Sure would be nice to see some "official" input on this thread. Maybe we could get Ken or Chip to weigh in on this discussion after the holiday. There is so much "win" in this concept and I'm really pleased to see it being explored. Now... If we can only get past the chat with coffee stage..
My direction is to use a larger FPGA board, with ARM cores, and add a custom processor (inspired by the P1V and RISC V) between their I/O logic and the ARMs to create a souped up version of the BeagleBone Black. I would LOVE to be able to go to Parallax and shop for a P1V+ or P2V variant to do what I want.
A series of boards based on the Altera Max10 series (as an example) could be made as drop in System-on-Modules that could be designed to be easy to adapt to other systems or uses. Complete flexibility! Education, hobby, custom commercial applications all pulling from the same investment.
Times change, and are changing even faster now. By the time Parallax gets a P2 chip in hand it will be obsolete with little to no real market potential.
Perfect example is their 123 board. Spent a bunch of time and money to
design and manufacture it and now there is an A9 based board for 1/3 the price already
to market while we still wait for theirs.
Some self-contradiction going down there.First a plea for flexible modules, then an example of how that fails to meet price.
The P2 is far from release, so it is way too early to market that, but the P1V is real, and anyone is free to make a module targeting that.
I'm sure if there is enough critical mass for such products, some will arrive.
eg The recent MAX 10 numbers suggest a Board with MAX 10 P1V plus a P1 could give 8 std COGS plus 1-2+ high speed COGs.That could make sense for Parallax, but first they need to focus on P2
Keep in mind also the A9 and MAX10 are very new - you cannot yet buy all the MAX 10's and the A9 software support is very recent.
This dynamic nature of FPGA's underlines the risks of doing a module - the design life is short, and it is hard to find a large enough market sweetspot with a singe module.
If I read Chips comments correctly... he intends to release the verilog code for the P2. I have no doubt that Parallax has already considered licensing the design and may already be licensing the P1. Of course, It is their business (and none of ours) how they handle the issue. But the conversation is still worth having, just to make sure the details are considered.
As things progress, I think the brilliance of Parallax's approach will become more obvious.
The market at large has pretty much dismissed the roll your own peripheral set in favor of dedicated, industry standard/expected h/w. For this niche that do use the Prop for more than small qty enthusiast/low volume production, you may be able to get a subset of that interested in FPGA Px.
Again, what are the costs currently to get a P1 equivalent on FPGA? FPGA, board, discreets, etc. A $50-75 board will fly with the community, however doesn't make much sense in almost any manf. scenario.
And if people really want a P3, the thing to consider is wasting Parallax' limited resources with these ideas. Heck, why does Parallax need to make their own FPGA board when its going to be low volume run, and thus most likely more expensive than someone churning out thousands of them? Wasted time effort for slim pickings.
I could be wrong (it happened once before) however P1 has been open sourced for almost a year, and a quick site search on Google shows no responses to Parallax or the Prop. If the top FPGA enthusiast site doesn't show any interest, I fail to understand where/why anyone believes there is any sort of money to be made in licensing FPGA images.
The Prop has virtual devices, and the FPGA is a virtual machine.
Yo, we heard you like virtual, so we put your virtual devices inside a virtual machine.
Step 3. Profit ?
Get a P2 out, and have it prove itself.
If it does, then it has a chance of licenseability.
However, industry acceptance is absolutely not required for either the silicon or a license, FPGA solution to be viable and profitable.
If industry acceptance were the rule, we would never have P1, and would all center on the dominant solutions.
And that happens, but not always. And where people are working in niches that will pay for well differentiated, high value products, using Prop tech can make great sense. Those are small relative to the market overall. Who cares? Seriously.
Taking any kind of significant market share is also not required for it to make sense and be profitable.
The key thought here is to think products, solutions and value. Where that can be sold, the industry overall is darn near irrelevant. No Joke. I have done that analysis many times in my career where I get visibility into hundreds of companies, some of whom use methods that would surprise anyone thinking the mainstream, industry chatter about what works and what does not is all there is to say on the matter.
People say stateside manufacturing doesn't make sense. Happens every day, and there are companies making good money doing it too. People say BOM cost is king, when buyer value perception really is king too.
Anyone making zillions of things is going to have to think through those ideas and will very likely be centered in on that industry chatter. People making thousands of things really don't have to think through that as much as they do how to max out value perception.
Finally, the world is made on margins. Always defaulting to the rock bottom BOM, assembly, development, etc... costs is an easy way to survive on thin margins and the default associated with that is volume, which is where the real money is, right? In common consumer goods, yes. All true.
Leave that niche, and the whole game changes.
Solve a problem, and the value of that solution can very easily be expressed in terms of what the solution actually does as a benefit to the buyer. Point out all those benefits and quantify them and the buyer will spend the money to get the return. And it's that benefit to them that determines what things are worth and why they are worth what they are.
Plenty of room to play in this way. Again, happens every single day. Those businesses won't be found on the cover of Fast Company, won't be reviewed in the various journals and trades too much. Why? Because they aren't of scale, don't fit the narrative, aren't a source of drama, etc...
But they will make those who run them a very nice living.
Heck, why does Parallax need to make their own FPGA board when its going to be low volume run, and thus most likely more expensive than someone churning out thousands of them? Wasted time effort for slim pickings.
I think this comes down to test coverage, and availability.A FPGA-1-2-3 board that Parallax does, can have the support devices around the FPGA to test more of the P2. That's pretty important.
It was also started before the BeMicro CV A9 was around, and A9 support in WebPACK is brand new.
Of course, now the BeMicro CV A9 is real, that is a clear P2 build target, but not all tests can be run on BeMicro CV A9. Plenty of core SW tests can.
In going back through the Dark Silicon thread, I ran across Ken's comment on $100K+ development costs for the A7 board, now upgraded to the bigger FPGA.
I'm a little confused though, and have little knowledge of FPGA boards.
However, before going ahead with an R&D effort that is primarily focused on a P2 validation effort, and thus low/no recurring sales/profit expected, did Parallax expect to be able to undercut someone like http://www.terasic.com.tw/cgi-bin/page/archive.pl?CategoryNo=167&No=816
enough to make it worthwhile?
Or, were there features needed that simple couldn't have been found on a different board?
For in-house validation work, it would seem from the outside to have been cheaper to just buy x15-20 boards and stay focused vs an expensive R&D effort to try and compete against a deep pocket like Terasic.
Considering the FPGA fatigue in the forum, was any sort of official poll done to see how many people would actually stump up for yet another board?
Parallax has in-house expertise, however even when that is so, it doesn't necessarily follow that using that over buying a pre-made product is going to be cheaper/better.
Heck, 15 boards for in-house and 15 for solid community SME's @$400 = $12K vs $100K.
Or, were there features needed that simple couldn't have been found on a different board?
Yes, for testing Parallax need DAC and ADC that are somewhat similar to P2, and VGA and SDRAM of the older type, not DDR3, plus SPI parts the same as P2 will use for boot .. etc ..
Of course now the BeMicro CV A9 is real, just maybe a (cheaper) daughter card could get to almost the same place ?That's the benefit of hindsight
Since this is a serious discussion, then please consider the following: what if parallax could handle the educational / hobbyist portion of the xmos product line. Production of boards, support, and more friendly non-commercial tools. Because something xmos is not particularly good at, to say the least, is the individual person's experience or those trying to start out in programming or electronics.
Cooperate to focus on strengths, instead of competing. Xmos isn't good at education or hobbyists, and Parallax isn't good at chip design.
Because Chip probably can't be tainted by what is competition, what should still happen is someone else close within parallax should evaluate xmos and report back their findings. Looking at it objectively, instead of relying on other parties biased opinions.
Not fair to say Parallax isn't any good at chip design. More fair would be to say that Parallax' one-man chip design just isn't scaleable or able to compete against any other company that has half-dozen to 50/100 engineers on staff.
CXMOS has some pretty heavy-hitter funding and strategic investors. I doubt they'd be interested in Parallax as anything other than some form of distributor, and thats doubtful.
whicker,
"please consider the following: what if parallax could handle the educational / hobbyist portion of the xmos product line."
That would be a huge undertaking. Recreating educational materials, designing boards, writing manuals, creating tools.
It would be doing a lot of work to support and promote another companies products.
It would probably mean abandoning the PII, and the FPGA board, and pretty much the P1 and all the products that use it.
