Ship it!
Baggers
Posts: 3,019
There's a saying in my industry, for things that have over run their dev time.
And it's something that P16X64A needs.
All the big users of Parallax items, want the next iteration of the Prop, and they want it ASAP.
Even in it's simple state, of no HUB-EXEC, no quad-read, no super cog, no this, no that, it's still far outperforms the P8X32A, and I for one, and I know a good many others of you, especially the big users, would love to have this sooner, and would be quite happy to have 16 cogs, at whatever speed, with 512KB hub-ram and whatever Chip has it doing by the end of this week's rejig from the ground up.
Get it shipped, then by all means, start work on the next iteration, with HUB-EXEC or whatever needs or can be added.
A lot of P8X32A projects were held down by IO count, and HUB-RAM limitations, these have already been far surpassed, and we even have a good speed boost too! so for me, that is more than enough for me, for the next version of the Propeller, yes, we all got greedy wanting this and that to go into the chip, and yes, we were all just helping Chip out, as he was after inspiration, but the things just got more complex as Chip competently added the functionality that we all requested, but it came at a big cost, time! We're now further down the line, and no closer to the goal of having a shippable item.
In my honest opinion, I think Chip and Ken need to just get this new chip out, we're now in May, and there's not been a shuttle run, so even if Chip finishes the simple version by the end of this coming week, it's still going to be a long while before it's out on the shelves.
The problem is, each week waiting, is a week that we're not making boards/programs/projects with the new chip, and therefore not bringing money into Parallax to be able to continue development. And I know this because I have had a bit of downtime with work over the past couple of months, yet not had anything to do, because I've been waiting for the next chip, the issue being, I know it's coming! which is awesome, but I don't know what it is, or what it can or can't do, so there's point in starting a project if it's gonna have to be re-written further down the line when the chip gets finalised, I'm not even doing any P8X32A stuff, because I'm wanting to do something with the new one, and I know I'm not the only one who is doing this.
So what we need to do, is stop the requests of "alternatives" and no that isn't an attack at anyone we've all been wanting to help Chip figure out a way to make the P16X64A the best chip ever, but it already is, it's a Propeller! and who knows, if Parallax get it out soon, it could even help them get the sales from it to afford to make the PyXzA in 65nm tech. which would make the next iteration even better, without even changing the design, just by making it smaller will make it faster and allow even more hub-ram, heck it could even just be two P16X64As on one chip! and that would be like a free version upgrade.
Anyway, that's enough from me, I have to go take my girls out shopping.
P.S. this message isn't a dig at anyone, so please nobody get offended in anyway, it's the weekend, it's supposed to be fun time! this is just my personal thoughts on the issues at hand!
And it's something that P16X64A needs.
All the big users of Parallax items, want the next iteration of the Prop, and they want it ASAP.
Even in it's simple state, of no HUB-EXEC, no quad-read, no super cog, no this, no that, it's still far outperforms the P8X32A, and I for one, and I know a good many others of you, especially the big users, would love to have this sooner, and would be quite happy to have 16 cogs, at whatever speed, with 512KB hub-ram and whatever Chip has it doing by the end of this week's rejig from the ground up.
Get it shipped, then by all means, start work on the next iteration, with HUB-EXEC or whatever needs or can be added.
A lot of P8X32A projects were held down by IO count, and HUB-RAM limitations, these have already been far surpassed, and we even have a good speed boost too! so for me, that is more than enough for me, for the next version of the Propeller, yes, we all got greedy wanting this and that to go into the chip, and yes, we were all just helping Chip out, as he was after inspiration, but the things just got more complex as Chip competently added the functionality that we all requested, but it came at a big cost, time! We're now further down the line, and no closer to the goal of having a shippable item.
In my honest opinion, I think Chip and Ken need to just get this new chip out, we're now in May, and there's not been a shuttle run, so even if Chip finishes the simple version by the end of this coming week, it's still going to be a long while before it's out on the shelves.
The problem is, each week waiting, is a week that we're not making boards/programs/projects with the new chip, and therefore not bringing money into Parallax to be able to continue development. And I know this because I have had a bit of downtime with work over the past couple of months, yet not had anything to do, because I've been waiting for the next chip, the issue being, I know it's coming! which is awesome, but I don't know what it is, or what it can or can't do, so there's point in starting a project if it's gonna have to be re-written further down the line when the chip gets finalised, I'm not even doing any P8X32A stuff, because I'm wanting to do something with the new one, and I know I'm not the only one who is doing this.
