Will the propII be cheaper or more expensive?
ErNa
Posts: 1,752
More cogs will allow for more parallelized program concepts. The same goal can be reached with better result, using more cogs (sometimes). But anyway, the price of the product should not change. So the question: will the more powerfull Prop II be also more expensive?
Comments
If Moor's law is somewhat correct, There are no bounds to what the propeller may do in the future !
Rob7
It will most like end with a compact black whole on everybody's desk
On linear CPU's (mostly linear with cores and all that) reaching peak speed on single threaded, largely sequential tasks is easy. Just compile the Smile out of it and it's gonna run fast. Also, pre-execute as much as you can and slap a big *** heat sink on to deal with the inefficiency.
On the prop, getting that peak speed for largely single threaded stuff is tough, if not impossible. However, doing lots of things at once is a lot less complex. And, given the overall speeds we can do today, a growing niche, IMHO.
There is no free lunch. If the Prop makes multi-processing easier and more robust (and it does), then peak speeds will be lower, or complexity will rise. It's the classic triangle. Pick two and deal.
I'm thinking ahead to Prop III. By then it will be a full on CPU, if Parallax goes that way. The kinds of devices it will generate will be very interesting variations on what we have now. We are at the limits of peak compute. Stepping up from here is really ugly. Either there is a lot of heat and latency, or we distribute tasks.
Multi-core CPUs are common now, so distributing tasks is in!
We may well see the Prop well positioned in a few years, with software techniques that follow this line of thinking growing more mature industry wide in general.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller Wiki: Share the coolness!
You know what would be really cool? A chip that allowed you to build an actual computer system that put you somewhere between a Commodore 64 & whatever is out there today.
It would need to be powerful enough to get a real operating system and real productivity software running, but limited enough that you couldn't just throw code bloat at the "faster processor". Maybe to create something like what RiscOS is today. Or what the Amiga was. Or Classic Macs.
Then what would be a good next step would be to create a variation of the Hydra that wasn't geared towards game consoles, but comuter systems. It would be sort of like when the Apple was created. Computers weren't just a bunch of boxes from Newegg. they were fun in a different way that seems to be missing today.
I know the current Propeller is a start, but maybe another line altogether. Call it the Jet or Rocket or something like that.
There's a bunch of us 'old timers' who have fallen in love with Propeller.
However, you don't have to wait, grab a Protoboard and start building..
(All that lacks is a case) -- I did build a picaxe keyboard controller for my
Propeller, so infact I do have one mounted in a Commodore 64 keyboard casing. [noparse]:)[/noparse]
Oldbitcollector
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
The comments and code above are proof that a million monkeys with a million propeller chips *could* write Shakespeare!
The biggest problem is that you need to reconsider how the programs are designed. And you probably have to give up on WYSIWYG wordprocessing.
(While a GUI is nice, it doesn't define what is or isn't a real computer system. Back in the heyday of mainframes, you could have multi-user systems with 64KB RAM with the users sitting at text-terminals.)
Other than that, most of the work is already done if you look at the Propeller OS thread, or the FORTH thread.
There's a couple of BASIC threads, too, and with the forthcoming C compiler(using the 'large memory model' concept) it should be easier to port 'utility' programs.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Don't visit my new website...
Prop II looks to be powerful enough to boot strap that. With the C complier, and an on chip memory size big enough to do things, we are highly likely to see a step forward that direction.
The big obstacle is addressing memory period. That's really the only downside. A prop MMU, handled by a paging COG, running some sort of low level kernel an OS could leverage would do the trick. I'm sure there are lots of other options as well. Maybe that's not the biggie either. It's just something I wonder about.
I'm also really curious about relocatable large memory model code. The up and coming speed increases in COG - HUB communication bandwidth are gonna be sweet. I can't yet visualize how code moves into and out of main on-chip RAM though. Assuming those things happen nicely, somehow, we then end up with this interesting environment.
One COG doing memory paging, another one running an OS in large memory mode. A few of them doing graphics, sound, user input I/O and other I/O, leaving a few for user programs. Perhaps we might be able to start up a COG, ask for an on chip memory space, knowing the other things running will respect it, ask for a off-chip memory space, in like kind, and go do stuff on the free I/O pins, not being used by the system.
We might also ask the OS for a thread too. Maybe supervisory kinds of programs can run this way. Lots of very interesting ways to go with it.
Once running, the user could then just use the on-chip tools, perhaps some system debugger, interpeter, etc... to query the program, watch it's RAM, change values, etc...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller Wiki: Share the coolness!
To be able to get the X, Y coordinate values off a color blob ?
So as to drive a bot towards it.
Or is it only a dream ?
JP
BTW, it's bad form (and inefficient) to hijack an old thread (start a new discussion on a different, possibly related topic). If your topic is valuable, start your own thread with the question.
more expensive than what ?
More expensive than a Prop 1, certainly. ISTR ~ 1.5x-2x as indicators ?
The die is as large as will fit into the TQFP128 package, and testing time (cost) will certainly be non-zero..