One of the things Ken mentioned it another thread, ..., that Parallax was not about attacking competitors or doing comparisons.
I don't think it would help to sell more Props if Parallax did show a comparison table between the Prop and other similar competative devices. To be honest, the Prop wouldn't fair well in a traditional comparision using benchmarks, and peformance versus price. Of course, that shouldn't prevent us as users from developing such a table. Perhaps Leon already has one handy.
The Prop certainly has some unique features, such as the parallel processors and video generators. I'm not sure if there are similar chips that have those features. Maybe Leon could enlighten us on that.
I think the allure of the Prop to the hobbyist is the ease of use, the uniquess of the chip and the support from Parallax. Personnally, I find the Prop to be a very interesting chip to tinker with. The Spin language and it's interpreter are fasinating, and I've enjoyed the challenge of understanding how it works. I've often wondered why Parallax developed the Spin language when they could have just extended the Stamp Basic language or chosen a subset of an existing standard language. I suppose a lot of it had to do with implementing the interpreter in 496 longs of memory, but they probably could have done that with Basic or a subset of C.
The video generators, counters, and the PLL, perhaps, seem to be the unique features to me, and something similar to the video generator has been implemented in software on the device which must not be named, by one of the regulars on this forum! I think that most of the counter capability could be emulated, as well.
I could draw up a table comparing that device with the Propeller, but I don't think that it would go down very well. It has been said that comparisons are odious, anyway!
The device which must not be named has eight hardware threads per core which can be treated as separate 50 MHz processors running in parallel. They switch in one clock. The single-core chip sells for $7.50.
There is also Chuck Moore's new chip with 144 processors. It will cost $20 per chip, initially, but the price is bound to come down if it takes off.
I'm just trying to point out that that no one chip is the best, it all depends on the application. A competent designer chooses the best chip for a particular design and doesn't get fixated on a single device. There is no place for "fan boys" when it comes to designing hardware and software systems.
People here are remarkably easy to wind up with a few innocuous statements. This forum seems to be unique in that respect!
I came to this discussion late but I'm part of a community that was introduced to programming through Pbasic and Spin. I've no urge to learn how to program interupts. I always thought Apple was a better innovator but Microsoft was a better marketer. Personally, I think other chipmakers should catch up to multicore technology. The Parallax company and community are well supplied with tech genius. Marketing requires different abilities. Maybe more focus needs to be aimed at promotion. (Just my 2 cents)
I'm just trying to point out that that no one chip is the best, it all depends on the application. A competent designer chooses the best chip for a particular design and doesn't get fixated on a single device. There is no place for "fan boys" when it comes to designing hardware and software systems.
People here are remarkably easy to wind up with a few innocuous statements, of course.
I understand Leon, I totally get the "fan boy" thing.
I just was hoping for a positive comment about the Prop, even if it was as simple as the DIP package is good for quick prototyping...
Leon, is the chip you can't name the LPC2148 ARM7TDMI that you mentioned in an earlier post? When you say it does a one-clock thread switch does that include the stack pointer and all the CPU registers? How often can the content be switched? Is it possible to execute a single instruction, and then switch to the next task? How much peformance loss would there be if that was done -- 50%?
I looked at a few PDFs on the "X" chip. It's a nice chip, and I can see that it would work well for a lot of applications. The hardware task-swapping would make interrupt handling almost painless, and it almost simulates having 8 independent processors. The 16 registers per instance may be a bit limiting for some applications. A prop cog can trade-off using its 496 longs for register space or program space. The "X" chip won't be able to achieve the minimum latency that a cog can do. The "X" chip is certainly a better solution for C programming. The Prop is better for applications that require low-latency interaction. Both chips have their pros and cons. I think it's good to point them out, but I don't think it serves a useful purpose to do it on a weekly basis.
I only mention it when people claim that the Propeller is unique in terms of parallel processing and deterministic operation. I suppose that the message will get across eventually.
The video generators, counters, and the PLL, perhaps, seem to be the unique features to me, and something similar to the video generator has been implemented in software on the device which must not be named, by one of the regulars on this forum! I think that most of the counter capability could be emulated, as well.
