Shop OBEX P1 Docs P2 Docs Learn Events
Consensus on the P16X32B? — Parallax Forums

Consensus on the P16X32B?

RossHRossH Posts: 5,462
edited 2014-04-13 03:03 in Propeller 2



UPDATE: Chip has started a new thread on this chip
here.

All,

In this thread about the P2, Chip proposed a new chip, tentatively called the P16X32B:
cgracey wrote: »
So, what about this:


100 pin 14x14mm exposed thermal pad TQFP package (Tja=20) with internal down-bonds to GND, so no pins needed for GND

16 x 1.8V VDD pins, at four per side, with internal down-bonds for GND
XI, XO, RESn, BOEn pins
64 I/O pins with a unique 3.3V VDD pin for every 4 pins (and internal down-bond for GND) - this is important for analog and high-speed switching
(that makes 100 pins, not including 32 internal down-bonds for GND)


16 two-clocks-per-instruction Prop1 cogs
256KB hub RAM with simple round-robin cog access - this maintains the same 8:1 instruction:hub-cycle ratio as the current Prop1
200MHz clock - cogs run at 100 MIPS, CTRs at 200 MHz
1600 total MIPS - 10x faster than current Prop1 chip

Center of die is only 4.3 x 4.3mm. Die is 5.4 x 5.4mm with pad frame, or 29.16 square mm (54% of current projected Prop2 die size)


This chip could be called P16X32B, as per current convention.

Manufacturing cost would be ~$1.90, assuming $950/wafer, 950 die per wafer, 80% yield, $0.50 per package, $0.15 total testing. 1K piece price would be ~$6.00.

This chip would behave like Prop1 with twice the cogs, all running 5x faster, and with 8x the hub RAM, plus 2x the I/O pins with new analog capabilities.

This would be really easy to make happen and the chance of latent bugs would be very low.

A little later, Chip posted about another chip, which I will call the P32X32B:
cgracey wrote: »
This would be a very simple chip to complete. It would be kind of a relief, for now.

Prop2 still needs a serdes and some thorough testing. I'd hate to have another false start.

I could see making a 32-cog 3200 MIPS Prop1 w/64 analog I/Os and 512KB RAM for now and then continue to get Prop2 ready for a smaller process.

If I felt some consensus here about doing that, it would be pretty easy to attack that project in short order.

These posts define an envelope for a P1 variant that Chip believes would be quick and easy to achieve.

Since that thread is dedicated to the P2, and this new proposal is not intended to either replace or derail the current P2, I thought I'd start another thread specifically to gauge the level of consensus for a P16X32B, a P32X32B or some other P1 variant within the same envelope.

This is not a thread about the features of such a chip, or bewailing the current state of the P2 development, or complaining that this new chip might undermine the P2 - any further discussion on these subjects should continue in the original thread. But I also decided not to make this a simple "poll", thread since I do expect and want more than just a "+1" response, and I think most forum members will not abuse that opportunity.

In this thread, please just let Chip know if you want to see a variant within the envelope Chip has said would be quick and easy to achieve, and also whether you would be willing to fund its development (should Parallax choose to use some kind of non-equity based crowd-funding model). If you are against it, by all means say so. Just try to keep it brief in either case. In summary:

If you want a variant within this envelope, vote Yes, and indicate if you would be willing to assist in funding it.

If you don't want to see a variant within this envelope, vote No.

Just to get things rolling, and to avoid any doubt, here are my answers:

Yes I want it, and Yes I would be willing to help fund it.

Let the games begin!

Ross.

Running tally (up to post #267) **

For:
  1. RossH (and will fund)
  2. David Betz (and will fund)
  3. dr hydra
  4. jazzed
  5. Roy Eltham (and will fund)
  6. Peter Jakacki
  7. mindrobots (and will fund)
  8. Bob Lawrence (VE1RLL)
  9. kwinn
  10. william chan
  11. Lawson (and may fund)
  12. Cluso99 (and will fund, preference for P32X32)
  13. ospropdev
  14. AntoineDoinel
  15. W9GFO
  16. msrobots
  17. Ken Gracey (and will inherently fund!)
  18. Ramon (and will fund)
  19. Roger Lee
  20. Brian Fairchild (and will fund)
  21. Baggers (and will fund)
  22. rjo__
  23. John Abshier (and will fund)
  24. 4x5n
  25. FredBlais (and will fund)
  26. LeoD (and will fund)
  27. pgbpsu
  28. tomcrawford
  29. Mickster (and will fund)
  30. KMyers
  31. Electrodude
  32. Publison (and will fund)
  33. base2design (and will fund)
  34. dnalor
  35. hippy
  36. Gordon McComb
  37. Invvebt0O-Doc (and will fund)
  38. JRetSapDoog (and will fund)
  39. NWCCTV
  40. Coley
  41. localroger
  42. Tracey Allen (and will fund)
  43. altosack
  44. T Chap (and will fund)
  45. Dr_Acula
  46. koehler
  47. pik33
  48. Heater.
Against:
  1. Dave Hein
  2. Bill Henning
  3. rogloh
  4. Kerry S


