Shop OBEX P1 Docs P2 Docs Learn Events
The New 16-Cog, 512KB, 64 analog I/O Propeller Chip - Page 74 — Parallax Forums

The New 16-Cog, 512KB, 64 analog I/O Propeller Chip

17172747677144

Comments

  • ElectrodudeElectrodude Posts: 1,638
    edited 2014-09-24 18:27
    brucee wrote: »
    Basically any ARM to update the BASIC Stamp (2K Flash and 32 whole bytes of RAM, really, it WAS great in its day, but that was 25 years ago), let me check my email on my Win95 machine...
    Or a propeller... There could be an official interpreter/compiler (probably would just use the spin interpreter) for something like the Spin Stamp that takes extended PBASIC that runs in one cog, and drivers could run in all of the other cogs.
    And the Prop, done in a process technology of 1995.
    That size process was all that was necessary.
    Yes the educational community doesn't much care, there are probably still people out there using Apple2's, and the hobbyist community doesn't care as much. But really, in the days of RasbPi, now the Intel Edison for embedded with video, WiFi boards <$30 with dual ARMs, other ARM boards <$20, and who knows what else might come out in the next 2 years, I don't get it.
    As David Betz said, none of those processors can do deterministic I/O.
    Yes playing with an open source CPU on an FPGA might be fun, even education, but it's not going to sell lots of units.
    But it already has...
  • bruceebrucee Posts: 239
    edited 2014-09-24 18:33
    But you can get deterministic execution on any ARM, just turn off interrupts around that section of code and run out of RAM. And at 62 cents each I can have 8 for less than a Prop 1 (remember no external components needed, boot Flash or crystal) and the IOs are 5 V tolerant. Comes in SO8s if you need small.
  • ElectrodudeElectrodude Posts: 1,638
    edited 2014-09-24 19:02
    brucee wrote: »
    But you can get deterministic execution on any ARM, just turn off interrupts around that section of code and run out of RAM.
    But if you only have one core and turn off interrupts, you might miss something else. The propeller can handle multiple things deterministically all at once, unlike an ARM.
    . . . remember no external components needed, boot Flash or crystal . . .
    Yes you do - you need tons of wires and other parts to connect the 8 separate ARM's. It would be much easier just to use a propeller and to deal with 2 extra parts and a few wires.
  • David BetzDavid Betz Posts: 14,511
    edited 2014-09-24 19:04
    brucee wrote: »
    But you can get deterministic execution on any ARM, just turn off interrupts around that section of code and run out of RAM. And at 62 cents each I can have 8 for less than a Prop 1 (remember no external components needed, boot Flash or crystal) and the IOs are 5 V tolerant. Comes in SO8s if you need small.
    You could lash together 8 ARM chips and that would work fine as long as they each had an independent task to perform. If they need to coordinate with each other then they lack fast access to shared memory like the Propeller provides and I expect most applications will require some coordination between cores.
  • bruceebrucee Posts: 239
    edited 2014-09-24 19:15
    Not all code has to be deterministic, it only has to be on a Prop because it has no peripherals. A UART on an ARM is deterministic and builtin, as are I2C or SPI or the other peripherals available, and with thousands of ARM variants you can find the combination you need including video and SDRAM. You can even get multicore ARMs (I've seen 4 CPUs, and you can run deterministic code on some of those, and those cost a lot less than a Prop 1).

    While the COG idea is flexible it is not cost effective (a UART takes up less silicon space than a COG + RAM). Add to that the old process technology being used (which means your die cost are 4x every generation you are behind.

    And then there is performance, even the 62 cent ARM runs rings around the Prop 1. Prop 2 might be about the same, but it is still vaporware, and now getting way too late to market. Maybe the Prop 2 will be good for education, but just too expensive, too slow, and too late for the general market or I think even the hobbyist market.
  • Invent-O-DocInvent-O-Doc Posts: 768
    edited 2014-09-24 19:21
    Ok Brucee, you are entitled to your opinion. Then why not hang out on the ARM forums instead of here?
  • jmgjmg Posts: 15,155
    edited 2014-09-24 19:46
    brucee wrote: »
    You can even get multicore ARMs (I've seen 4 CPUs, and you can run deterministic code on some of those, and those cost a lot less than a Prop 1).