Hi Leon,
Which device are you talking about, and why must we not name it? Is this particular device a wizard that can detect when its name is spoken, like Voldemort? And then comes and fries all your ICs so that you have to submit and buy a batch of 100,000? j/k of course
I think the Propeller has a bright future ahead of it. Just as long as we can keep the Arduino craze at bay until the Prop 2 comes out, than we should do fine. The Propeller has many things that other hobbyist uC's do not have, the main being video/ audio generation... The Prop doesn't really have much of a hold in commercial markets, because these big companies think Parallax only makes hobbyist equipment/ novelties. Although that assertion is far from the truth, what company would really want to switch over to using a Propeller for it's products? The Propeller is overkill for small applications, like thermometers and detection systems and it is underpowered for intense graphic and more computer like functions. That narrows it right down:) The current Propeller is a interesting anomaly. It is a mixture of a easy to use device, but it is more flexible than other hobby micros... The Propeller 2, i think, will really kick some butt. In the hobbyist and commercial world alike. It has some REAL potential...
Speaking purely from my first-hand experience of trying to sell the Propeller to my various bosses for a specific embedded application, the two biggest reasons the Prop gained no traction was:
* No integrated 12-bit ADC & DAC
* Cost was 2x the ARM which *did* have a multiplexed ADC and several DACs
I think that ultimately I could have overcome the other objections. But I couldn't overcome these.
I guess the point is that it's a very competitive world...and a very fluid world.
I simply can't imagine the Prop ever selling outside the relatively small universe of individuals and organizations that appreciate its unusual aspects or else have a particular application optimally suited to it, like quadcopter flight control.
I only mention it when people claim that the Propeller is unique in terms of parallel processing and deterministic operation. I suppose that the message will get across eventually.
It is funny you say that, I was just checking out t'other chip out of interest and watched a video about motor control using t'other chip:
I would have to say that Parallax's web site is WAY better than XMOS's The OBEX is awesome and the user forums are top notch here. I'm currently learning the XMOS chip because I need more speed than the propeller can provide but don't (at this point) see myself choosing XMOS for a design unless I need a lot of speed.
As it stands, I love how easy the propeller is to use and am choosing it for new designs that aren't cost sensitive.
Okay I realize there in no single best chip, and there will almost always be the right chip for the job.
But let's face it, unless there are some serious comparisons, Parallax's Propeller will have a hard time being taken seriously. Without some comparsions, I think it will be an uphill struggle to get into the commercial and industrial markets.
Once again, that is just my opinion, and not based on any known facts.
There is a battle currently going on between TI and Microchip over which company makes the chips with the best low-power performance. Both companies have produced documentation claiming that their chips outperforms the other's, but it's virtually impossible to say which is right. I think that the same will apply when comparing the Propeller with other devices, because it is so application-specific.
There is an old saying that bad press, is better than no press. The whole idea is to get the capabilities and the Propeller name in front of the engineers.
The Propeller model of concurrency is different than the Xmos one. I think it's worth recognizing those differences. On the Prop, the cogs run in a entirely separate memory space. This is unique, in that the COG is completely decoupled from both the other COGs and the HUB memory when executing. That's true concurrency. The round robin HUB access scheme also means deterministic shared memory access, again in true concurrent form, in that each COG access window does not depend on what the other COGs are doing.
The Xmos system does not have those attributes. Scenarios can be contrived that approach what the Propeller does "in the box, with NO EFFORT". Let's make that clear right along with the overall, "it can be done on Xmos", which it can clearly, but at a significantly higher effort and overall complexity.
Re: Something similar to video generators. Yes. A Xcore can be used in that fashion, but again, it's not the same as what the Propeller does. Any Propeller COG can use the video generator, IN PARALLEL with ordinary compute functions, to generate a signal, with data contained in the COG, or the HUB, or other storage source. (arguably, "other source" is complex and limited). Multiple displays are possible, just as easily as single displays, only limited by overall RAM. Again, the same basic things are in play. It's possible to do, but more complex, and with some different dynamics, limitations and implications in play. Those are not insignificant things.
Just my own take on that. The scale of the devices is different, and co-mingling that with the basic functional attributes is often a poor comparison.
Comments
The Prop certainly has some unique features, such as the parallel processors and video generators. I'm not sure if there are similar chips that have those features. Maybe Leon could enlighten us on that.
I think the allure of the Prop to the hobbyist is the ease of use, the uniquess of the chip and the support from Parallax. Personnally, I find the Prop to be a very interesting chip to tinker with. The Spin language and it's interpreter are fasinating, and I've enjoyed the challenge of understanding how it works. I've often wondered why Parallax developed the Spin language when they could have just extended the Stamp Basic language or chosen a subset of an existing standard language. I suppose a lot of it had to do with implementing the interpreter in 496 longs of memory, but they probably could have done that with Basic or a subset of C.
Dave
I could draw up a table comparing that device with the Propeller, but I don't think that it would go down very well. It has been said that comparisons are odious, anyway!
There is also Chuck Moore's new chip with 144 processors. It will cost $20 per chip, initially, but the price is bound to come down if it takes off.