I'd love to see all these things for the XMOS devices, to get them into the hands of more people, hobbyists, students and others, to get more hacker friendly XMOS boards and addons. But no way is it of interest to Parallax.
1. Experienced P1 user wants to solve some problems just a bit bigger than the P1. This one should be fairly obvious.
2. Low volume, high margin, niche product. Advantage = time to market, ease of development, potential to put more advanced product development within reach of ordinary people.
3. Commercial P1 product manufacturer. Let's assume success with P1 and some product features / additional products are in the pipe. Advantage = not having to adopt new architecture, time to market, additional features, potential integrated device using FPGA to shrink BOM some to help with costs and margins.
4. Other device manufacturer. They run into limits, or desire features that prove painful to them in some way. Could be dev time. Could be device limits. Could be other things. Be creative here. Think backward and take the P2 features we understand today and frame them as possible limits / pain for other devices that see common use in smaller scale, higher margin products.
Any others?
Feel free to add to, shoot down, whatever on the ones I put here as a start.
It is my experience, getting past the "nice chat over coffee" stage means actualizing the idea. I suggest starting from a few plausible use cases. Forget the implausible ones. Those are easy, and the issues are present on this thread already. That's a "no can do" dialog.
The dialog I am suggesting is a "can do" dialog, and it starts by thinking through what actually needs to or can happen to enable "can do." Of course, we argue those, and some die, some grow, whatever. Maybe they all die. In that case, idea is great, but moot; otherwise, some favorable use cases leads to the next step, which is:
Taking those and identifying specifics and profiling what a business model might look like, and a big part of that might be something about value perception and where pricing might need to be, assuming a lean, generic BOM. Think of this as the floor qualifier. It will likely shoot down another use case, or niche, or advantage or two. No worries. The remaining ones now have some baseline numbers attached.
This can take a couple paths. One is to assume there won't be real silicon for a long time. That may well be a reasonable assumption, but it's not a death sentence at all. Another is to assume a path to silicon on a short term, say 1 year timeline. Yet another is to think about there never, ever being silicon.
Each of those has implications on the use cases.
From there, we can take a look at what the license could look like and what that means to everybody involved. For this part, assume some of something is better than a lot of nothing and that maximizing the some of something is the goal of the day.
If it gets here, then we are talking in some terms that are meaningful to everyone and the overall merits of the idea take shape to a point where it's really starting to make sense to hammer out license details and for Parallax to do the work to qualify the effort on their end.
How many of you have been through something like this?
I'm not sure how to work up a sample given a public forum. But, we need to. Or a private dialog must happen along similar lines. I can help with that.
***And this is how I know to do it. Anyone else with better means and methods? Speak up please! I, for one, will read and participate with great interest. My intent here is an attempt to have the dialog Jeff is looking for and either qualify this idea out or in.
***To those who are major industry trend / best practice focused. I get it, and you aren't wrong in those contexts. My advocacy here centers on those possibilities and opportunities that lie outside those well trodden norms. It can be tough to think this way, but if you do, that's where nice little businesses get formed. Sometimes they grow too. Other times they are just nice businesses that pay somebody well enough to be worth doing.
potatohead, I think you are still letting your Prop enthusiasm vastly color the reality of the situation.
Would you be willing to take a years sabbatical to work at Parallax, and live on the licensing you were able to bring in? Not being snarky, just a hypothetical.
Seriously, I think a lot of this type of talk does to Ken what most of us are trying not too do to Chip.
I think if the P1 was nearly as successful as some here think, Beau would never have been laid off to help Parallax cover license fees. Magically having it available as an FPGA doesn't to me, make any difference.
I am actually wondering about thinking only slightly out of the box.
What if Chip were to use opencores and one of the ARM-compaitble cores as a starting block? Granted it wouldn't be V7/V8, however I believe it would be free as in beer, avoid licensing issues, and I'm sure he could somehow fuse it to the P1/P1.5/P2 in such a way that more people would actually be interested in.
Or, do this with the lowRISC?
Since the Prop doesn't seem to be a Page 1 item even to Parallax, maybe its time to get out of the core business and into the uncore where the differentiation Chip likes can be focused upon?
1. Experienced P1 user wants to solve some problems just a bit bigger than the P1. This one should be fairly obvious.
2. Low volume, high margin, niche product. Advantage = time to market, ease of development, potential to put more advanced product development within reach of ordinary people.
3. Commercial P1 product manufacturer. Let's assume success with P1 and some product features / additional products are in the pipe. Advantage = not having to adopt new architecture, time to market, additional features, potential integrated device using FPGA to shrink BOM some to help with costs and margins.
Those are very niche, and notice how they need to offer something more than P1 to be viable.On present FPGA offerings, it still costs more to do a P1 in FPGA, than simply buy a P1.
That metric means it makes most sense to offer a P1 and a P1V, as the COGs in a std P1 come cheaper than P1V COGS and then you use P1V COGS where the std ones are not quite enough.
One potential area here, is speed - and the numbers suggest MAX 10 could boost MHz quite significantly, but a MAX 10 speed-optimized version is not there yet...
There may be very rare use cases, where fewer COGS are ok and the user desperately needs a smaller package. Small BGA FPGAs may help here.
Here's a sample thought, which may actually have relevance in some of my professional work right now.
Having worked with professional grade industrial robots for a while, I've come to the understanding that the system software supplied on the controller is quite capable, and is generally a totally closed thing, but it's also a very robust thing. They have a programming language that you do everything through. No lower level access is possible. You are just a user. That language can be exported as text, and programs can be written that way, with some limitations.
This is all done specifically to define how the bot is controlled and establish the liabilities as well as fit, form, function. If one can do it with that bot, it gets done with the supplied interface which comes with support and assurances. This is thick stuff, and it's hard to operate with on a smaller budget. Given one larger robot killed somebody recently, the approach is understandable. As robots move to smaller companies seeking to automate things, some add on or complimentary devices are going to be needed.
But what if somebody just wants to do a few simple things or inter-operate with novel machinery they are building?
Margins on products for a niche like this are HUGE. Seriously. It's pretty amazing. And, if you are off the beaten path, attempting a novel solution, buying computers, things like LABVIEW, custom sensor devices, I/O devices, etc... just getting help and some gear can start at $10K and go rapidly from there. The robots, software, controller, etc... start at $20K or so and go up quick. A typical project or cell can easily be $100K and up.
Without identifying a specific task, it's safe to say a smart device capable of basic signaling, sensor data collection / measurement / qualification / filtering, etc... that offers some display capability, perhaps a console for interaction with the industrial robots (helpful when programming as the hand held tool makes one input letters and numbers game console style...) could present very high value perception. This niche could land in the "make thousands" range quite easily, and it would feature support services, some education, and maybe connector and sensor sales in tandem.
The problem to solve is getting useful data to and from a novel machine intended to work in tandem with a robot, or robots. Today, that's a quite expensive proposition. There are a number of bigger, standard players with very high price tags attached. IMHO, this niche is ripe for some creative and disruptive products. Perfect for this kind of thing. I do not yet have the expertise to drill down more. Still learning. I have gotten my bots to do stuff and am just now exploring the I/O and how to make them do stuff together, or even respond to basics like a switch.
Primary attributes would be ease of use, very robust operation, ability to bridge the gap between industrial controllers and various things one might encounter when attempting to automate something. There is room on cost. The better the solution, the more room on cost there actually is.
Target buyers would be robot integrators, and manufacturing / automation / potentially IT related people in whatever enterprise is looking for ways to make use of industrial robotics. A direct sales channel and distributor (perhaps through a few integrators) model would be the primary ways this kind of product would be sold and supported.
A useful product selling for a couple thousand might generate half a million to a million per year on hundreds sold type numbers, and many of those could be build to ship, or at the least delivered from a lean inventory.
The product manufacturer is going to want to rapidly develop interfaces for a variety of things, offer useful UI / UX dynamics, and be able to improve on the product over time. They will want to package up complex things and present them in simple, useful, script-able ways. Robust, glitch free, deterministic operation is a must, and concurrent signaling, sensor read, display and many other tasks would be in play.
There are many niche products in this space. Typically, they require specific expertise, or are obtuse, difficult to use, due to minimal interaction capabilities. Many assumptions are made too. Packaging up similar capability in a more accessible solution would open automation doors for many who have the need but do not have the domain expertise needed to get started.