** If anyone feels I have misrepresented their position, just let me know!
«13456789

Comments

  • David BetzDavid Betz Posts: 14,516
    edited 2014-04-04 18:36
    Sure, I'd invest in a Kickstarter for this intermediate chip. I also volunteer to work on any GCC porting that might be necessary to accomodate any instruction set changes from P1.
  • dr hydradr hydra Posts: 212
    edited 2014-04-04 18:42
    +1....build it....it sounds awesome:)

    Can I vote more than once?
  • RossHRossH Posts: 5,462
    edited 2014-04-04 18:46
    dr hydra wrote: »
    Can I vote more than once?

    Only if you intend voting Yes. :smile:
  • jazzedjazzed Posts: 11,803
    edited 2014-04-04 18:46

    I'm all for a bigger HUB RAM, and more COGs etc... before P2 comes.

    But, for which chip do you want consensus? There are 2 described in the opening post.

    RossH wrote: »
    All,

    In this thread about the P2, Chip has proposed a new chip, tentatively called the P16X32B:


    quote_icon.png Originally Posted by cgracey viewpost-right.png
    So, what about this:


    100 pin 14x14mm exposed thermal pad TQFP package (Tja=20) with internal down-bonds to GND, so no pins needed for GND

    16 x 1.8V VDD pins, at four per side, with internal down-bonds for GND
    XI, XO, RESn, BOEn pins
    64 I/O pins with a unique 3.3V VDD pin for every 4 pins (and internal down-bond for GND) - this is important for analog and high-speed switching
    (that makes 100 pins, not including 32 internal down-bonds for GND)


    16 two-clocks-per-instruction Prop1 cogs
    256KB hub RAM with simple round-robin cog access - this maintains the same 8:1 instruction:hub-cycle ratio as the current Prop1
    200MHz clock - cogs run at 100 MIPS, CTRs at 200 MHz
    1600 total MIPS - 10x faster than current Prop1 chip

    Center of die is only 4.3 x 4.3mm. Die is 5.4 x 5.4mm with pad frame, or 29.16 square mm (54% of current projected Prop2 die size)


    This chip could be called P16X32B, as per current convention.

    Manufacturing cost would be ~$1.90, assuming $950/wafer, 950 die per wafer, 80% yield, $0.50 per package, $0.15 total testing. 1K piece price would be ~$6.00.

    This chip would behave like Prop1 with twice the cogs, all running 5x faster, and with 8x the hub RAM, plus 2x the I/O pins with new analog capabilities.

    This would be really easy to make happen and the chance of latent bugs would be very low.



    A little later, Chip posted:


    quote_icon.png Originally Posted by cgracey viewpost-right.png
    This would be a very simple chip to complete. It would be kind of a relief, for now.

    Prop2 still needs a serdes and some thorough testing. I'd hate to have another false start.

    I could see making a 32-cog 3200 MIPS Prop1 w/64 analog I/Os and 512KB RAM for now and then continue to get Prop2 ready for a smaller process.

    If I felt some consensus here about doing that, it would be pretty easy to attack that project in short order.



    Since that thread is dedicated to the P2, and this new proposal is not intended to either replace or derail the current P2, I thought I'd start another thread specifically to gauge the level of consensus for the P16X32B.

    This is not a thread about the features of the P16X32B, or bewailing the current state of the P2 development, or complaining that the P16X32B might undermine the P2 - any further discussion on these subjects should continue in the original thread. But I also decided not to make this a simple "poll", thread since I do expect and want more than just a "+1" response, and I think most forum members will not abuse that opportunity.

    In this thread, please just let Chip know if you want such a beast, and also whether you would be willing to fund its development (should Parallax choose to use some kind of non-equity based crowd-funding model). If you are against it, by all means say so. Just try to keep it brief in either case!