So what we need to do, is stop the requests of "alternatives" and no that isn't an attack at anyone we've all been wanting to help Chip figure out a way to make the P16X64A the best chip ever, but it already is, it's a Propeller! and who knows, if Parallax get it out soon, it could even help them get the sales from it to afford to make the PyXzA in 65nm tech. which would make the next iteration even better, without even changing the design, just by making it smaller will make it faster and allow even more hub-ram, heck it could even just be two P16X64As on one chip! and that would be like a free version upgrade.
Anyway, that's enough from me, I have to go take my girls out shopping.
P.S. this message isn't a dig at anyone, so please nobody get offended in anyway, it's the weekend, it's supposed to be fun time! this is just my personal thoughts on the issues at hand!
Comments
Agreed, this has been far too long n the making, I know Chip encouraged the forum's input and some great discussion came out of it but in reality it has just delayed the project even more.
I think most of what you say fits in with what a lot of P1 users come up against with the I/O and HubRAM, this is where Prop 1.5 would have been a sure fire winner.
But I think your words will fall on deaf ears matey and you will no doubt be accused of being a 'naysayer' and not backing up your argument with a technical submission will most likely mean it will be ignored anyway ;-)
Ultimately there is a limited window of opportunity for Parallax and the P2 (or whatever its called this week), let's hope they make it before it closes.....
Coley
Understood and agreed!
Err, and how exactly are you determining that this 'window' is so limited and rapidly closing?
Is this coming from personal experience as a large volume customer, or some conversations with such?
Because, otherwise, I'd call shenanigans on that.
Yes, everyone and their mother wants something newer, and they want it 3-4 years ago.
To me, it sounded like some of the real features that would have helped the P2 compete were found to be a bit more difficult than originally thought, and Ken/Chip have decided to nix them in favor of shortening the timeline.
One can hope there is actual revenue to come in from it to support a future revision in a year or two with those features.
However, after the P2 starts cannibalizing sales of the P1, and potentially increasing its market penetration, it might be 1,2,4 years before there is a 'next' revision.
Might be better to let Chip get to a good 90%+ completion, and then perhaps take another look at quads and hub sharing with a lot less stress involved. Hypothetically, would an extra month delay be worthwhile for something that will probably be the main product for the next 2+ years?
My opinion is based on the rapidly increasing number of multicore processors out there, mainly ARM but increasing at a rapid rate.
When I first stated using P1 many years ago, multicore processing in anything other than PC's was fairly unique, not so much these days.
Also video generation on chip is becoming more commonplace (especially with ARM SOC), it's just not a unique feature anymore.
I remember that same question being asked months ago, most thought it was a good idea then and look where we are now!
I also wonder how many of the large volume customers really need all of the added razzmatazz that we here on this forum keep asking for?
I would wager it's more likely they need more I/O, more speed and more HubRAM, happy to be proven wrong though ;-)
Don't get me wrong, I'm not desperate for P2, I'll wait as long as it takes but if we keep on opening up the design for input it will never be completed.
I'll happily keep using P1 in my existing products and where I need more power I'll just use two or more.
My products aren't price conscious and they aren't big volumes either so using more than one isn't a problem for me.
I would imagine this is more of a problem for the volume users though.
I'm certain everybody's input has been with the best of intentions, i.e. to make it the best microcontroller out there and that has to be applauded.
Of all of the 'features' put forward Hub Execution made the most sense to me, I'm not a C user but I understand the need for C compatibility and therefore performance, let's hope LMM get's us somewhere near it's certain to be better than P1 anyway.
I truly don't know what's best for Parallax - a basic x64 sooner or a fancier x64 later? But I can say that on a personal level it doesn't matter. The setbacks with the Px provided an opportunity to tackle projects that were too long ignored. The most recent one involves implementing a J1-like engine on an Altera FPGA. Between that and Clyde Phillips' outstanding FISH forth for the LPC1114, it will be a while before I'm desperate for an x64.
In this part of the world, the phrase is a little more blunt. F{inish} It, Ship it. But it translates the same way.
