From what I understand, the Zii platform is a collection of a bunch of "generic" processing elements (or "PEs" as the Zii people call them) working in conjunction together to make a larger processor. My guess are that these PEs are somehow "general purpose" ALUs/FPUs - a discrete circuit that is capable of adding two numbers together and returning the sum. In computers today, it isn't really this simple because different types of operations are best performed by different circuit elements. The obvious example is that ALUs are good at adding two integer numbers together - by far the most common type of calculation - while FPUs are used for dealing with floating point numbers (decimal numbers) - much more important in graphical and mathematical applications. My guess is that the Zii PEs are ALUs that can also somehow perform floating point calculation, thus making them "general purpose" processing elements.
The advantage of this, as their promotional material suggests, is that you have more flexibility in bringing the resources of the chip to bear for the task at hand. A Zii chip won't necessarily be more powerful than a conventional chip of the same density (unlike what the marketing suggests), and the operating system is still under the onus of smartly managing its limited resources but will be able to bring a greater maximum proportion of its resources to the task at hand, whether that be stuff as different as gaming and word-processing.
All of what I've just said is conjecture, as I really haven't been able to look at the design specs of the Zii chip, so I can't really say what's inside. But if what I'm guessing is true, there are also a few doubts/problems with this architecture that the pot-of-gold marketing glosses over:
One problem is whether or not Zii PEs can truly be considered "general purpose" calculators - and by that I mean - be able to handle fixed and floating point calculations. With the marketing "Zii is only good for the computing that matters to you" grok, I'm guessing the answer is no. Zii probably doesn't truly support floating point calculations, but instead uses fixed-point calculations to some degree of accuracy to simulate floating point operations. If the Zii is really smart, it probably links some number of PEs together to increase the degree of accuracy/range to support calculations that require more significant digits. (For example, if one PE can handle 32 "significant digits", chaining two together will allow you to handle 64.) This is probably enough for most basic floating point calculations, but really doesn't cut it when high degrees of accuracy are required.
The second problem is whether or not the Zii can really scale up as the marketing suggests. You can't simply add more PEs to a Zii chip because some controller is likely responsible for doling out tasks and controlling the various PEs. To make use of an increased number of PEs, the controller (which is basically another specialized circuit) has to be buffed up to manage these new resources. That's why there are only 20-something PEs per Zii chip; that's probably as much as a not-too-powerful (or else it will eat up too much battery) controller on the Zii chip can handle. This is why when the promotional material says that you can string together a bunch of Zii >chips< to increase processing power; these are whole Zii chips we're talking about, and as discrete processors, this actually is not that much different that any other massively multiprocessor environment, with the same loss of performance and synchronization problems as scale increases.
Anyway, if conjecture holds true, the Zii is probably a good solution for the limited computing/power environment of mobile devices. In an environment with constrained resources, being able to bring more of what you already have to the important tasks at hand - this ability along with a smart OS can produce mobile devices that are more responsive and seem more powerful than a similarly dense conventional processor. I really don't see Zii-style architecture becoming popular in desktop or even most laptop environments where those kind of constraints are not that important nor as restrictive.
PS: I can't consider myself a "master" of computer architecture, so take my word with a grain of salt. Also, if you happen to know more than me and wish to correct, please feel free, but also please be kind :).
Now that we've thrown 'em off the trail, use the form below to get in touch with the people at Engadget. Please fill in all of the required fields because they're required.
*scratches head..*
Okay, SERIOUSLY! Are there any engineers in the audience? Can we get some hypotheses on WTH is so promising or revolutionary about this freakin' chip?
I say, it will be awesome.
Happy?
From what I understand, the Zii platform is a collection of a bunch of "generic" processing elements (or "PEs" as the Zii people call them) working in conjunction together to make a larger processor. My guess are that these PEs are somehow "general purpose" ALUs/FPUs - a discrete circuit that is capable of adding two numbers together and returning the sum. In computers today, it isn't really this simple because different types of operations are best performed by different circuit elements. The obvious example is that ALUs are good at adding two integer numbers together - by far the most common type of calculation - while FPUs are used for dealing with floating point numbers (decimal numbers) - much more important in graphical and mathematical applications. My guess is that the Zii PEs are ALUs that can also somehow perform floating point calculation, thus making them "general purpose" processing elements.
The advantage of this, as their promotional material suggests, is that you have more flexibility in bringing the resources of the chip to bear for the task at hand. A Zii chip won't necessarily be more powerful than a conventional chip of the same density (unlike what the marketing suggests), and the operating system is still under the onus of smartly managing its limited resources but will be able to bring a greater maximum proportion of its resources to the task at hand, whether that be stuff as different as gaming and word-processing.
All of what I've just said is conjecture, as I really haven't been able to look at the design specs of the Zii chip, so I can't really say what's inside. But if what I'm guessing is true, there are also a few doubts/problems with this architecture that the pot-of-gold marketing glosses over:
One problem is whether or not Zii PEs can truly be considered "general purpose" calculators - and by that I mean - be able to handle fixed and floating point calculations. With the marketing "Zii is only good for the computing that matters to you" grok, I'm guessing the answer is no. Zii probably doesn't truly support floating point calculations, but instead uses fixed-point calculations to some degree of accuracy to simulate floating point operations. If the Zii is really smart, it probably links some number of PEs together to increase the degree of accuracy/range to support calculations that require more significant digits. (For example, if one PE can handle 32 "significant digits", chaining two together will allow you to handle 64.) This is probably enough for most basic floating point calculations, but really doesn't cut it when high degrees of accuracy are required.
The second problem is whether or not the Zii can really scale up as the marketing suggests. You can't simply add more PEs to a Zii chip because some controller is likely responsible for doling out tasks and controlling the various PEs. To make use of an increased number of PEs, the controller (which is basically another specialized circuit) has to be buffed up to manage these new resources. That's why there are only 20-something PEs per Zii chip; that's probably as much as a not-too-powerful (or else it will eat up too much battery) controller on the Zii chip can handle. This is why when the promotional material says that you can string together a bunch of Zii >chips< to increase processing power; these are whole Zii chips we're talking about, and as discrete processors, this actually is not that much different that any other massively multiprocessor environment, with the same loss of performance and synchronization problems as scale increases.
Anyway, if conjecture holds true, the Zii is probably a good solution for the limited computing/power environment of mobile devices. In an environment with constrained resources, being able to bring more of what you already have to the important tasks at hand - this ability along with a smart OS can produce mobile devices that are more responsive and seem more powerful than a similarly dense conventional processor. I really don't see Zii-style architecture becoming popular in desktop or even most laptop environments where those kind of constraints are not that important nor as restrictive.
PS: I can't consider myself a "master" of computer architecture, so take my word with a grain of salt. Also, if you happen to know more than me and wish to correct, please feel free, but also please be kind :).
Whoa. Sorry, now that I look at it, that got loooong. >_
The Zii platform will allow scaleable, low-watt, cost-effective Engadget comment editing.
@Anthop: Please, please, no apologies! Thanks for much for punching all that out. Very much appreciated.