    Just to get things rolling, and to avoid any doubt, here are my answers:

    Yes I want it, and Yes I would be willing to help fund it.

    Let the games begin!

    Ross.
  • evanhevanh Posts: 15,915
    edited 2014-04-04 18:48
    A question for Chip (and excuse me if it's already been answered in the other thread): How come the enhanced P1 is so easy all of a sudden? Wasn't this in the too messy bin a year or two back?
  • Roy ElthamRoy Eltham Posts: 3,000
    edited 2014-04-04 18:48
    I vote for whatever P1B variant Chip thinks is easiest/quickest/safest to make.
  • Bill HenningBill Henning Posts: 6,445
    edited 2014-04-04 18:51
    My preference of the two is:
    Originally Posted by cgracey
    This would be a very simple chip to complete. It would be kind of a relief, for now.

    Prop2 still needs a serdes and some thorough testing. I'd hate to have another false start.

    I could see making a 32-cog 3200 MIPS Prop1 w/64 analog I/Os and 512KB RAM for now and then continue to get Prop2 ready for a smaller process.

    If I felt some consensus here about doing that, it would be pretty easy to attack that project in short order.

    In the other thread I suggested a 64 entry vector, of five bit fielts, indexed by a 6 bit counter running from the 200Mhz clock for assigning hub cycles.

    Default would be 0-31,0-31 (cogid) for who gets the slot, but it could be re-programmed for deterministic timing at whatever grain needed.
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2014-04-04 18:52
    Having just seen this thread after posting my approval in the 5W thread I take the opportunity to repost here as well.
    You have my vote for this proposal, wholeheartedly, but please, refer to this as the P2 which we originally wanted, otherwise a Superfortress is really just a B17 "B" variant. Your proposal quadruples the cogs, x20 speed, etc. This is no variant, it still has "Propellers" like the original but it is a very different beast, as is the still in-flux serdesless nextgen P3 design which when it is finalized and fabricated will be very warmly welcomed.

    EDIT: Some would like this P2 to be enhanced as well but it is already a thumping roaring beast, don't complicate this design, leave the all-nighters and trials and errors for the P3.
  • evanhevanh Posts: 15,915
    edited 2014-04-04 18:59
    evanh wrote: »
    A question for Chip (and excuse me if it's already been answered in the other thread): How come the enhanced P1 is so easy all of a sudden? Wasn't this in the too messy bin a year or two back?

    I guess the answer is it would be derived from the current P2 rather than any part of original P1.
  • RossHRossH Posts: 5,462
    edited 2014-04-04 19:01
    jazzed wrote: »
    But, for which chip do you want consensus? There are 2 described in the opening post.

    True. It was originally about the first one (the P16X32B). But I would also vote in favor of the second one. I'll update the first post to clarify.

    Ross.
  • mindrobotsmindrobots Posts: 6,506
    edited 2014-04-04 19:03
    Here we go again.

    Chip, you are the Chief Visionary and Chief Technical Officer.
    Ken you are the Chief Bean Counter (no offense intended) and Chief Business Realist.

    Between the two of you the future will be decided - INCLUDING THE DESIGN AND FEATURES of ANY chip.

    If there is to be a PaaXbb chip, and a mob funding program, I will support it with whatever cash I have ldftover from buying two Nanos and a DE2 for Prop2 testing. I'll also volunteer time and skills as able and needed.
  • jazzedjazzed Posts: 11,803
    edited 2014-04-04 19:08
    evanh wrote: »
    I guess the answer is it would be derived from the current P2 rather than any part of original P1.
    It would probably be a port of the P1 code from AHDL to Verilog with tweaks.
    @Chip if your watching, can you PLEASE do us all a favor and define a feature set that you can limit yourself to completing? Don't be too aggressive ;-)

    I'm dead serious. If you don't say where you're going, you will never get there.
  • Bob Lawrence (VE1RLL)Bob Lawrence (VE1RLL) Posts: 1,720
    edited 2014-04-04 19:16
    I could see making a 32-cog 3200 MIPS Prop1 w/64 analog I/Os and 512KB RAM for now and then continue to get Prop2 ready for a smaller process.

    Sounds great to me. I vote yes.

