| General > General Technical Chat |
| Pi "foundation" gets fatter |
| << < (15/18) > >> |
| rsjsouza:
--- Quote from: ve7xen on September 27, 2021, 10:43:46 pm --- --- Quote from: rsjsouza on September 27, 2021, 09:48:09 pm ---Don't forget that the Raspberry Foundation was run by the Director of Marketing of BCM. They had severe ties to their main supplier that, even still, were unable to break through their own datasheets, binary blobs and supply chain to enable others to manufacture or create their own boards. --- End quote --- From the beginning it's always felt to me like the RPi foundation was just a vehicle for BCM to stretch the profitability of some of its older process nodes / designs. They probably get some kind of charitable tax cuts for it too. It's a slimy arrangement. --- End quote --- I share the same sentiment, although I can only speculate. The BCM device was an ARM11 that was well on its way out of the mobile communications market (Cortex A was all the rage) and the Pi was an interesting way to offset the losses of these devices. --- Quote from: ve7xen on September 27, 2021, 10:43:46 pm ---Nobody serious about designing an open platform would ever even consider BCM, and here we are several generations later when they have the volume to get competitive pricing on more open options without the 'friend price' connections, and they are still using opaque and closed ICs. At least they don't really sell it as 'open'. But however they spin it, it seems much more like the PR arm of Broadcom to me than any legitimately-charitable, independent enterprise. --- End quote --- Any analysis of cost of the RPis at the time evidenced a heavy subsidy going on. Although I don't follow modern instances of it, I imagine the same still goes. |
| djacobow:
--- Quote from: tom66 on September 27, 2021, 09:23:44 pm ---I've had Pi systems run for years on the SD card. --- End quote --- So, every time someone brings up the issues of Pi's an SD corruption, someone says this. You do see that this proves nothing, right? I've had a LOT of failures. I've got a jar here of at least a dozen dead SD cards from Pi's. I've also run a LOT of pi's. I've had cluster of of them, I've deployed them in monitoring stations by the dozen. It just so happens that the compelling use-case for most Pi projects do not lend themselves to net booting. Could that corruption be my fault? Sure! Brownouts, bad power, lightning, WTF knows? I don't really care very much. SD cards make for shitty OS disk. They fail. A lot. And when they do, it is not something that fsck can help with. The card is just toast. |
| bd139:
Oh it's even worse than that. Seeing as this is an engineering forum, lets do a top down on their PMC situation. There is literally a single NCP1117 between 5V0 USB input and 3V3. If you've got your usual wall wart with added inductor then any current dips can sink that below the dropout voltage of the regulator which means your 3V3 rail bombs. I've actually seen this on a scope on an RPi2 and it's the same circuit AFAIK now. That drives all the other LDOs and the BCM core as well which all lose regulation. This causes the BCM core to BOD and reboot outside of OS control. How can you do that? Well use a 1m long USB cable adapter and hot plug something on USB a few times. You can watch it floating around 4.2 volts under load as well as even the best of those supplies are pretty dire and there is I2R loss on the cable. This is pretty normal activity and pushing those LDOs close to the dropout voltage. What's the outcome? There the kernel and filesystem relies on two guarantees which are not met: firstly that the filesystem driver honours sync (it pretends it does here - another BCM turd) and that when it does a sync that all the data is persisted (it isn't due to SD - hence bad filesystem choice). Better hope it's not writing to disk when you're near it :-DD This is just simply poor engineering. |
| langwadt:
--- Quote from: bd139 on September 28, 2021, 08:26:08 am ---Oh it's even worse than that. Seeing as this is an engineering forum, lets do a top down on their PMC situation. There is literally a single NCP1117 between 5V0 USB input and 3V3. If you've got your usual wall wart with added inductor then any current dips can sink that below the dropout voltage of the regulator which means your 3V3 rail bombs. I've actually seen this on a scope on an RPi2 and it's the same circuit AFAIK now. That drives all the other LDOs and the BCM core as well which all lose regulation. This causes the BCM core to BOD and reboot outside of OS control. --- End quote --- which version is that? afaict everything except a ~10yo schematic shows a switcher |
| bd139:
See RPi2 in the message. This was an issue I raised on the forum, which was of course deleted like my other post... You can't evaluate the new ones as the public schematics are incomplete. |
| Navigation |
| Message Index |
| Next page |
| Previous page |