I've left off a lot of detail on this. I'm not where I need to be in order to flesh it out some. However, this is typical of the kind of niche "profile" info needed to understand where this kind of licensed P2 product might play well. And this could be done on other products too. Some thought about P2 type features and how they might play out favorably, and for that matter P1 features, is important to qualify this one.
Maybe some of us here have more domain expertise. I don't have as much as I need yet. Learning is happening, but it's a lot of work! I could absolutely use this kind of product.
The overall market for small to mid-sized robotics is exploding! For the next 10 years, small to mid sized companies are going to want to adopt the kinds of tech currently in use in the major vertical markets. (auto, aero, medical, energy) Right now, the major players are making more accessible and lower cost product offerings. People are getting used gear too.
One last thing. In a sense, this is competing with the big boys. But they are slow, and they have a problem:
That problem is they really can't undercut their current business without incurring very significant opportunity costs. Established means, methods, value, prices, etc... are delivering margins they need. Eventually, some of this will shake out, cheaper offerings will move, and there will be an adjustment. This timeline could be on the order of 20 years.
Typically, this means new products that don't necessarily inter-operate with established ones, or products that feature limits that segment into small to mid size, or the opportunity just sits out there, like it is right now.
Again, incomplete, but food for thought. Do you see a niche? What does
it look like? This post intended to stimulate some discussion to
support plausible use cases.
...
Margins on products for a niche like this are HUGE. Seriously. It's pretty amazing. And, if you are off the beaten path, attempting a novel solution, buying computers, things like LABVIEW, custom sensor devices, I/O devices, etc... just getting help and some gear can start at $10K and go rapidly from there. The robots, software, controller, etc... start at $20K or so and go up quick. A typical project or cell can easily be $100K and up.
....
Again, incomplete, but food for thought. Do you see a niche? What does
it look like? This post intended to stimulate some discussion to
support plausible use cases.
Certainly there are openings for P2 in Instrumentation, Labs, and Process control, and plenty of areas where software costs totally dominate over chip costs.
The Achilles Heel I see in P2, as currently specified, is in the Connectivity area.That means it will need another device to manage any networking.P2 can manage the control and determinism very well
Some quick comparisons of parts available in 2015, or announced recently :Renesas RX231 : CAN, USB and SDHI, FPU 48~100pins Wide VccNuvoton NUC472 : CAN, HS-USB, Ethernet, FPU 100~176 pins Wide VccInfineon XMC4800: CAN(6) USB, Ethernet, EtherCAT, FPU, 100~196 pins 2MBF 352kR
SDHI is spec'd as allowing "Connect to Wi-Fi, Wi-SUN, BLE, and other wireless ICs easily and at high speed"
>>potatohead, I think you are still letting your Prop enthusiasm vastly color the reality of the situation.
Hey, this may make no sense at all. However, I see "reality" defined in very non-inclusive terms. If we are to qualify this idea AT ALL, it makes great sense to put "reality" into some appropriate context.
That's all I'm doing here. And I've spent a lot of my career doing just that in many contexts. If I can get you to think a little differently, that's the goal.
See JMG processing on some of it:
>>Those are very niche, and notice how they need to offer something more than P1 to be viable.>>On present FPGA offerings, it still costs more to do a P1 in FPGA, than simply buy a P1.
Good thoughts, and yes narrow. But narrow isn't a killer. It does however require some more thought, a sample of which I put above.
>>Would you be willing to take a years sabbatical to work at Parallax, and
live on the licensing you were able >>to bring in? Not being snarky,
just a hypothetical.
Who are you kidding? That's snarky as hell. Perfectly OK though. That's a reasonable question to ask.
I don't think qualifying the idea would take a year. It would take a man month or maybe two though, and that should be clear given my recent posts. Mostly, somebody would have to write a check for that to happen in any real detail, though I may well help somebody out who is very seriously attempting to understand the opportunity, if any. How that gets compensated, if at all, depends on everything related to said attempt.
This question was asked: "Can we get some official discussion?"
And I put the sort of background work necessary for that dialog to make any real sense here for the purpose of potentially stimulating said dialog. Without that work, what exactly is there to discuss? Talking about tech issues and specs is fun, but this kind of discussion isn't always as fun, but it's necessary all the same.
A point many of you are making. And it's valid, but unqualified as true in that the idea itself makes no sense.
Do you understand the difference?
The difference is I'm suggesting ways to get at actually doing that work so that the idea is qualified in or out properly. There is either an opportunity or not, and it's either worth it or not.
And that actually is the answer to your question! If qualified, I might make an investment of that kind. See how it works now? Here's another way to put it: If they offered me the job, I would move to qualify that before doing much else. It's not ethical to just take it and pray. So the dialog would be, "does this make sense?" and at the end of that dialog would be a risk / reward profile to be evaluated.
In case, it's not clear, no matter what, for this conversation to actually be meaningful, some work does need to be done. A great way to start that work is to think through some cases and work backwards to understand what it means to everyone potentially involved. Which was the point of my posts.
I'm surprised somebody didn't mention another obvious case. I danced around it:
One familiar with P2 has a product idea NOW and could use a license.
So, there is a public offering, which will take considerable effort. And there may be a case for a private offering too, a lot less effort.
Both of those play out very differently, and the latter could be used as a way to help qualify this. Frankly, I would evaluate a private effort linked to a plausible, reasonably qualified product idea very closely. It's a great way to understand and make an investment at the same time. Consider that a standing recommendation.
A lot of what you take for "prop enthusiasm" is a deliberate way of thinking. Engineers are typically risk adverse people. Entrepreneurs are typically risk accepting people.
Both can be good, or a lot of trouble, depending.
I've spent a lot of years working on both sides of that fence, and qualification is the single most important thing to be done to avoid the trouble part of risk, be it avoiding it or taking it on. It all costs the same when the actual risk / reward isn't qualified in some way to be meaningful beyond a gamble of sorts.
The most important part of that whole mess is understanding where the potentials actually are and one doesn't get that unless some potential positive thoughts happen and get worked through.
Of course, I can just go away too. Plenty of other things to do.
I must go now, but I'll return tomorrow. That's the right kind of dialog.
But for this quick bit:
>>The Achilles Heel I see in P2, as currently specified, is in the Connectivity area.That means it will need another device to manage any networking.P2 can manage the control and determinism very well
In the license scenario, could the connectivity be bundled on the FPGA for something far better aligned with your use case? Or the use case? What does that look like, and how can it be done simply and easily?
See, different points of view.
What you consider snarky and still acceptable, I considered applying a more level-set POV from what I consider reality.
I had a buddy who was way on the entreprenurial (sp) side of things, and occasionally he did make money. But he also lost quite a lot, and was always either onthe high side buying me dinner to discuss ideas or on the low side renting in Oakland.
Maybe similar to Greece vs Germany, which is interesting to watch lately.
Parallax being able to branch out and find more hidden niches by which they can prosper would be great. The problem is the general market has not shown much interest, and Parallax needs to do a lot more work to prove they are AS capable as the current players BEFORE they can try to sell the idea that they are better for reason x, y, or z.
With 16 cores (Cogs, hah!) becoming a reality, I think Parallax should find an industrial niche like lab equipment or automotive, where such multicore, non-interrupt would likely be of interest. Unfortunately, the thing everyone heralds as the primary benefit of the Prop, soft peripherals, just seems like its not been much accepted by many at all, whether XMOS, Cypress, etc.
JMG or Leon can correct me on the latter two. If either have been successful, then it may be time to look at how they approach it and mirror it rather than still going it alone.
Of course, thats based upon the assumption that Parallax has the interest or financial ability. In some responses Parallax seems to indicate it would happy as a clam if they could just continue with the P2 doing sales similar to P1.
No one outside of a prop-head is going to spend the money on an FPGA board to do what a P1 can do. If they need more, then the next most economical thing to do would be to plop 2 P1's on a board and move ahead with minimal expense and effort, and potential troubleshooting.
FPGA for P2 seems to me to be even more uneconomical.
Thats why I think its mostly a pipedream. Parallax has a lot of other options with a much better bang for the buck potential return than this idea. And their option$ are probably more limited than we think.
No one outside of a prop-head is going to spend the money on an FPGA board to do what a P1 can do. If they need more, then the next most economical thing to do would be to plop 2 P1's on a board and move ahead with minimal expense and effort, and potential troubleshooting.
I agree that P1 Chip represents cheapest-per-cog, but the above presumes 'more COGS' is all that the user needs. MHz matters too.