    Which 4 core ARMs come for less than a P1 ?
    brucee wrote: »
    And then there is performance, even the 62 cent ARM runs rings around the Prop 1. Prop 2 might be about the same, but it is still vaporware, and now getting way too late to market. Maybe the Prop 2 will be good for education, but just too expensive, too slow, and too late for the general market or I think even the hobbyist market.

    There are a great many markets, and a 100pin part, is never going to be on the same shortlist as an 8 pin part.

    The P2 can find a niche in the areas between Microcontrollers, Microprocessors and small FPGA
  • potatoheadpotatohead Posts: 10,260
    edited 2014-09-24 21:37
    Additionally, it does not need market dominance of any kind. It only needs to be successful enough to fund the P3, Parallax and the related ecosystem.
  • bruceebrucee Posts: 239
    edited 2014-09-24 22:23
    Ok Brucee, you are entitled to your opinion. Then why not hang out on the ARM forums instead of here?

    OK, why am I here in the first place. In the mid 1990s I was working at a startup doing graphic Vacuum Fluorescent Displays (something no one in the US had seen much of, but Japan was playing with them). Everyone knows of VFDs, if you saw a blue-green display in your microwave about the time, that was a VFD. Anyway we were treading new ground and there was a lot of basic physics/research we were doing and we needed to do lots of characterization. At the time I had a little 68HC11 based 4x6 inch board to do various control functions. It worked quite well, but it was programmable in ASM language, and everytime a change was needed it was Bruce can you do this? One day one of my colleagues (a physicist not a programmer) said have you seen one of these BASIC stamps? I took a look, and heck yeah it could do a lot things we needed to get done and it was programmed in BASIC, so I was off the hook. Over a few years, we probably put about 100 into use in various test or burnin functions.

    So I came to like the BASICstamp for what it was good for at the time, and really liked the idea of Parallax training the next generation of engineers. Time marches on and that VFD display company got chewed up as the vendors for cellphone LCDs dropped the bottom on prices just to pay for their sunk capital costs as displays got bigger. Even though if they had started at the same time VFDs would have won, as all they consist of are arrays of wires.

    Back to the subject, price and performance are measurable and not opinions. Any single core $8 ARM runs about 50 times faster than a Prop 1 CGU. And some of those $8 ARMs are dual cores, so really no comparison.

    As for the FPGA market, the leaders (X and A) really don't care much about the low volume customers (read hobbyists), so prices there are high. Most of what I do with FPGAs uses high end chips and we maybe build 10 boards, so we don't care that the FPGA costs $100. It's performance improvement over a Prop is many orders of magnitude faster. X and A make their money at Cisco building lots of routers with hardware acceleration based on FPGA. So while it is opinion, the Prop 1 or 2 are not competition for any FPGA. Maybe there is a market segment for one of the secondary FPGA makers, if they could do an Arduino like thing (the other A and L companies might be interested).

    My concern is Parallax is headed down the same path as the Prop 1 now 9 years ago. At the time the BASICstamp ruled the DIY/education market, but the Arduino had just come out. At the time I was talking to one of their distributors who admitted they didn't understand it as it seemed too limited in memory and performance. But if you look at what DIYers were doing with BASICstamps, it had enough memory and speed to do all that, and at 1/2 the price.

    The the Prop 1 came out, and as I read the specs, I said to myself 8 CPUs kind of cute, and use some of them to do peripherals (makes some sense like the Scenix parts). But then it had non-contiguous memory (nightmares of x86 days), no 5V tolerance (really?), and programmed in a different language and that was interpreted which means a 30x performance hit. No wonder Arduino dominates the DIY/education market today.

    Back to opinion, it seems to me the Prop2 will be way too late and cost way too much. Is there some small niche for it out there; maybe. But the assumption that millions will be sold (I'll quote Ken on that) won't happen when compared to RasPi, Edison or other boards coming on the market. I can understand a business plan with the assumption you will sell millions a year, but reality out there is the competition. Parallax has a wonderful user base, and had good contacts to take advantage of Chinese manufacture. But the reality (opinion) is that CPUs are commodities now (what else would you call 50 cent parts), it seems foolish to chase that market having spent $4M (quoting Ken, I think it is larger) to get into the business of selling an obsolete mid $teen part. Maybe if it came out 5 years ago it might make sense, though I doubt it.

