$/GFLOPS for both GPU and CPU are horrible for BB CPU. AllWinner makes much cheaper chips with much higher raw speed.
Well, this is only fair if you are only interested in raw computing power. In a case like this, you most probably don't need (or want to) design a new board from the chip itself.
On the other hand, the BB chip has some very nice features for embedded developments: both in hardware (such as the PRU and ADCs), and in documentation and support. And it is a very capable CPU, able to even implement SDR with GnuRadio (I did this myself). Most applications do not require much computing power beyond what the BB offers.
If I wanted to do that and be extremely lazy, I would use new version of BB's BGA system on module (that encapsulates the AM chip, memory, USB and PMIC in a single module in BGA).
That allows me to get things done with only 4-layer PCB.
Sure! This actually shows how the market that TI wants to reach is different from the AllWinner's.
I know Octave systems is the one designing it, but surely they have plenty of support from TI, to get the bare chips and all.
If I wanted to do that and be extremely lazy, I would use new version of BB's BGA system on module (that encapsulates the AM chip, memory, USB and PMIC in a single module in BGA).
That allows me to get things done with only 4-layer PCB.
I just wish they were priced better. The OSD3358 alone is the same price as a complete BB Green, but you still need flash etc.
Seen an eMMC not in a fine pitch BGA lately? (Everything available I can find is 0.5mm pitch.)
The 100LBGA at 1mm pitch which is pretty easy...
All this could still be ok for small volume applications with reasonable margin, but that generally means some sort of industrial market at least in Oz... But then they're not available in industrial temp grade.
I am not sure what design software you use, but if Altium, have a look at OpenRex. You can download complete Altium files and adjust them to your needs.
It is open source even for commercial use + components are from Digikey:
http://www.imx6rex.com/open-rex/Hope this helps.
IMX / TI stuff is probably the way to go.
The frightening thing is that both TI and the-company-formerly-known-as-Freescale are getting out of the ARM applications-processor market. TI is just leaving it (AFAIK no plans for ARMv8 chips), and Freescale is leaving by virtue of being bought out by NXP, then Qualcomm, which probably spells the death of future iMX development. I hope I am wrong but it does appear that the remaining players will all have similar policies toward the "little guy"--no data without NDA, no parts in small quantities, etc. Allwinner could start looking better in this scenario. When no company will document and publish as we used to take for granted, the next best choice may be a company which is bad at keeping secrets.
I wonder what beage.org is going to do if they cant get any more parts.
Beagleboard just introduced the x15, actually I think it's impossible to buy one yet in the US. There are also a gazillion cool projects for which the BeagleBone Black and similar boards are well suited, even if outdated in terms of the CPU. Also, the PRU (Programmable Realtime Unit) on those processors is so insanely useful it's a real shame that TI is getting out of the biz. But anyway, the Beagle_____ ecosystem won't be going away anytime soon.
I'm sure there will be plenty of stock for a while but if they can't find an new (open) replacement for the next version
The frightening thing is that both TI and the-company-formerly-known-as-Freescale are getting out of the ARM applications-processor market. TI is just leaving it (AFAIK no plans for ARMv8 chips), and Freescale is leaving by virtue of being bought out by NXP, then Qualcomm, which probably spells the death of future iMX development. I hope I am wrong but it does appear that the remaining players will all have similar policies toward the "little guy"--no data without NDA, no parts in small quantities, etc. Allwinner could start looking better in this scenario. When no company will document and publish as we used to take for granted, the next best choice may be a company which is bad at keeping secrets.
This has, more or less, been the play for ARM in general. Until ARM-GCC was a thing, the whole family was priced out of the market for anyone smaller than 'large enterprise' in terms of compilers. Really, the whole sense of ARM comes off as "go away."
Can anyone provide more of a source on TI getting out of the ARM processors? Best I can find is an article from 4 years ago regarding their OMAP processors.
Can anyone provide more of a source on TI getting out of the ARM processors? Best I can find is an article from 4 years ago regarding their OMAP processors.
I don't have a reference, but AFAIK they haven't announced anything to follow the OMAP5xxx series, so it would seem to be a foregone conclusion by now.
Can anyone provide more of a source on TI getting out of the ARM processors? Best I can find is an article from 4 years ago regarding their OMAP processors.
I don't have a reference, but AFAIK they haven't announced anything to follow the OMAP5xxx series, so it would seem to be a foregone conclusion by now.
TI was stopping making OMAP for the mobile industry. It does not make sense to make very complicated chips and then sell some millions of it with less than 1 dollar profit on it, if you can sell that many opamps with more profit. As far as I understand, ARM for industrial is here to stay.
TI was stopping making OMAP for the mobile industry. It does not make sense to make very complicated chips and then sell some millions of it with less than 1 dollar profit on it, if you can sell that many opamps with more profit. As far as I understand, ARM for industrial is here to stay.
Since Mobile space isn't my concern here, I'm okay then!
For what it's worth, here is the radio with the OrangePi mounted currently.
All things considered, I might just keep with buying OPIs due to costs.
For what it's worth, here is the radio with the OrangePi mounted currently.
All things considered, I might just keep with buying OPIs due to costs.
Interesting. Which Linux are you using? The ones I saw were pathetic, where basic things like detecting (or setting) screen resolution werent working.
Interesting. Which Linux are you using? The ones I saw were pathetic, where basic things like detecting (or setting) screen resolution werent working.
Just ARMBian. The screen resolution side of things isn't actually important to me, as my end-goal is for this thing to run headless. It's more of an appliance than a PC, I mainly needed the horsepower to crunch 8192pt FFTs, do some decoding, and post results to a website. This project frees a radio and PC, as it consolidates things into a single stand-alone unit.
IMX / TI stuff is probably the way to go.
The frightening thing is that both TI and the-company-formerly-known-as-Freescale are getting out of the ARM applications-processor market. TI is just leaving it (AFAIK no plans for ARMv8 chips), and Freescale is leaving by virtue of being bought out by NXP, then Qualcomm, which probably spells the death of future iMX development. I hope I am wrong but it does appear that the remaining players will all have similar policies toward the "little guy"--no data without NDA, no parts in small quantities, etc. Allwinner could start looking better in this scenario. When no company will document and publish as we used to take for granted, the next best choice may be a company which is bad at keeping secrets.
NXP leaving ARM? not even close.
i.MX family is still under development (i.MX8) and are all under extended life (over 10 years) due to automotive market requirements.
Automaker are really conservative guys, that's why Qualcomm had to buy a big player to really enter the market.