P1V can hit other spaces - The latest NIOS numbers in MAX10, suggest 200+MHz equiv P1 performance should be possible.
P1V can also easily add HW modes like Quadrature counting, or HW LMM support..The appeal of MAX10, is it has a good size variation - choose the FPGA that fits only what you need.
Seems we are talking about similar things, but with different terms, or maybe framing on it.
In terms of "the industry" or people outside of Propeller land, yes. Licensing for use on FPGA devices does not make much sense.
That leaves people who are familiar, and that's precisely where I was centering on.
I wrote snarky, because there are better ways to ask that question. You will note I did agree it was valid, and my response was to qualify the whole thing too.
Those people you know who lost, very likely didn't do the work outlined in a few posts here. I do that work BTW, and I don't lose often because of that work.
So that leaves us with the bold recommendation left a post or two earlier. A private agreement to allow for product development on FPGA, with P1, V, P2, whatever, would make for a great exercise and some shared risks.
Success there would then be used to expand on that effort.
Again, just ruling it out for a general lack of easy path to greater acceptance isn't inclusive enough to qualify as "serious discussion" which is what Jeff asked for.
So we can leave it here for now. I don't see the kind of dialog needed to complete the exercise, though I did like where JMG seems to be with speed, feature, and other potentials bundled into one FPGA chip.
To me, that's the most compelling thing I've seen on the thread so far as P1 is concerned. We don't have a P2 image and until we do, that part of the discussion is early, as mentioned.
Early doesn't necessarily mean wait. It does mean there is a window to do the kinds of things needed to maximize it. The robotics related niche info is one of many that could be profiled and put into some simple, "maybe I can do that", or "there is a problem I can solve" type terms needed to much better understand how Propellers, licensed or not, could enjoy some acceptance outside the Prophead club.
That dialog is useful either way, and I didn't note it earlier. Whoops.
Finally, you are speaking strongly from the risk adverse angle. Great. Jeff is asking the question from the strong risk taking angle.
I did not see any real attempt to address Jeff's query and thought to stimulate that.
Please do not take it as some gung ho, blow a bunch of money type discussion. Please do take it as some food for thought as to how to resolve this in a higher value way.
Expanding on P1V and MAX10 (as for now, P1V is real, and smaller than P2).
As time progresses, P2 will naturally get more viable in FPGA space too.
The reported build numbers suggest a single P1V COG can (just) fit in the smallest MAX10 - which is a 3mm x 3mm BGA with 27io
Looks like those are due in stock @ Digikey later this month $2.84/1k
That could make a small module of [P1 + Smallest MAX10] viable ?
Gives 8 Standard COGS, and one TurboCOG, with special features possible.
Unfortunately, the thing everyone heralds as the primary benefit of the Prop, soft peripherals, just seems like its not been much accepted by many at all, whether XMOS, Cypress, etc.
JMG or Leon can correct me on the latter two. If either have been successful, then it may be time to look at how they approach it and mirror it rather than still going it alone.
XMOS has been very successful in high-end audio processing with lots of companies including Sony using their devices: http://www.xmos.com/products/silicon/xcore-audio http://www.xmos.com/applications/audio/high-resolution
Here are some case studies illustrating the speed with which designs can be brought to market using XMOS technology: http://www.xmos.com/news/casestudies
Xilinx has put money into the company, although one of the applications is FPGA replacement in low-end systems.
However, realistically, what are we talking about BOM-wise/end board cost for a 200mhz P1 FPGA solution?
I'm sure there are some products which are not penny-pinching, lowest cost production. However, going from a $8 chip that almost gets you there, to a $30, 40 or more board is an issue of more than mere pennies.
I can't remember Ken's 5 primary P2 needs, speed may or may not have been on it.
Without any idea of actual customer requests, which only Parallax would know, I think I'd be safe in saying that at double, triple or more the cost of P1 alone, thats getting a bit expensive and another mcu might be worth looking at.
But thats just me trying to be rational without knowing User requests, volume, etc. I could be wrong certainly, however I'll play bad cop until/unless I get more information that indicates there is any real market potential for such.
OK, from the perspective of potential Prop-users looking to bump up a P1 FPGA before P2.
Sure, Parallax could do that, however I'm kinda thinking just the lawyers fees alone in getting that all set up as an ongoing concern might eclipse the potential profit.
The only way it make sense in the 3-4 minutes I've thought about it, would be for those people who want a P1 Hi-Speed variant as JMG suggested. Otherwise, $8 P1 works whether you need 4/5/6/7/8 cores.
So, P1 Hi-speed, what can Parallax reasonably charge?
Is that just for the P1v, does that come rated at a certain speed with a certain FPGA, what if they want 200Mhz? Do the fees take into account Buyer engineering R&D, or does the image come able to run at that speed?
What happens when there are liability concerns, those get expensive when both sides have lawyers. Does Parallax spend the R&D to determine what FPGA families Xilinx/Altera/Intel are warranted, or when FPGA features change or spec changes, who responsible?
Biggest thing is, are there any current volume users asking for such a thing, and, are they willing to buy enough to even make it worth Parallax's while to investigate?
I personally think you'd have to be the hardest of hardcore Prop devotees to be willing to seriously look at this. Many in the forum would do it.
But someone would have to spend the dollars on R&D. Although, now that I think of it, Parallax has just gone through this process, albeit for the P2.
I suppose they could reuse a lot of that work for a full blown P1 (is that done already?) on a specific FPGA, with board layout, etc, and see if there are any takers.
Before any of that is done though, Parallax should be able to give an actual board cost figure.
If Parallax were to estimate it would come in at $45 in Qty 1 from the website, they could check with their volume/specialty customers and the forum to see if there were any takers before expending any capitol.
I think JMG could probably post an approximate cost just of the top of his head.
Just not sure there is any market of real worth for a 200Mhz P1, though interested in seeing myself proven wrong.
However, realistically, what are we talking about BOM-wise/end board cost for a 200mhz P1 FPGA solution?
See above, $2.84/1k gives you (roughly) 1 P1V COG - if you are talking about chip level customers, then chip level prices apply
I'm sure there are some products which are not penny-pinching, lowest cost production. However, going from a $8 chip that almost gets you there, to a $30, 40 or more board is an issue of more than mere pennies.
You know, I just realized something in this discussion:
There is the revenue potential for Parallax, whoever is seeking a license, and potentially some middle person / integrator type.
Attorney costs can be kept very low by not doing a public license offering. I'm really serious when I say it should be limited to those able to produce some business model or other. Just for a while. That can keep the risk and cost compartmentalized, which would be good.
Some other use cases are falling out here:
1. Modules. Psomething go small, big, fast, more I/O.
Wouldn't the "more I/O" case be worth an FPGA frequently? We've all seen that a P1 could easily be driving 64 pins, if not more. Dual props, feature limits, compromises all have happened due to lack of pins, speed and memory.
2. FPGA version of the Proto Board.
It runs a Prop image, provides I/O, and the bare minimums beyond that.
How to wade into this thread? Lets start with some reference points
Re "Ken's 5 key features" - here Ken identified Code Protection, Analog, bigger RAM, more pins as fundamental. This coincides with Chip's epic thread title, and his recent confirmation of expected feature set.
There may be more cogs in the mix, but lets see how the design fits and what power it takes before setting cog number expectations in stone.
Its worth noting that 4 of these key features are proven and/or relatively straightforward using P1V. And analog should be easy with Max10. OzPropDev's generator tool breaks down several barriers and comes recommended for use with the expanding range of cheap fpga dev boards.
If you want to do commercial/industrial P1V development, as Heater pointed out recently and Ken indicated in the past, contact Parallax (after Ken surfaces for air)
P2 "Hot" option:- don't rule this out, at this point it is still a potent option (at least for development), and it already exists, and entry price is a DE0 Nano. Breakouts are available. Lots of what Chip is aiming for is already in this incarnation, it won't be that hard to migrate. Its got hubexec, more pins, more ram, and similar mips per cog as likely in the final incarnation. Recognize the math will be done via a hub cordic, and hardware tasks will need to migrate to their own cog.
I'm interested in why the above two options don't appeal to the Osborne club waiting for the next incarnation. What are the key turn-offs, in rank order?
I want to flag a risk/threat I haven't seen raised by others - hardware patents. There are a lot out there. Finding/proving P2 (accidently) infringes another patent/submarine patent is a lot easier if Parallax releases P2 verilog in a wide public release, so the solution is to release binaries or go narrow, at least until development money is recouped.