    And that is my 2 cents, ban me from the forum or whatever...
  • evanhevanh Posts: 15,340
    edited 2014-09-24 23:19
    brucee wrote: »
    So I came to like the BASICstamp for what it was good for at the time, and really liked the idea of Parallax training the next generation of engineers.

    Right from the start when the BASICstamp first came out I remember laughing at it as just your typical education tool that was Smile for real control. It had no multitasking capabilities and was slow to boot. It taught simplistic batch processing, just like the computing books. I never even took a look at Parallax.

    When I heard about the Propeller I totally changed my mind and immediately participated. Here was something obviously targeting open ended development.

    The situation hasn't changed much imho. In theory, ARM chips running Linux can do decent multitasking but it's never that easy to develop nor fast with the multilayered wrappers.
  • rod1963rod1963 Posts: 752
    edited 2014-09-24 23:35
    Electrodude

    The Prop is a non player on the Stamp board/DIY scene. Parallax doesn't officially offer a StampBasic on the Prop for obvious reasons - it would eat into BS2 sales. The fact is a lot of BS2(I started using them in 2000 for work) users like myself waited years in the hopes Parallax would release a BS3 and instead got ignored by them. We went to other vendors who offered a much more up to date and price competitive product line. The more artistic types went to the PickAxe or one of the many Arduino variants and that is continually adding new micros to it's ecosystem.

    It's too bad Parallax didn't offer a BS3, then a BS4 as newer micros hit the scene to capitalize on the Stamp line when there was a interest for them, if they did, I don't think Arduino would have ever taken off the way it did and become the dominant hobby/DIY player.
  • jmgjmg Posts: 15,155
    edited 2014-09-24 23:45
    brucee wrote: »
    ... But the assumption that millions will be sold (I'll quote Ken on that) won't happen when compared to RasPi, Edison or other boards coming on the market. .

    Those are both Board-level products, running an operating system. - rather a Chalk and cheese comparison.
    My expectation is that P2 sales will be helped by the RasPi, Edison or other boards, not hindered.

    Microchip continue to release PIC16 variants, so clearly they consider there is space 'outside ARM'
  • jmgjmg Posts: 15,155
    edited 2014-09-25 00:35
    rod1963 wrote: »
    It's too bad Parallax didn't offer a BS3, then a BS4 as newer micros hit the scene to capitalize on the Stamp line when there was a interest for them, if they did, I don't think Arduino would have ever taken off the way it did and become the dominant hobby/DIY player.

    What stops Parallax doing that in 2014/2015 ?

    I would nominate the highest-end 5V operation, TQFP48, 32bit uC as a candidate.
    That gives an important backward compatible pathway
    From
    Nuvoton - perhaps a M453LG6AE ?
    Spansion(Fujitsu) perhaps a MB9BF566KPMC (M4 160MHz, 512KF 64KR ) ?

    and maybe also a bottom end, same-form-factor, to give some product spread.

    A couple of good bottom-end modules examples are (selected for 5V operation parts)

    KIT_XMC_2GO_XMC1100 (M0 32MHz, 64KF, 16KR 5 euros Module )

    CY8CKIT-049-42XX ( M0, 48MHz, 32KF, 4KR, US$3 Module )
  • Invent-O-DocInvent-O-Doc Posts: 768
    edited 2014-09-25 04:12
    Good answer Brucee. I was just wondering. You have some valid points, though not everyone has concerns about the things that have been limiting to your applications.
  • David BetzDavid Betz Posts: 14,511
    edited 2014-09-25 04:57
    jmg wrote: »
    What stops Parallax doing that in 2014/2015 ?
    I don't know for sure but it seems to me that Parallax believes that Spin is the next step beyond PBasic and so they have already made the next generation modules. They even once made a product called the SpinStamp which is what makes me think they intentionally left PBasic behind for Spin. Spin is certainly a more capable language so that makes some amount of sense. Maybe they just need a plan for moving BSx users to the Propeller and Spin rather than making follow-ons to the BSx series.
  • Invent-O-DocInvent-O-Doc Posts: 768
    edited 2014-09-25 08:44
    Interesting David,