    Re:P2 @chip

    We really don't have enough features for the prop2 yet anyway LOL
  • mindrobotsmindrobots Posts: 6,506
    edited 2014-04-04 19:17
    jazzed wrote: »
    @Chip if your watching, can you PLEASE do us all a favor and define a feature set that you can limit yourself to completing? Don't be too aggressive ;-)

    I'm dead serious. If you don't say where you're going, you will never get there.

    Thanks! People value your opinion! I'm starting to feel like Rodney Dangerfield around here! :lol:
  • jmgjmg Posts: 15,173
    edited 2014-04-04 19:18
    It is not possible to vote, without knowing the features, and the 180nm Power envelopes.
    Many different variants have been proposed.

    eg
    In the other thread I suggested a 64 entry vector, of five bit fielts, indexed by a 6 bit counter running from the 200Mhz clock for assigning hub cycles.

    Default would be 0-31,0-31 (cogid) for who gets the slot, but it could be re-programmed for deterministic timing at whatever grain needed.

    Agreed, but I would tweak to add a 'No-COG' mapping for power envelope control
    - Power will still be an important issue on this device.

    I would also expand the RAM x2, and reduce the COGs a little, especially if the device wants to target the large LCD driver business.

    Needed here is a P1E COG to memory equivalent figure.

    Other things should be lifted from P2, like Timers, and HubExec (non threaded) , and I think a device playing in the LCD area needs QuadSPI in silicon, to allow high bandwidth/low power FLASH access, and allow other micros, to access this as a slave.
  • jmgjmg Posts: 15,173
    edited 2014-04-04 19:20
    jazzed wrote: »
    It would probably be a port of the P1 code from AHDL to Verilog with tweaks.

    Chip has compiling a 2 Clock, Dual Port design of the P1.

    I think the P2 Peripheral ADC/DAC/Pins, are implicit in this P1E.
  • mindrobotsmindrobots Posts: 6,506
    edited 2014-04-04 19:22
    Do $10 chips compete in the LCD driver business?? What is currently being used?
  • kwinnkwinn Posts: 8,697
    edited 2014-04-04 19:24
    Roy Eltham wrote: »
    I vote for whatever P1B variant Chip thinks is easiest/quickest/safest to make.

    +1

    To be perfectly honest everything I have done so far has been so simple that the P1 is overkill, but I have used the P1 for it's potential expandability, ease of use, and the help available from Parallax and this forum. There is only one project I would like to tackle that needs more than the P1 provides and the P16X32B exceeds that by a large margin.
  • RossHRossH Posts: 5,462
    edited 2014-04-04 19:30
    jmg wrote: »
    It is not possible to vote, without knowing the features

    Of course it is possible to vote. Chip has defined an envelope in which he believes he could easily and quickly develop some kind of P1 variant (i.e. a P1 with more pins, more Hub RAM and more speed). Before we get bogged down on specifics, let's see if we can achieve a consensus on whether we want such a variant.

    If you want a variant within this envelope, vote yes.

    If you don't want to see a variant within this envelope, vote no.

    I personally don't believe there is any point in arguing about the detailed specifications of the variant (look where that has gotten us so far!) and am willing to leave these up to Parallax.

    Ross.
  • jazzedjazzed Posts: 11,803
    edited 2014-04-04 19:36
    RossH wrote: »
    If you want a variant within this envelope, vote yes.

    If you don't want to see a variant within this envelope, vote no.
    Perfect. Yes. +1

    Maybe that is the poll?
  • David BetzDavid Betz Posts: 14,516
    edited 2014-04-04 19:40
    RossH wrote: »
    I personally don't believe there is any point in arguing about the detailed specifications of the variant (look where that has gotten us so far!) and am willing to leave these up to Parallax.
    I agree with this. Let Chip decide what features can be added to P1. He already has lots to choose from in the current P2 design and I'm sure he has a good handle on which ones (if any) can be safely added to P1E.
  • Dave HeinDave Hein Posts: 6,347
    edited 2014-04-04 19:41
    I vote no. I prefer that Parallax continues working toward the goal of producing the P2.
  • jmgjmg Posts: 15,173
    edited 2014-04-04 19:47
    mindrobots wrote: »
    Do $10 chips compete in the LCD driver business?? What is currently being used?

    Chip has indicated this would be ~ $6, and the reference device I am thinking of is the SSD1963, which looks ~ $5 - but that SSD1963, is basically 'a memory playback' device, with a parallel host interface of moderate speed.
    It lacks even Colour-table compression, and has no fonts.