Yes, Please. Do NOT make continued features requests which are proving to be a distraction to the designers. Let's keep this forum open for actual Project/Product idea for which the chip's existing specifications have been defined. There will be plenty of time for a Propeller III, IV, and V and personally I hope to be around to get collect them all.
Again, Thank you Baggers.
Don't be mad at the OP, be concerned with the situation that has developed.
Could a "mod" please sticky this thread?
Jeff
/edited for forum decorum reasons...
That said even if Chip releases a perfect design image for the FPGA's next week, it won't be until next year before they can do a commercial roll-out. The FPGA version has to be thoroughly tested, compilers written and validated, software libraries polished, App notes and white papers written, documentation in general.
And this assumes the shuttle runs don't discover any flaws that need a do-over and another shuttle run.
And will it eat into P1 sales? Probably, especially those who feel the current offering is I/O and memory constrained will migrate to it.
But the $64,000 question is: Will the P16x64 be good enough to get Parallax enough volume design wins to allow them to recoup their investment..
The thing is, the Prop really isn't unique anymore. The PSOC does a lot of it sans the multicore. There's Xmos and it's offerings, Freescale and TI have multicore PPC and ARM's respectively that are very I/O capable. FPGA's and CPLD's are a lot more powerful than they were when the P1 was introduced.
One last thing: Heater's idea of stuffing a ARM into a Prop seems to have caught on at Xmos seeing as they are developing a multicore with a M3 core to handle tasks not suited for the Xcore processors. It looks like even they ran into certain design constraints.
Amen!! Even if the only difference between the P1 and P2 is the extra hub ram, IO and increase in speed it'll be a significant improvement!! Get this thing done!!
I know I'm ready to get going on my next prop project, no matter what the chip can do, I'll work with what it does, and I know it'll be amazing, because quite frankly, the original was amazing and was always a joy to program on.
We all have done some amazing things with the Prop1, some things that at the time were thought not possible, but we did it. so even at 4* the speed, and 16* the hub ram and 2* the IO and 2* the cog count, we're onto a winner of a chip.
David, I'm not saying any of the features we ALL have requested of the past few months/years have been a bad idea, heck it's inspired Chip, and that in itself is an awesome fact! all I'm saying is yeah by all means release the next gen prop next year or two away from now, but get something simple yet awesome out now, we already have simple yet awesome! then get some money back from sales of this new chip, to help fund either just more time to add features for the next release, or heck who knows, it may do really well, and put Parallax in a position where it could even afford to do the 65nm tech for the next iteration, if not, they'd have still had time to put more features in like hub-exec or quad read/writes or whatever, where even just that would be a spec update and enough to warrant an updated chip release. And allow users to update their devices, rather than have them potentially unupdate-able for many years, thus possibly missing out on more chip sales.
Jeff, no worries, and we say that here too, it's just that on a forum, I try to show some decorum
Rod, yes I think it will be, Ken has even said that many of Parallax's big users have stated they want a newer version, and with the addition of another 32 IO users can add more SDRAM if they so desire, the HUB-RAM is now a massive 16* more! able to store double buffered displays of an 8bit 320x240 bitmap screen, or single buffered 640x480 8bit bitmap, and the cogs are faster so can do more calculations, and if the cordic hub stays in, that'll vastly help improve the maths capabilities of the chip, so can be used in even more designs, plus with the IP protection it will be safer for businesses to put it into their designs without too much worry of people just cloning it.
4x5n, exactly lol
user name, having an x64 now, doesn't mean we won't get an x64 later, the aim is to get a fancier x64 later too, just that we'll all get to play with something now, and hopefully help get Parallax some cash in from sales of the new chip to help continue funding the fancy x64.
RossH thanks bud haha
Sorry if my decorum has been slipping.. I've been hoping that part of my income this year would come from Propeller II projects. I'll try to suck it up a bit, swallow hard and remain my pleasant, friendly self.
Jeff
Breaking the habit is hard, I know but it has to be done, even in the games industry, even operating systems have updates, people are used to updates, heck they even prefer updates.
Look at some big games Call of Duty for example, it has gone through many iterations over the years, releasing a new game every year or two, it's still technically exactly the same gameplay, it just has different features, new weapons etc, yet people buy it every year! same with Football, and hockey etc. people are happy to have the latest addition each year or two, no matter how small it is.