    Ken is always saying that tons of people are still buying BASIC Stamps. PBASIC sure is easy. Then they went into the Propeller C big time, though apparently that has somehow brought people into Propeller and increased interest in SPIN.
  • David BetzDavid Betz Posts: 14,511
    edited 2014-09-25 08:49
    Interesting David,

    Ken is always saying that tons of people are still buying BASIC Stamps. PBASIC sure is easy. Then they went into the Propeller C big time, though apparently that has somehow brought people into Propeller and increased interest in SPIN.
    I think one reason so many new Propeller users end up being interested in Spin is that whenever they ask a question on the forums existing Propeller users steer them in that direction.
  • RamonRamon Posts: 484
    edited 2014-09-25 09:19
    jmg wrote: »
    The P2 can find a niche in the areas between Microcontrollers, Microprocessors and small FPGA

    There is (There was?) a very small window timeframe to develop the P2. The small FPGAs are currently made in a two-digit-nm process and the P2 is currently being developed in a 180nm. The longer it takes to develop the P2 less chances to make an IC that can be competitive and be sold.

    Data about Altera MAX 10 FPGA (using 55nm Flash process):

    - Integrated 55 nm dual Flash (dual boot + user flash)
    - Integrated ADC, DSP, DDR interface, Oscilator, PLLs, LDOs
    - from 2K to 50K Logic Elements (P1V uses around 15K LE in Cyclone IV/V).
    - Block RAM from 108Kb to 1,638 Kb
    - Standby power 10mW - 100mW. Max operational power: 1W
    - 144-EQFP package
    - Price : unknown yet

    First images: http://www.eeworld.com.cn/FPGA/2014/0924/article_3505.html
    Leaked confidential document: search google for its own name "Altera 10M08"
  • jmgjmg Posts: 15,155
    edited 2014-09-25 13:25
    David Betz wrote: »
    I don't know for sure but it seems to me that Parallax believes that Spin is the next step beyond PBasic and so they have already made the next generation modules. They even once made a product called the SpinStamp which is what makes me think they intentionally left PBasic behind for Spin. Spin is certainly a more capable language so that makes some amount of sense. Maybe they just need a plan for moving BSx users to the Propeller and Spin rather than making follow-ons to the BSx series.

    Perhaps, but that 'let the customers relearn/recode everything' stance is rather optimistic, and history shows how well that has worked.. it also is rather contradicted by the large efforts to release GCC for Prop.

    Smarter would be to expand the BasicSTAMP language, and align with a compiled PC version like FreeBASIC, and the importance of backward-compatible superset is also why I selected only 5V 32bit cores.

    32b & 5V choices are expanding, so the product-choice timing of this is now very good.
  • jmgjmg Posts: 15,155
    edited 2014-09-25 13:40
    Ramon wrote: »
    There is (There was?) a very small window timeframe to develop the P2. The small FPGAs are currently made in a two-digit-nm process and the P2 is currently being developed in a 180nm. The longer it takes to develop the P2 less chances to make an IC that can be competitive and be sold.

    Data about Altera MAX 10 FPGA (using 55nm Flash process):

    - Integrated 55 nm dual Flash (dual boot + user flash)
    - Integrated ADC, DSP, DDR interface, Oscilator, PLLs, LDOs
    - from 2K to 50K Logic Elements (P1V uses around 15K LE in Cyclone IV/V).
    - Block RAM from 108Kb to 1,638 Kb
    - Standby power 10mW - 100mW. Max operational power: 1W
    - 144-EQFP package
    - Price : unknown yet

    First images: http://www.eeworld.com.cn/FPGA/2014/0924/article_3505.html
    Leaked confidential document: search google for its own name "Altera 10M08"

    Great info, thanks, I'll start a thread in P1V on the MAX 10.
  • prof_brainoprof_braino Posts: 4,313
    edited 2014-09-25 16:28
    brucee wrote: »
    $8 low chip cost???