Or just do the soft-release that can be patched, but that'd be an underwhelming result from where Chip says things are at now
Comments
Sure would be nice to see some "official" input on this thread. Maybe we could get Ken or Chip to weigh in on this discussion after the holiday. There is so much "win" in this concept and I'm really pleased to see it being explored. Now... If we can only get past the chat with coffee stage..
My direction is to use a larger FPGA board, with ARM cores, and add a custom processor (inspired by the P1V and RISC V) between their I/O logic and the ARMs to create a souped up version of the BeagleBone Black. I would LOVE to be able to go to Parallax and shop for a P1V+ or P2V variant to do what I want.
A series of boards based on the Altera Max10 series (as an example) could be made as drop in System-on-Modules that could be designed to be easy to adapt to other systems or uses. Complete flexibility! Education, hobby, custom commercial applications all pulling from the same investment.
Times change, and are changing even faster now. By the time Parallax gets a P2 chip in hand it will be obsolete with little to no real market potential.
Perfect example is their 123 board. Spent a bunch of time and money to
design and manufacture it and now there is an A9 based board for 1/3 the price already
to market while we still wait for theirs.
Some self-contradiction going down there.First a plea for flexible modules, then an example of how that fails to meet price.
The P2 is far from release, so it is way too early to market that, but the P1V is real, and anyone is free to make a module targeting that.
I'm sure if there is enough critical mass for such products, some will arrive.
eg The recent MAX 10 numbers suggest a Board with MAX 10 P1V plus a P1 could give 8 std COGS plus 1-2+ high speed COGs.That could make sense for Parallax, but first they need to focus on P2
Keep in mind also the A9 and MAX10 are very new - you cannot yet buy all the MAX 10's and the A9 software support is very recent.
This dynamic nature of FPGA's underlines the risks of doing a module - the design life is short, and it is hard to find a large enough market sweetspot with a singe module.
As things progress, I think the brilliance of Parallax's approach will become more obvious.
I feel very lucky right now. Do you feel lucky?
The market at large has pretty much dismissed the roll your own peripheral set in favor of dedicated, industry standard/expected h/w. For this niche that do use the Prop for more than small qty enthusiast/low volume production, you may be able to get a subset of that interested in FPGA Px.
Again, what are the costs currently to get a P1 equivalent on FPGA? FPGA, board, discreets, etc. A $50-75 board will fly with the community, however doesn't make much sense in almost any manf. scenario.
And if people really want a P3, the thing to consider is wasting Parallax' limited resources with these ideas. Heck, why does Parallax need to make their own FPGA board when its going to be low volume run, and thus most likely more expensive than someone churning out thousands of them? Wasted time effort for slim pickings.
I could be wrong (it happened once before) however P1 has been open sourced for almost a year, and a quick site search on Google shows no responses to Parallax or the Prop. If the top FPGA enthusiast site doesn't show any interest, I fail to understand where/why anyone believes there is any sort of money to be made in licensing FPGA images.
The Prop has virtual devices, and the FPGA is a virtual machine.
Yo, we heard you like virtual, so we put your virtual devices inside a virtual machine.
Step 3. Profit ?
Get a P2 out, and have it prove itself.
If it does, then it has a chance of licenseability.
Wait until we have PII silicon and see if industry likes it. If it does then consider options.
However, industry acceptance is absolutely not required for either the silicon or a license, FPGA solution to be viable and profitable.
If industry acceptance were the rule, we would never have P1, and would all center on the dominant solutions.
And that happens, but not always. And where people are working in niches that will pay for well differentiated, high value products, using Prop tech can make great sense. Those are small relative to the market overall. Who cares? Seriously.
Taking any kind of significant market share is also not required for it to make sense and be profitable.
The key thought here is to think products, solutions and value. Where that can be sold, the industry overall is darn near irrelevant. No Joke. I have done that analysis many times in my career where I get visibility into hundreds of companies, some of whom use methods that would surprise anyone thinking the mainstream, industry chatter about what works and what does not is all there is to say on the matter.
People say stateside manufacturing doesn't make sense. Happens every day, and there are companies making good money doing it too. People say BOM cost is king, when buyer value perception really is king too.
Anyone making zillions of things is going to have to think through those ideas and will very likely be centered in on that industry chatter. People making thousands of things really don't have to think through that as much as they do how to max out value perception.
Finally, the world is made on margins. Always defaulting to the rock bottom BOM, assembly, development, etc... costs is an easy way to survive on thin margins and the default associated with that is volume, which is where the real money is, right? In common consumer goods, yes. All true.
Leave that niche, and the whole game changes.
Solve a problem, and the value of that solution can very easily be expressed in terms of what the solution actually does as a benefit to the buyer. Point out all those benefits and quantify them and the buyer will spend the money to get the return. And it's that benefit to them that determines what things are worth and why they are worth what they are.
Plenty of room to play in this way. Again, happens every single day. Those businesses won't be found on the cover of Fast Company, won't be reviewed in the various journals and trades too much. Why? Because they aren't of scale, don't fit the narrative, aren't a source of drama, etc...
But they will make those who run them a very nice living.
I think this comes down to test coverage, and availability.A FPGA-1-2-3 board that Parallax does, can have the support devices around the FPGA to test more of the P2. That's pretty important.
It was also started before the BeMicro CV A9 was around, and A9 support in WebPACK is brand new.
Of course, now the BeMicro CV A9 is real, that is a clear P2 build target, but not all tests can be run on BeMicro CV A9. Plenty of core SW tests can.
In going back through the Dark Silicon thread, I ran across Ken's comment on $100K+ development costs for the A7 board, now upgraded to the bigger FPGA.
I'm a little confused though, and have little knowledge of FPGA boards.
However, before going ahead with an R&D effort that is primarily focused on a P2 validation effort, and thus low/no recurring sales/profit expected, did Parallax expect to be able to undercut someone like http://www.terasic.com.tw/cgi-bin/page/archive.pl?CategoryNo=167&No=816
enough to make it worthwhile?
Or, were there features needed that simple couldn't have been found on a different board?
For in-house validation work, it would seem from the outside to have been cheaper to just buy x15-20 boards and stay focused vs an expensive R&D effort to try and compete against a deep pocket like Terasic.
Considering the FPGA fatigue in the forum, was any sort of official poll done to see how many people would actually stump up for yet another board?
Parallax has in-house expertise, however even when that is so, it doesn't necessarily follow that using that over buying a pre-made product is going to be cheaper/better.
Heck, 15 boards for in-house and 15 for solid community SME's @$400 = $12K vs $100K.
Or, were there features needed that simple couldn't have been found on a different board?
Yes, for testing Parallax need DAC and ADC that are somewhat similar to P2, and VGA and SDRAM of the older type, not DDR3, plus SPI parts the same as P2 will use for boot .. etc ..
Of course now the BeMicro CV A9 is real, just maybe a (cheaper) daughter card could get to almost the same place ?That's the benefit of hindsight
Since this is a serious discussion, then please consider the following: what if parallax could handle the educational / hobbyist portion of the xmos product line. Production of boards, support, and more friendly non-commercial tools. Because something xmos is not particularly good at, to say the least, is the individual person's experience or those trying to start out in programming or electronics.
Cooperate to focus on strengths, instead of competing. Xmos isn't good at education or hobbyists, and Parallax isn't good at chip design.
Because Chip probably can't be tainted by what is competition, what should still happen is someone else close within parallax should evaluate xmos and report back their findings. Looking at it objectively, instead of relying on other parties biased opinions.
CXMOS has some pretty heavy-hitter funding and strategic investors. I doubt they'd be interested in Parallax as anything other than some form of distributor, and thats doubtful.
"please consider the following: what if parallax could handle the educational / hobbyist portion of the xmos product line."
That would be a huge undertaking. Recreating educational materials, designing boards, writing manuals, creating tools.
It would be doing a lot of work to support and promote another companies products.
It would probably mean abandoning the PII, and the FPGA board, and pretty much the P1 and all the products that use it.
I'd love to see all these things for the XMOS devices, to get them into the hands of more people, hobbyists, students and others, to get more hacker friendly XMOS boards and addons. But no way is it of interest to Parallax.
So, use cases?
Here's a few I can think of:
1. Experienced P1 user wants to solve some problems just a bit bigger than the P1. This one should be fairly obvious.
2. Low volume, high margin, niche product. Advantage = time to market, ease of development, potential to put more advanced product development within reach of ordinary people.