    A correctly scaled P1E can swallow all of a SSD1963, and easily add Fonts/Icons and Colour-table compression, and use a serial interface to the host device.

    FTDI have an EVE video device, but that has much lower memory, so cannot compete on the same sized LCDs.
    FTDI are making the market aware of Smarter LCD Drivers, which would help Parallax.
  • william chanwilliam chan Posts: 1,326
    edited 2014-04-04 19:52
    We will buy 200 P1B chips a year when it is released.
  • LawsonLawson Posts: 870
    edited 2014-04-04 19:52
    I would buy this, and consider funding it. The analog IO are compelling by themselves. A ton more memory, 2x cogs, and 4x instruction throughput per cog is extra nice. Almost completely backward compatibility is another useful perk. If I was greedy, I'd ask for the multiply instruction as reserved in the P1 machine code docs. (I have one application that could use all the extra MACs)

    Marty
  • RossHRossH Posts: 5,462
    edited 2014-04-04 19:52
    jazzed wrote: »
    Perfect. Yes. +1

    Maybe that is the poll?

    I don't think a poll gives quite enough scope for discussion, but I've updated the first post to reflect this is the intent.

    Ross.
  • David BetzDavid Betz Posts: 14,516
    edited 2014-04-04 20:21
    Chip also mentioned this as a possibility:
    About the Prop1 ideas: I was thinking that hub exec could be realized by having no more than a single cache line that would afford 100% execution speed in a straight line. Also, this would allow the dual-port cog RAM to be cut in half, from 0.292 sq mm to ~0.15 sq mm, since we only need it for deterministic code and variables. Add in the ~0.1 sq mm of cog logic and the cog area drops from 0.4 sq mm to 0.25 sq mm - a ~38% shrink, while affording execution from hub. This is all at 180nm.

    When I went in to look at the old Prop1 code, I was blown away how there is almost nothing there. It doesn't have the creature comforts of the Prop2, but it is as lean as can be. It's so compact that you can just add more of them, rather than inflate them.
  • Cluso99Cluso99 Posts: 18,069
    edited 2014-04-04 20:29
    I vote a resounding YES, and preference for the P32X32B (32 P1 cogs with 512KB).

    I am willing to help funding and think we would be better to do it via the forum and direct to Parallax (no 10% commission).

    Perhaps a possible way would be for us to buy chip(s) from the shuttle run and pay up front. Whether they work or not its our problem and not Parallax's.
    Perhaps $100 / chip ??? I know this does not cover the shuttle run, but if Parallax sold 100 then that is $10K. Might make for a decent quantity to be made on the first shuttle, and if they work we have working chips early. Obviously we need Ken's input, but this is a starting point.

    BTW I agree with Bill's 32 hub slot mechanism - easy to describe, fairly simple to implement (I think).
  • mindrobotsmindrobots Posts: 6,506
    edited 2014-04-04 20:32
    Cluso99 wrote: »
    I vote a resounding YES, and preference for the P32X32B (32 P1 cogs with 512KB).

    I am willing to help funding and think we would be better to do it via the forum and direct to Parallax (no 10% commission).

    Perhaps a possible way would be for us to buy chip(s) from the shuttle run and pay up front. Whether they work or not its our problem and not Parallax's.
    Perhaps $100 / chip ??? I know this does not cover the shuttle run, but if Parallax sold 100 then that is $10K. Might make for a decent quantity to be made on the first shuttle, and if they work we have working chips early. Obviously we need Ken's input, but this is a starting point.

    I'd need a chip on a minimal dev board. A chip by itself does me no good. But I'm sure deals can be struck!
  • jazzedjazzed Posts: 11,803
    edited 2014-04-04 20:37
    Cluso99 wrote: »
    Perhaps a possible way would be for us to buy chip(s) from the shuttle run and pay up front. Whether they work or not its our problem and not Parallax's.
    Perhaps $100 / chip ??? I know this does not cover the shuttle run, but if Parallax sold 100 then that is $10K. Might make for a decent quantity to be made on the first shuttle, and if they work we have working chips early. Obviously we need Ken's input, but this is a starting point.
    IIRC, minimum for a chip-spin is $50K at TSMC for 300nm process. Heard that OnSemi NRE is cheaper (and more available although costlier for production).

    @Rick if the past is any indicator the forum could have 100 new board designs for you pretty quick. ;-)
Sign In or Register to comment.