Should we just pack up and go home because we aren't "the best".
I know you can find something positive to say, I think you just have more fun poking us.
C.W.
People here are remarkably easy to wind up with a few innocuous statements. This forum seems to be unique in that respect!
I understand Leon, I totally get the "fan boy" thing.
I just was hoping for a positive comment about the Prop, even if it was as simple as the DIP package is good for quick prototyping...
C.W.
It's an easy chip to use.
Low entry cost.
The DIP option is nice for hobbyists.
Support is very good.
It can be very cost-effective in some applications.
It's very reliable.
Low-cost boards are available.
Lots of fun things can be done with it.
Etc., etc.
Leon, you made me smile today, and that isn't always easy.
Thanks,
C.W.
Second that:)
I looked at a few PDFs on the "X" chip. It's a nice chip, and I can see that it would work well for a lot of applications. The hardware task-swapping would make interrupt handling almost painless, and it almost simulates having 8 independent processors. The 16 registers per instance may be a bit limiting for some applications. A prop cog can trade-off using its 496 longs for register space or program space. The "X" chip won't be able to achieve the minimum latency that a cog can do. The "X" chip is certainly a better solution for C programming. The Prop is better for applications that require low-latency interaction. Both chips have their pros and cons. I think it's good to point them out, but I don't think it serves a useful purpose to do it on a weekly basis.
Dave
I only mention it when people claim that the Propeller is unique in terms of parallel processing and deterministic operation. I suppose that the message will get across eventually.
Hi Leon,
Which device are you talking about, and why must we not name it? Is this particular device a wizard that can detect when its name is spoken, like Voldemort? And then comes and fries all your ICs so that you have to submit and buy a batch of 100,000? j/k of course
I thought I read the documentation properly, but I must have gotten something wrong. Can you educate me on which of my statements was incorrect?
Thanks,
Dave
* No integrated 12-bit ADC & DAC
* Cost was 2x the ARM which *did* have a multiplexed ADC and several DACs
I think that ultimately I could have overcome the other objections. But I couldn't overcome these.
I guess the point is that it's a very competitive world...and a very fluid world.
I simply can't imagine the Prop ever selling outside the relatively small universe of individuals and organizations that appreciate its unusual aspects or else have a particular application optimally suited to it, like quadcopter flight control.
It is funny you say that, I was just checking out t'other chip out of interest and watched a video about motor control using t'other chip:
http://www.xmos.com/development-kits/motor-control-platform
They make exactly the same claims for t'other chip at around 1:45
Graham
Apologies, I came over all Yorkshire.
As it stands, I love how easy the propeller is to use and am choosing it for new designs that aren't cost sensitive.
PM me, please. Parallax doesn't like discussions of competing products on the forum.
Ee, bah gum! That sounds messy!
They probably don't know about the Propeller, perhaps you ought to enlighten them.
The Propeller won't work in that application, of course. The Prop II might, but I doubt it.
But let's face it, unless there are some serious comparisons, Parallax's Propeller will have a hard time being taken seriously. Without some comparsions, I think it will be an uphill struggle to get into the commercial and industrial markets.
Once again, that is just my opinion, and not based on any known facts.
Bruce
Leon, PM sent.
There is an old saying that bad press, is better than no press. The whole idea is to get the capabilities and the Propeller name in front of the engineers.
Bruce
The Propeller model of concurrency is different than the Xmos one. I think it's worth recognizing those differences. On the Prop, the cogs run in a entirely separate memory space. This is unique, in that the COG is completely decoupled from both the other COGs and the HUB memory when executing. That's true concurrency. The round robin HUB access scheme also means deterministic shared memory access, again in true concurrent form, in that each COG access window does not depend on what the other COGs are doing.
The Xmos system does not have those attributes. Scenarios can be contrived that approach what the Propeller does "in the box, with NO EFFORT". Let's make that clear right along with the overall, "it can be done on Xmos", which it can clearly, but at a significantly higher effort and overall complexity.
Re: Something similar to video generators. Yes. A Xcore can be used in that fashion, but again, it's not the same as what the Propeller does. Any Propeller COG can use the video generator, IN PARALLEL with ordinary compute functions, to generate a signal, with data contained in the COG, or the HUB, or other storage source. (arguably, "other source" is complex and limited). Multiple displays are possible, just as easily as single displays, only limited by overall RAM. Again, the same basic things are in play. It's possible to do, but more complex, and with some different dynamics, limitations and implications in play. Those are not insignificant things.
Just my own take on that. The scale of the devices is different, and co-mingling that with the basic functional attributes is often a poor comparison.