It's something new, that ups the bar each year, for the sales teams to push to the end users.
Heck, if we'd have had the P8X32B back in the christmas of the Eindhoven trip, we'd have all done more stuff, and helped push the parallax chips out there.
And now with the P16X64A we'll push them again, and again when the old P2 design comes out, in whatever strain it appears, or even if it's something completely different, people will want that also! so you're getting double sales at a minimum.
The thing is, we're all here because at some point we've loved programming the Prop. be it professionally or in the hobby stage. Some even started in hobby stage and moved it to professional after making boards with the prop.
No worries Jeff
Yes I know you're still using the P8X32A so it's technically not losing sales, the issue is it quite easily could have.
Although RAM, pins and speed are the most-requested features, I'm fairly certain that I read that some volume purchasers could benefit from or need more cogs. To me, additional cogs offers sooooo much more flexibility over the P1, and makes some things possible that might otherwise not be in one chip. Additional cogs permit the flexibility that soft peripherals need, and 8 is quite limiting, especially when some objects take 2, 3 or 4 cogs (and some still will even with the more powerful chip). I just think it takes a lot of hubris to try to open up a discussion at this point when Chip and Ken have decided on and announced design goals that they are happy with and Chip is already "down in the dungeon" hard at work on implementing those goals.
The refactoring of the new chip was totally understandable, as Chip/Parallax are still developing experience with the tools and the new fab process/house. But if Parallax were to significantly reformulate the design goals at this point (after all that has happened) without a very strong reason for doing so, such as to scale back to 8 cogs (remember, 32 were briefly considered), I would actually lose a significant degree of my confidence in them (maybe 25% or more). Fortunately, I don't think Parallax is considering any such radical changes, but I worry that the lack of a voice countering proposals for other designs could be taken as a sign that we don't care or don't think it could impact Parallax's credibility. Please don't misunderstand: I don't mean this as a threat and I know that Parallax welcomes, nay encourages, our input/ideas, but I can't see how I wouldn't lose a significant portion of my high degree of confidence in Parallax if additional changes are made without strong justification, even though I hope/plan to apply the new Propeller where possible/appropriate. Hence this message of concern.
Yes, even more worrisome and in the spirit of the original post is concern about the release timeline. But additional changes might well make for additional delays (in addition to possibly reduced functionality/applicability). Folks, we have a great design on the drawing board! And it's now in the capable hands of Chip & Co. Yeah, I suppose there's room for meaningful, thought-provoking discussions as to, for example, whether 8 cogs represents some kind of magical sweet spot and anything more is a deviation from the ideal. But the ship has sailed! It sailed about three weeks ago. And if Chip has entertained the thought of 16 cogs at various points over the last few years (yeah, including the big discussion where we had to choose between having more cogs or more RAM, i.e., couldn't have both), then I'm inclined to believe that such a so-called sweet spot spans a considerable area. So, in conclusion, my two cents would be design the chip as announced and ship it as soon as feasible.
And yeah ship it as soon as it's feasibly possible, after it's ready and tested of course, our DE2s and DE0s are poised ready!
Chip agrees, and I'd say Hubexec has made the cut from this
Chip : ["I think those features are important enough, though, that the disruption they cause must be accommodated."]
Cool, Yup Great....
Ship what exactly ? There is no it at the moment.
Skip testing ? Sure, That gets it faster and Ship it does not mention if working matters, so I guess the timeline trumps that too... unlike SW, Silicon cannot download a batch onto your field-tester-customers.
I'm totally with you on this. Ship it, ship it...all together now, ship it...louder...I can't here you...SHIP IT!
Or as I have said recently "Get me it...NOW!"
All I ever wanted from a PII was, more RAM, more pins and more speed. And importantly I wanted it 2 years ago!
The current design, as pared-down as it, is already exceeds my wildest dreams regarding a Propeller 2. *
1) More RAM -- check
2) More Pins -- check
3) More Speed -- check
4) More Cogs -- check, that's a bonus.
All present and correct. Ship It!
* Well except that really wild recurring dream about the P3 with 64 bit wide COGs, allowing megabytes of COG RAM space, floating point support and a JavaScript interpreter instead of Spin...
Keep It Simple Stupid (KISS). Just more RAM, more pins, more speed, more COGs and we are good to go.
