultrasonic distance measurements. If you need additional PWM out you can get a HAT for that, if you need additional GPIO in, there are a lot of options avalible as well. You can do things like high speed serial to an Arduino MEGA, or even talk to it via I2C or SPI bus, giving you ~50 additional GPIO. At that point you're up to like, $60 for all the RPi stuff plus another $50 for an Arduino MEGA, plus wiring plus custom code. If you want some integrated solution, check out the Nvidia Jetson TK1 it has ~50 GPIO and enough CPU horsepower to melt paint off the wall at 50 yards for $179. http://elinux.org/Jetson/GPIO http://www.nvidia.com/object/jetson-tk1-embedded-dev-kit.html It looks like there's the TX1 now too, but it's $599
|
|
#
?
Nov 22, 2015 00:40
|
|
|
This a surprise to anyone? https://www.raspberrypi.org/blog/raspberry-pi-zero/ It's five bucks. It's made in Wales! It's really small so it probably needs cables to do anything useful to me, but, five bucks. edit: ten bucks says it'd cost fifteen to get it anywhere near me. doctorfrog fucked around with this message at 09:22 on Nov 26, 2015 |
|
#
?
Nov 26, 2015 09:17
|
|
|
The C.H.I.P. $9 computer people recently announced that they would be selling their device for $8, like, yesterday. Those Broadcom chips must be crazy cheap these days. And yeah they're probably feeling pressure from the NodeMCU/ESP8266 folks who are shipping devices for under $10 The big disadvantage of the Pi at this point is it still relies on SD memory, rather than on board storage, and lack of built in WiFi, which all of their competitors have now. The Pi Zero at $5 looks pretty interesting, but you still need a $10 Wi-Fi nub and a $5 SD card, which at least brings it down to $20, but that's still double the price of it's fully integrated competition.
|
|
#
?
Nov 26, 2015 10:00
|
|
|
I don't understand why they didn't put wifi on that if they cut ethernet. Still, $5 and made somewhere with labour laws is nice.
|
|
#
?
Nov 26, 2015 10:12
|
|
|
There's an interesting hackaday article, Olimex claims the BOM for the CHIP is closer to $25 http://hackaday.com/2015/06/11/does-the-worlds-first-9-computer-cost-9/ If you sift through the comments, general consensus seems to be that the company building the CHIP is in cahoots with a tech incubator in China, and furthermore, Allwinner is a controlling partner in that incubator. The CHIP runs some IoT variant of the A20 made by Allwinner. Having access to the raspberry pi tool chain is nice, but it'll be really interesting to see what happens when the English version of the Allwinner tool chain comes online. I paid $15 shipped for an Orange Pi with the Allwinner H3 a couple of months ago, my guess is that in another year something like the ESP8266 on steroids will hit the market, and in 2017 we'll see some quad core monstrosity for under $10 with Wi-Fi and eMMC running Linux.
|
|
#
?
Nov 26, 2015 10:37
|
|
|
$5 Raspberry Pi is pretty amazing. It would have been great if they could have gotten wifi in there, or at least left ethernet intact. I don't think there is much use for these things without networking of some kind. It seems to be that networking is more important to most RPi applications than output to a computer monitor.
|
|
#
?
Nov 26, 2015 15:35
|
|
|
It sort of sounds like it will do data over the charging USB, like on the Beaglebone. Otherwise yeah, like you said without networking it's going to be extremely difficult to program these things. Most of my interaction with my Raspberry Pi has been via SSH, generally I only connect it via HDMI long enough to get a root level user and SSH setup, and the HDMI never gets used again. With a $30 , 7" touchscreen, this would make a great base for an offline informational kiosk.
|
|
#
?
Nov 26, 2015 16:09
|
|
|
PBCrunch posted:$5 Raspberry Pi is pretty amazing. It would have been great if they could have gotten wifi in there, or at least left ethernet intact. I don't think there is much use for these things without networking of some kind. It seems to be that networking is more important to most RPi applications than output to a computer monitor. The on-board ethernet of previous models required both the ethernet IC and a USB hub IC so those were probably the first casualties when it got pared down.
|
|
#
?
Nov 26, 2015 19:32
|
|
|
Hi thread, A 10 year old in my life is is kind of going in the wrong direction and I'd like to redirect him onto computer stuff. He's into computer games (as in, he's not actually allowed to play them, but watching other people play them on YouTube is okay?  ) and thinks MineCraft and Garry's Mod are baller. Do you guys have any ideas for how to get this kid involved in building stuff/learning programming? I'm thinking about preordering the raspberry pi zero but all the project ideas I've come across look too sophisticated and maybe a little boring for a kid who is new to it. Any inspiring projects or places to start learning programming that would appeal to a 10 year old?
|
|
#
?
Nov 26, 2015 21:53
|
|
|
Get a Pi 2, not the Pi zero. The zero will be really frustrating for a first time user--you need to solder in the GPIO header, add a mini HDMI adapter, and a USB OTG cable, and even then you only have 1 USB port and no networking. You're going to end up buying all those things and the price will be more than just getting a Pi 2 to start with. The zero is neat but it's better for highly portable or other embedded uses. The Pi 2 is a much better general computer board to mess around with. Grab him/her a Pi 2, a good power supply & 8gb micro SD card, and maybe a GPIO breakout cable (if you wanna go all out get a big kit like this https://www.adafruit.com/products/2380 otherwise just the power supply https://www.adafruit.com/products/1995 and 8GB micro SD card https://www.adafruit.com/products/1294 are all else you need). Then get a fun book like programming minecraft on the pi (https://www.raspberrypi.org/blog/adventures-in-minecraft/) or maybe the new official Pi projects book (https://www.raspberrypi.org/blog/#official-raspberry-pi-projects-book).
|
|
#
?
Nov 26, 2015 22:24
|
|
|
|
| # ? Apr 25, 2024 04:48 |
|
|
cov-hog posted:Hi thread, Don't buy anything to start with. Get them used to Scratch first to see if they have the aptitude for programming... https://scratch.mit.edu/
|
|
#
?
Nov 26, 2015 22:58
|
|