3. Commercial P1 product manufacturer. Let's assume success with P1 and some product features / additional products are in the pipe. Advantage = not having to adopt new architecture, time to market, additional features, potential integrated device using FPGA to shrink BOM some to help with costs and margins.
4. Other device manufacturer. They run into limits, or desire features that prove painful to them in some way. Could be dev time. Could be device limits. Could be other things. Be creative here. Think backward and take the P2 features we understand today and frame them as possible limits / pain for other devices that see common use in smaller scale, higher margin products.
Any others?
Feel free to add to, shoot down, whatever on the ones I put here as a start.
It is my experience, getting past the "nice chat over coffee" stage means actualizing the idea. I suggest starting from a few plausible use cases. Forget the implausible ones. Those are easy, and the issues are present on this thread already. That's a "no can do" dialog.
The dialog I am suggesting is a "can do" dialog, and it starts by thinking through what actually needs to or can happen to enable "can do." Of course, we argue those, and some die, some grow, whatever. Maybe they all die. In that case, idea is great, but moot; otherwise, some favorable use cases leads to the next step, which is:
Taking those and identifying specifics and profiling what a business model might look like, and a big part of that might be something about value perception and where pricing might need to be, assuming a lean, generic BOM. Think of this as the floor qualifier. It will likely shoot down another use case, or niche, or advantage or two. No worries. The remaining ones now have some baseline numbers attached.
This can take a couple paths. One is to assume there won't be real silicon for a long time. That may well be a reasonable assumption, but it's not a death sentence at all. Another is to assume a path to silicon on a short term, say 1 year timeline. Yet another is to think about there never, ever being silicon.
Each of those has implications on the use cases.
From there, we can take a look at what the license could look like and what that means to everybody involved. For this part, assume some of something is better than a lot of nothing and that maximizing the some of something is the goal of the day.
If it gets here, then we are talking in some terms that are meaningful to everyone and the overall merits of the idea take shape to a point where it's really starting to make sense to hammer out license details and for Parallax to do the work to qualify the effort on their end.
How many of you have been through something like this?
I'm not sure how to work up a sample given a public forum. But, we need to. Or a private dialog must happen along similar lines. I can help with that.
***And this is how I know to do it. Anyone else with better means and methods? Speak up please! I, for one, will read and participate with great interest. My intent here is an attempt to have the dialog Jeff is looking for and either qualify this idea out or in.
***To those who are major industry trend / best practice focused. I get it, and you aren't wrong in those contexts. My advocacy here centers on those possibilities and opportunities that lie outside those well trodden norms. It can be tough to think this way, but if you do, that's where nice little businesses get formed. Sometimes they grow too. Other times they are just nice businesses that pay somebody well enough to be worth doing.
potatohead, I think you are still letting your Prop enthusiasm vastly color the reality of the situation.
Would you be willing to take a years sabbatical to work at Parallax, and live on the licensing you were able to bring in? Not being snarky, just a hypothetical.
Seriously, I think a lot of this type of talk does to Ken what most of us are trying not too do to Chip.
I think if the P1 was nearly as successful as some here think, Beau would never have been laid off to help Parallax cover license fees. Magically having it available as an FPGA doesn't to me, make any difference.
I am actually wondering about thinking only slightly out of the box.
What if Chip were to use opencores and one of the ARM-compaitble cores as a starting block? Granted it wouldn't be V7/V8, however I believe it would be free as in beer, avoid licensing issues, and I'm sure he could somehow fuse it to the P1/P1.5/P2 in such a way that more people would actually be interested in.
Or, do this with the lowRISC?
Since the Prop doesn't seem to be a Page 1 item even to Parallax, maybe its time to get out of the core business and into the uncore where the differentiation Chip likes can be focused upon?
Here's a few I can think of:
1. Experienced P1 user wants to solve some problems just a bit bigger than the P1. This one should be fairly obvious.
2. Low volume, high margin, niche product. Advantage = time to market, ease of development, potential to put more advanced product development within reach of ordinary people.
3. Commercial P1 product manufacturer. Let's assume success with P1 and some product features / additional products are in the pipe. Advantage = not having to adopt new architecture, time to market, additional features, potential integrated device using FPGA to shrink BOM some to help with costs and margins.
Those are very niche, and notice how they need to offer something more than P1 to be viable.On present FPGA offerings, it still costs more to do a P1 in FPGA, than simply buy a P1.
That metric means it makes most sense to offer a P1 and a P1V, as the COGs in a std P1 come cheaper than P1V COGS and then you use P1V COGS where the std ones are not quite enough.
One potential area here, is speed - and the numbers suggest MAX 10 could boost MHz quite significantly, but a MAX 10 speed-optimized version is not there yet...
There may be very rare use cases, where fewer COGS are ok and the user desperately needs a smaller package. Small BGA FPGAs may help here.
Having worked with professional grade industrial robots for a while, I've come to the understanding that the system software supplied on the controller is quite capable, and is generally a totally closed thing, but it's also a very robust thing. They have a programming language that you do everything through. No lower level access is possible. You are just a user. That language can be exported as text, and programs can be written that way, with some limitations.
This is all done specifically to define how the bot is controlled and establish the liabilities as well as fit, form, function. If one can do it with that bot, it gets done with the supplied interface which comes with support and assurances. This is thick stuff, and it's hard to operate with on a smaller budget. Given one larger robot killed somebody recently, the approach is understandable. As robots move to smaller companies seeking to automate things, some add on or complimentary devices are going to be needed.
But what if somebody just wants to do a few simple things or inter-operate with novel machinery they are building?
Margins on products for a niche like this are HUGE. Seriously. It's pretty amazing. And, if you are off the beaten path, attempting a novel solution, buying computers, things like LABVIEW, custom sensor devices, I/O devices, etc... just getting help and some gear can start at $10K and go rapidly from there. The robots, software, controller, etc... start at $20K or so and go up quick. A typical project or cell can easily be $100K and up.
Without identifying a specific task, it's safe to say a smart device capable of basic signaling, sensor data collection / measurement / qualification / filtering, etc... that offers some display capability, perhaps a console for interaction with the industrial robots (helpful when programming as the hand held tool makes one input letters and numbers game console style...) could present very high value perception. This niche could land in the "make thousands" range quite easily, and it would feature support services, some education, and maybe connector and sensor sales in tandem.
The problem to solve is getting useful data to and from a novel machine intended to work in tandem with a robot, or robots. Today, that's a quite expensive proposition. There are a number of bigger, standard players with very high price tags attached. IMHO, this niche is ripe for some creative and disruptive products. Perfect for this kind of thing. I do not yet have the expertise to drill down more. Still learning. I have gotten my bots to do stuff and am just now exploring the I/O and how to make them do stuff together, or even respond to basics like a switch.
Primary attributes would be ease of use, very robust operation, ability to bridge the gap between industrial controllers and various things one might encounter when attempting to automate something. There is room on cost. The better the solution, the more room on cost there actually is.
Target buyers would be robot integrators, and manufacturing / automation / potentially IT related people in whatever enterprise is looking for ways to make use of industrial robotics. A direct sales channel and distributor (perhaps through a few integrators) model would be the primary ways this kind of product would be sold and supported.
A useful product selling for a couple thousand might generate half a million to a million per year on hundreds sold type numbers, and many of those could be build to ship, or at the least delivered from a lean inventory.
The product manufacturer is going to want to rapidly develop interfaces for a variety of things, offer useful UI / UX dynamics, and be able to improve on the product over time. They will want to package up complex things and present them in simple, useful, script-able ways. Robust, glitch free, deterministic operation is a must, and concurrent signaling, sensor read, display and many other tasks would be in play.
There are many niche products in this space. Typically, they require specific expertise, or are obtuse, difficult to use, due to minimal interaction capabilities. Many assumptions are made too. Packaging up similar capability in a more accessible solution would open automation doors for many who have the need but do not have the domain expertise needed to get started.
I've left off a lot of detail on this. I'm not where I need to be in order to flesh it out some. However, this is typical of the kind of niche "profile" info needed to understand where this kind of licensed P2 product might play well. And this could be done on other products too. Some thought about P2 type features and how they might play out favorably, and for that matter P1 features, is important to qualify this one.
Maybe some of us here have more domain expertise. I don't have as much as I need yet. Learning is happening, but it's a lot of work! I could absolutely use this kind of product.