Anything else adds time. This last "quick tweak" whilst waiting for the next shuttle run has taken what? Over a year now!
No more.
As Ken recently said, more frequent releases would be a better way to go. So start now with KISS. Think about and add extra junk at leisure, if it even turns out that anybody still really wants it.
And no, I don't think anyone is suggesting skimping on testing. Many of us here have FPGA boards at the ready for exactly that. And obviously the actually chip when it arrives needs a good looking at.
Personally I have not bothered much with exercising PII on my nano board. Things were just changing too much all the time.
OK, I agree entirely with all of that.
Those are good points.
However after 6-8 years, the truth is, a few months difference doesn't matter unless Parallax is already having financial disfficulties.
I think my comments are still valid though, when the Prop is essentially being cut feature and performance-wise, again and again in the rush to get something, anything 'out', how does something less and less performant save anything except for the very short term?
Other people are starting to do the who multi-core thing, including dedicated video, etc as you rightly said.
.
I agree mostly completely with that.
I just fail to see where/how Parallax will be able to suddenly start dumping considerable R&D $$$ into new revisions on a yearly or biennial cycle, unless the P2 somehow becomes some large hit, or margins are far better on it.
If thats true, just seems like pushing the envelope for this revision should be strongly considered as we can see the competition on the horizon, and more rapid revisioning is not a given.
I know a lot of folks who are hobbyists are going to dislike this idea. Anything other than 'better than P1" and 'NOW" is just not acceptable to them. Parallax making a business decision based on the multitudes of "NOW" from them, is potentially more costly than any delay, in the longer run.
Anyway, here and now, this moment is the "long term". It is the end of the long term that was envisaged starting 8 years ago!
As many of said Parallax and the Prop cannot compete with the likes of the ARM. Better not to even go that way. They have to provide something unique. So far the Props Unique Selling Points (USP) have been it's simplicity and it's simplicity and most of all it's simplicity.
That simplicity and flexibility should not be overlooked or underestimated.
++Very well said.
However, I think many people are still expecting a product in store in time for Christmas or earlier.....
Chip may vary well be able to solve the speed bumps of Quads/hub exec during the time it takes to get the design and FPGA to 99%.
I can just see the crowd pushing to avoid it with the expectation that it will somehow result in the P2 getting to them months and months earlier.
In any discipline, on any project, its always the final stretch where frustration leads to mistakes and a short-sightedness as the goal posts come into view.
John Abshier
Daniel
You may say three months doesn't make much difference,
But 3 months could potentially miss a shuttle run, which could make 3 months into 5-6 months, which then people say wow we have lots of time left, can we add this. And you can't deny this, as it's happened all along the P2's development.
KISS keep it simple stupid, ship it soon as possible, ( and always AFTER testing, as nobody has ever said skip testing. ) then you can add more bits, yes people will want that too, but in the mean time you'll have had 6+ months of giving your big customers what they've wanted for the past few years! more cogs, more hub-ram, more IO, security, yet keeping the same simplicity! win win situation if you ask me!
What big customers. Parallax has no large design wins with the Prop.
Shuttle runs aren't production runs. They are there to allow the company to test drive the real silicon and do revisions if necessary. Rush it and you end up with the Pic32MZ or AVRx series with a fat load of errata and non-functional silicon and ****** off customers. Microchip can eat those mistakes and keep on going, Parallax can't.
The bottom line for Parallax is that it has to recoup the $4 million + they've invested in it, they can't make it with hobbyists they need wins that allow them sell hundreds of thousands of units per year to do that.
You can't even being to make those kind of sales without a robust software development suite, a variety of eval boards for different needs(one won't do it), vetted software libraries, complete documentation, app notes, white papers, trained support engineers to handle calls from prospective clients, etc. All this takes time to set up and you cannot skimp. This will all take close to a 9 months to a year to do.
Now if they screw up the design by short cutting the Design or T&E phases or by releasing Beta level software to shave off a few months here and there, those wins won't be forthcoming and that will be the end of the line for the Prop. Even then there is no guarantee the P16x64 will be competitive. This isn't BasicStamp territory, Parallax is playing in the big boy's sandbox and they can't afford bungle it by speeding up the process too much.
Maybe the applicable quote is "perfect is the enemy of the good".