    Sorry, I meant low chip COUNT. I don't care about cost, as long as its under $10. You should see my car.
  • evanhevanh Posts: 15,340
    edited 2014-09-25 18:47
    jmg wrote: »
    32b & 5V choices are expanding, so the product-choice timing of this is now very good.

    I assume most are dual powered? Eg: 5V I/O + 1.8V core. I think that's what I'd prefer for Prop2.
  • jmgjmg Posts: 15,155
    edited 2014-09-25 20:01
    evanh wrote: »
    I assume most are dual powered? Eg: 5V I/O + 1.8V core. I think that's what I'd prefer for Prop2.

    Yes, the process core is not 5V, so there is usually an on chip regulator, and the finer/smaller process is used for the Logic core.
    Some parts have a Vcore decoupling pin, but the actual core voltage used is up to the chip vendor.
    ( Infineon manage to avoid a VCore pin, but Cypress have one )

    The Infineon parts are well under $1, and Cypress ones ~ $1, so there is no noticeable end-price impact.

    Relative to the lowest cost ASIC, I think this needs a thicker oxide in the IO areas, thinner for the core, and some proper level shifters to give lowest power voltage translation.
    I believe the P2 OnSemi process does not allow for this ?
    (so gives 1.8V core and 3.3V Vcc limit )

    Addit: I see data is up on the new Cortex M7 devices
    eg ST STM32F756NG - this specs 166 x 5V tolerant IO and 1.7-3.6 Vcc.
    It also has 200MHz timers, HS USB, CAN, Ethernet, QuadSPI, 12b ADC
    Packages :
    LQFP100 (14
  • rjo__rjo__ Posts: 2,114
    edited 2014-09-25 20:21
    Boy... you guys sure worry a lot.

    The P2 will be far from a niche player. In fact, the FPGA version will probably approach niche status. I don't expect to buy my first P2... I sort expect it to be free... in the form of an FPGA image. After that, I expect to license it as sort of a soft "hardcopy" with entrails open to my various tinkerings.... and finally after all of that I expect to buy lots of P2 demo boards... I intend to mount the first one in the middle of a large mirror and hang it on my wall... that way, when I comb my hair in the morning, I can say "pure genius" and not have to lie about it as I occasionally do now:)
  • evanhevanh Posts: 15,340
    edited 2014-09-25 20:35
    Oh for the days of hair combing ... hehe, didn't comb my hair even then. :P
  • Erik FriesenErik Friesen Posts: 1,071
    edited 2014-09-26 05:28
  • Dave HeinDave Hein Posts: 6,347
    edited 2014-09-26 06:50
    In the grand scheme of things the Prop does exist in a niche market. Part of that is due to the proprietary Spin language. Part of that is due to the interrupt-less multi-core architecture that uses cogs for soft peripheral drivers and requires polling loops. The video circuitry is a nice feature of the Prop, but the small hub RAM limits it usefulness.

    I see the P2 as an enhancement to the P1, and not a competitor to any chip from other manufacturers. People who currently use the P1 will use the P2 for applications that require more processing power and/or memory. The P2 does need to have a reasonable price tag or people won't use it. I think the price needs to be below $10.

    It seems that the price of the P1 will need to drop after the P2 is introduced. I'm hoping that the P1 will drop to $5 or below.
  • User NameUser Name Posts: 1,451
    edited 2014-09-29 20:33
    rjo__ wrote: »
    Boy... you guys sure worry a lot.

    Yeah, roger that. I've never figured out what brucee's negativity and hand wringing is all about, but it certainly is tiring.
  • SeairthSeairth Posts: 2,474
    edited 2014-09-30 10:08
    Just got the official Max 10 release email from Altera:

    http://www.altera.com/devices/fpga/max-10/max-10-index.html
  • TubularTubular Posts: 4,646
    edited 2014-09-30 14:12
    Thanks Seairth. Digikey and Mouser have small stock of the dev kit and 10M08 devices, with mouser being a little cheaper.
Sign In or Register to comment.