The overall market for small to mid-sized robotics is exploding! For the next 10 years, small to mid sized companies are going to want to adopt the kinds of tech currently in use in the major vertical markets. (auto, aero, medical, energy) Right now, the major players are making more accessible and lower cost product offerings. People are getting used gear too.
One last thing. In a sense, this is competing with the big boys. But they are slow, and they have a problem:
That problem is they really can't undercut their current business without incurring very significant opportunity costs. Established means, methods, value, prices, etc... are delivering margins they need. Eventually, some of this will shake out, cheaper offerings will move, and there will be an adjustment. This timeline could be on the order of 20 years.
Typically, this means new products that don't necessarily inter-operate with established ones, or products that feature limits that segment into small to mid size, or the opportunity just sits out there, like it is right now.
Again, incomplete, but food for thought. Do you see a niche? What does
it look like? This post intended to stimulate some discussion to
support plausible use cases.
Margins on products for a niche like this are HUGE. Seriously. It's pretty amazing. And, if you are off the beaten path, attempting a novel solution, buying computers, things like LABVIEW, custom sensor devices, I/O devices, etc... just getting help and some gear can start at $10K and go rapidly from there. The robots, software, controller, etc... start at $20K or so and go up quick. A typical project or cell can easily be $100K and up.
....
Again, incomplete, but food for thought. Do you see a niche? What does
it look like? This post intended to stimulate some discussion to
support plausible use cases.
Certainly there are openings for P2 in Instrumentation, Labs, and Process control, and plenty of areas where software costs totally dominate over chip costs.
The Achilles Heel I see in P2, as currently specified, is in the Connectivity area.That means it will need another device to manage any networking.P2 can manage the control and determinism very well
Some quick comparisons of parts available in 2015, or announced recently :Renesas RX231 : CAN, USB and SDHI, FPU 48~100pins Wide VccNuvoton NUC472 : CAN, HS-USB, Ethernet, FPU 100~176 pins Wide VccInfineon XMC4800: CAN(6) USB, Ethernet, EtherCAT, FPU, 100~196 pins 2MBF 352kR
SDHI is spec'd as allowing "Connect to Wi-Fi, Wi-SUN, BLE, and other wireless ICs easily and at high speed"
Hey, this may make no sense at all. However, I see "reality" defined in very non-inclusive terms. If we are to qualify this idea AT ALL, it makes great sense to put "reality" into some appropriate context.
That's all I'm doing here. And I've spent a lot of my career doing just that in many contexts. If I can get you to think a little differently, that's the goal.
See JMG processing on some of it:
>>Those are very niche, and notice how they need to offer something more than P1 to be viable.>>On present FPGA offerings, it still costs more to do a P1 in FPGA, than simply buy a P1.
Good thoughts, and yes narrow. But narrow isn't a killer. It does however require some more thought, a sample of which I put above.
>>Would you be willing to take a years sabbatical to work at Parallax, and
live on the licensing you were able >>to bring in? Not being snarky,
just a hypothetical.
Who are you kidding? That's snarky as hell. Perfectly OK though. That's a reasonable question to ask.
I don't think qualifying the idea would take a year. It would take a man month or maybe two though, and that should be clear given my recent posts. Mostly, somebody would have to write a check for that to happen in any real detail, though I may well help somebody out who is very seriously attempting to understand the opportunity, if any. How that gets compensated, if at all, depends on everything related to said attempt.
This question was asked: "Can we get some official discussion?"
And I put the sort of background work necessary for that dialog to make any real sense here for the purpose of potentially stimulating said dialog. Without that work, what exactly is there to discuss? Talking about tech issues and specs is fun, but this kind of discussion isn't always as fun, but it's necessary all the same.
A point many of you are making. And it's valid, but unqualified as true in that the idea itself makes no sense.
Do you understand the difference?
The difference is I'm suggesting ways to get at actually doing that work so that the idea is qualified in or out properly. There is either an opportunity or not, and it's either worth it or not.
And that actually is the answer to your question! If qualified, I might make an investment of that kind. See how it works now? Here's another way to put it: If they offered me the job, I would move to qualify that before doing much else. It's not ethical to just take it and pray. So the dialog would be, "does this make sense?" and at the end of that dialog would be a risk / reward profile to be evaluated.
In case, it's not clear, no matter what, for this conversation to actually be meaningful, some work does need to be done. A great way to start that work is to think through some cases and work backwards to understand what it means to everyone potentially involved. Which was the point of my posts.
I'm surprised somebody didn't mention another obvious case. I danced around it:
One familiar with P2 has a product idea NOW and could use a license.
So, there is a public offering, which will take considerable effort. And there may be a case for a private offering too, a lot less effort.
Both of those play out very differently, and the latter could be used as a way to help qualify this. Frankly, I would evaluate a private effort linked to a plausible, reasonably qualified product idea very closely. It's a great way to understand and make an investment at the same time. Consider that a standing recommendation.
A lot of what you take for "prop enthusiasm" is a deliberate way of thinking. Engineers are typically risk adverse people. Entrepreneurs are typically risk accepting people.
Both can be good, or a lot of trouble, depending.
I've spent a lot of years working on both sides of that fence, and qualification is the single most important thing to be done to avoid the trouble part of risk, be it avoiding it or taking it on. It all costs the same when the actual risk / reward isn't qualified in some way to be meaningful beyond a gamble of sorts.
The most important part of that whole mess is understanding where the potentials actually are and one doesn't get that unless some potential positive thoughts happen and get worked through.
Of course, I can just go away too. Plenty of other things to do.
I must go now, but I'll return tomorrow. That's the right kind of dialog.
But for this quick bit:
>>The Achilles Heel I see in P2, as currently specified, is in the Connectivity area.That means it will need another device to manage any networking.P2 can manage the control and determinism very well
In the license scenario, could the connectivity be bundled on the FPGA for something far better aligned with your use case? Or the use case? What does that look like, and how can it be done simply and easily?
See, different points of view.
What you consider snarky and still acceptable, I considered applying a more level-set POV from what I consider reality.
I had a buddy who was way on the entreprenurial (sp) side of things, and occasionally he did make money. But he also lost quite a lot, and was always either onthe high side buying me dinner to discuss ideas or on the low side renting in Oakland.
Maybe similar to Greece vs Germany, which is interesting to watch lately.
Parallax being able to branch out and find more hidden niches by which they can prosper would be great. The problem is the general market has not shown much interest, and Parallax needs to do a lot more work to prove they are AS capable as the current players BEFORE they can try to sell the idea that they are better for reason x, y, or z.
With 16 cores (Cogs, hah!) becoming a reality, I think Parallax should find an industrial niche like lab equipment or automotive, where such multicore, non-interrupt would likely be of interest. Unfortunately, the thing everyone heralds as the primary benefit of the Prop, soft peripherals, just seems like its not been much accepted by many at all, whether XMOS, Cypress, etc.
JMG or Leon can correct me on the latter two. If either have been successful, then it may be time to look at how they approach it and mirror it rather than still going it alone.
Of course, thats based upon the assumption that Parallax has the interest or financial ability. In some responses Parallax seems to indicate it would happy as a clam if they could just continue with the P2 doing sales similar to P1.
No one outside of a prop-head is going to spend the money on an FPGA board to do what a P1 can do. If they need more, then the next most economical thing to do would be to plop 2 P1's on a board and move ahead with minimal expense and effort, and potential troubleshooting.
FPGA for P2 seems to me to be even more uneconomical.
Thats why I think its mostly a pipedream. Parallax has a lot of other options with a much better bang for the buck potential return than this idea. And their option$ are probably more limited than we think.
No one outside of a prop-head is going to spend the money on an FPGA board to do what a P1 can do. If they need more, then the next most economical thing to do would be to plop 2 P1's on a board and move ahead with minimal expense and effort, and potential troubleshooting.
I agree that P1 Chip represents cheapest-per-cog, but the above presumes 'more COGS' is all that the user needs. MHz matters too.
P1V can hit other spaces - The latest NIOS numbers in MAX10, suggest 200+MHz equiv P1 performance should be possible.
P1V can also easily add HW modes like Quadrature counting, or HW LMM support..The appeal of MAX10, is it has a good size variation - choose the FPGA that fits only what you need.
Seems we are talking about similar things, but with different terms, or maybe framing on it.
In terms of "the industry" or people outside of Propeller land, yes. Licensing for use on FPGA devices does not make much sense.
That leaves people who are familiar, and that's precisely where I was centering on.
I wrote snarky, because there are better ways to ask that question. You will note I did agree it was valid, and my response was to qualify the whole thing too.
Those people you know who lost, very likely didn't do the work outlined in a few posts here. I do that work BTW, and I don't lose often because of that work.
So that leaves us with the bold recommendation left a post or two earlier. A private agreement to allow for product development on FPGA, with P1, V, P2, whatever, would make for a great exercise and some shared risks.
Success there would then be used to expand on that effort.
Again, just ruling it out for a general lack of easy path to greater acceptance isn't inclusive enough to qualify as "serious discussion" which is what Jeff asked for.
So we can leave it here for now. I don't see the kind of dialog needed to complete the exercise, though I did like where JMG seems to be with speed, feature, and other potentials bundled into one FPGA chip.
To me, that's the most compelling thing I've seen on the thread so far as P1 is concerned. We don't have a P2 image and until we do, that part of the discussion is early, as mentioned.
Early doesn't necessarily mean wait. It does mean there is a window to do the kinds of things needed to maximize it. The robotics related niche info is one of many that could be profiled and put into some simple, "maybe I can do that", or "there is a problem I can solve" type terms needed to much better understand how Propellers, licensed or not, could enjoy some acceptance outside the Prophead club.
That dialog is useful either way, and I didn't note it earlier. Whoops.
Finally, you are speaking strongly from the risk adverse angle. Great. Jeff is asking the question from the strong risk taking angle.
I did not see any real attempt to address Jeff's query and thought to stimulate that.
Please do not take it as some gung ho, blow a bunch of money type discussion. Please do take it as some food for thought as to how to resolve this in a higher value way.
As time progresses, P2 will naturally get more viable in FPGA space too.
The reported build numbers suggest a single P1V COG can (just) fit in the smallest MAX10 - which is a 3mm x 3mm BGA with 27io
Looks like those are due in stock @ Digikey later this month $2.84/1k
That could make a small module of [P1 + Smallest MAX10] viable ?
Gives 8 Standard COGS, and one TurboCOG, with special features possible.
JMG or Leon can correct me on the latter two. If either have been successful, then it may be time to look at how they approach it and mirror it rather than still going it alone.
XMOS has been very successful in high-end audio processing with lots of companies including Sony using their devices:
http://www.xmos.com/products/silicon/xcore-audio
http://www.xmos.com/applications/audio/high-resolution
Here are some case studies illustrating the speed with which designs can be brought to market using XMOS technology:
http://www.xmos.com/news/casestudies
Xilinx has put money into the company, although one of the applications is FPGA replacement in low-end systems.
OK, now thats a point I can understand.
However, realistically, what are we talking about BOM-wise/end board cost for a 200mhz P1 FPGA solution?
I'm sure there are some products which are not penny-pinching, lowest cost production. However, going from a $8 chip that almost gets you there, to a $30, 40 or more board is an issue of more than mere pennies.
I can't remember Ken's 5 primary P2 needs, speed may or may not have been on it.
Without any idea of actual customer requests, which only Parallax would know, I think I'd be safe in saying that at double, triple or more the cost of P1 alone, thats getting a bit expensive and another mcu might be worth looking at.
But thats just me trying to be rational without knowing User requests, volume, etc. I could be wrong certainly, however I'll play bad cop until/unless I get more information that indicates there is any real market potential for such.
OK, from the perspective of potential Prop-users looking to bump up a P1 FPGA before P2.
Sure, Parallax could do that, however I'm kinda thinking just the lawyers fees alone in getting that all set up as an ongoing concern might eclipse the potential profit.
The only way it make sense in the 3-4 minutes I've thought about it, would be for those people who want a P1 Hi-Speed variant as JMG suggested. Otherwise, $8 P1 works whether you need 4/5/6/7/8 cores.
So, P1 Hi-speed, what can Parallax reasonably charge?
Is that just for the P1v, does that come rated at a certain speed with a certain FPGA, what if they want 200Mhz? Do the fees take into account Buyer engineering R&D, or does the image come able to run at that speed?
What happens when there are liability concerns, those get expensive when both sides have lawyers. Does Parallax spend the R&D to determine what FPGA families Xilinx/Altera/Intel are warranted, or when FPGA features change or spec changes, who responsible?
Biggest thing is, are there any current volume users asking for such a thing, and, are they willing to buy enough to even make it worth Parallax's while to investigate?
I personally think you'd have to be the hardest of hardcore Prop devotees to be willing to seriously look at this. Many in the forum would do it.
But someone would have to spend the dollars on R&D. Although, now that I think of it, Parallax has just gone through this process, albeit for the P2.
I suppose they could reuse a lot of that work for a full blown P1 (is that done already?) on a specific FPGA, with board layout, etc, and see if there are any takers.
Before any of that is done though, Parallax should be able to give an actual board cost figure.
If Parallax were to estimate it would come in at $45 in Qty 1 from the website, they could check with their volume/specialty customers and the forum to see if there were any takers before expending any capitol.
I think JMG could probably post an approximate cost just of the top of his head.
Just not sure there is any market of real worth for a 200Mhz P1, though interested in seeing myself proven wrong.
OK, now thats a point I can understand.
However, realistically, what are we talking about BOM-wise/end board cost for a 200mhz P1 FPGA solution?
See above, $2.84/1k gives you (roughly) 1 P1V COG - if you are talking about chip level customers, then chip level prices apply
I'm sure there are some products which are not penny-pinching, lowest cost production. However, going from a $8 chip that almost gets you there, to a $30, 40 or more board is an issue of more than mere pennies.
Modules are a different market, and just how much the module costs depends on how you margin that $2.84 device. As this is Parallax, the main focus for them I'm thinking would be Education & low-medium industrial users.
It can be revealing to visit Digikey links like
http://www.digikey.com/product-search/en/programmers-development-systems/evaluation-boards-embedded-mcu-dsp
http://www.digikey.com/product-search/en/programmers-development-systems/evaluation-and-demonstration-boards-and-kits
http://www.digikey.com/product-search/en/programmers-development-systems/evaluation-boards-embedded-complex-logic-fpga-cpld
and sort by either Price, or Stocking levels, to get a feel for industry price-curves.
There is the revenue potential for Parallax, whoever is seeking a license, and potentially some middle person / integrator type.
Attorney costs can be kept very low by not doing a public license offering. I'm really serious when I say it should be limited to those able to produce some business model or other. Just for a while. That can keep the risk and cost compartmentalized, which would be good.
Some other use cases are falling out here:
1. Modules. Psomething go small, big, fast, more I/O.
Wouldn't the "more I/O" case be worth an FPGA frequently? We've all seen that a P1 could easily be driving 64 pins, if not more. Dual props, feature limits, compromises all have happened due to lack of pins, speed and memory.
2. FPGA version of the Proto Board.
It runs a Prop image, provides I/O, and the bare minimums beyond that.
Re "Ken's 5 key features" - here Ken identified Code Protection, Analog, bigger RAM, more pins as fundamental. This coincides with Chip's epic thread title, and his recent confirmation of expected feature set.
There may be more cogs in the mix, but lets see how the design fits and what power it takes before setting cog number expectations in stone.
Its worth noting that 4 of these key features are proven and/or relatively straightforward using P1V. And analog should be easy with Max10. OzPropDev's generator tool breaks down several barriers and comes recommended for use with the expanding range of cheap fpga dev boards.
If you want to do commercial/industrial P1V development, as Heater pointed out recently and Ken indicated in the past, contact Parallax (after Ken surfaces for air)
P2 "Hot" option:- don't rule this out, at this point it is still a potent option (at least for development), and it already exists, and entry price is a DE0 Nano. Breakouts are available. Lots of what Chip is aiming for is already in this incarnation, it won't be that hard to migrate. Its got hubexec, more pins, more ram, and similar mips per cog as likely in the final incarnation. Recognize the math will be done via a hub cordic, and hardware tasks will need to migrate to their own cog.
I'm interested in why the above two options don't appeal to the Osborne club waiting for the next incarnation. What are the key turn-offs, in rank order?
I want to flag a risk/threat I haven't seen raised by others - hardware patents. There are a lot out there. Finding/proving P2 (accidently) infringes another patent/submarine patent is a lot easier if Parallax releases P2 verilog in a wide public release, so the solution is to release binaries or go narrow, at least until development money is recouped.
Or just do the soft-release that can be patched, but that'd be an underwhelming result from where Chip says things are at now