Intel Atom: Did Pandora Choose The Wrong Cpu?

[icurafu]

Except for compatibility what would that bring you?

Except? Why would anyone buy a Pandora anyway? Except for all the things it can do, it's just a hunk of plastic.

I'm aware that the x86 architecture is old and not well suited to mobile devices. I know that there are better, more efficient alternatives and that software designed for these will run just as well/speedily. I'm much more interested in software that is not designed for this device, specifically.

The community here is great, I'm expecting lots of great things, but it's a big wide world out there and we can choose to join in and share and benefit, or we can choose to shut ourselves off in the name of efficiency.

This is all pretty speculative, but in my mind an ideal computing environment is not one that is so heavily segmented.


Seriously, get a Eee PC or another UMPC (like I did). Or wait until June and get a MID that you can install WinXP on.

That's not what this is.

 

[AireTamStorm]

x86 was designed 30 years ago. This essentially means that x86 was designed with the needs of people from 30 years ago. Did people 30 years ago have a big need for mobile computers? No.

Frankly, x86 is only so successful thanks to microsoft, seeing as they only compile their OS for that CPU architecture.

For Performance / Watt, Atom still lacks in comparison to some VIA chips. ARM is a very nice architecture, general math runs very fast for the clock speed, and there are some extensions to various ARM chips that help specialize them when optimized.

Intel's solution to everything is to up the clock speed, where most other processors up the instructions / clock cycle. Every 8 years they do a major overhaul of their CPUs but its normally "use more power!" or more recently: "slap a few more cores in there!"

 

[ldesnogu]

I'm aware that the x86 architecture is old and not well suited to mobile devices. I know that there are better, more efficient alternatives and that software designed for these will run just as well/speedily. I'm much more interested in software that is not designed for this device, specifically.

The community here is great, I'm expecting lots of great things, but it's a big wide world out there and we can choose to join in and share and benefit, or we can choose to shut ourselves off in the name of efficiency.

This is all pretty speculative, but in my mind an ideal computing environment is not one that is so heavily segmented.

If you are interested in existing software go buy an eeePC or any other forthcoming Atom based UMPC and choose one with a MS OS. Because the logical next step after choosing standardized hardware is of course choosing a standardized OS.

If you don't want segmentation, that means you want unified stuff, the best way to kill innovation. So OK let's all go to x86 and MS and let them control our HW, our wallet and the way we have fun with our computers

Frankly, x86 is only so successful thanks to microsoft, seeing as they only compile their OS for that CPU architecture.

That's not true, MS OS *is* portable to other architectures, there just is no market for that

QUOTE
Intel's solution to everything is to up the clock speed, where most other processors up the instructions / clock cycle. Every 8 years they do a major overhaul of their CPUs but its normally "use more power!" or more recently: "slap a few more cores in there!"

I don't like the x86 architecture that's a given But Intel has done tremendous things from an architectural point of view. And they had to do that or else their CPUs would be extremely poor performance wise. And why did they have to improve performance? Because we live in a segmented world where others were running circles around Intel CPUs

 

[shmuck]

Alright, y'all can quit it with the "go buy an Eee" nonsense. Obviously, I'm here because I like something that the Pandora has to offer. If I'd wanted an Eee, I'd have one.

I had thought of the Windows analogy. The fact is, if Windows were an open and configurable platform that did not subject me to the whims of some parent corporation, I would probably use it exclusively in spite of the existence of technically superior options. I choose Linux for the freedom, not because it crashes a little less often / is a little less vulnerable / etc.

Given that much of the focus of the Pandora is directed at emulation (i.e.: cross platform compatibility) I'm a bit surprised at the dedication to a less prominent architecture. Let me ask this: on this theorized Pandora 2, which configuration can run Fallout: yours, running on some Arm chip, or mine, running on an x86? How 'bout Starcraft? There are literally tens of thousands of games where the answer is "only mine".

I do not think that a unified platform kills innovation. I can't think of any evidence to support that. I can certainly think of examples where unique limitations have spurred innovation, but I don't think that those things are equivalent.

This is a good point:

And why did they have to improve performance? Because we live in a segmented world where others were running circles around Intel CPUs

I'm not suggesting that we stick to x86 forever though, I'm just suggesting that we be a bit more organized about it. Should be treating this more like a standard. Apple, I think, handled their switch to x86 fairly smoothly and in a manner that should really be exemplary to the industry as a whole.

It's the fragmentation of the market that prevents this sort of unified move and sticks us with the same architecture that we've been using for the last thirty years. Not that I don't realize that what I'm proposing wouldn't be an intense bureaucratic mess, but at least it would happen. At this rate x86 is never going to go away.

Anyway, to sum up, here are my options:


1. There's a computing emergency. This software needs to be run, and now, but there isn't a desktop or laptop in sight!

"I'll do it," I say confidently, pulling my Pandora out of my pocket, knowing that whatever it is, I can probably run it and save the day. And I can. And I do. And here come the ladies, oh so grateful.


2. There's a computing emergency, etc.

"Hey," I say. "Is the source available? Is this configured for Arm? We have time to get everything recompiled, right?" I think you can see where this is headed.

As it turns out, much as I may have talked it up, the ladies just weren't impressed by my slightly better performance per watt.

 

[ldesnogu]

Given that much of the focus of the Pandora is directed at emulation (i.e.: cross platform compatibility) I'm a bit surprised at the dedication to a less prominent architecture. Let me ask this: on this theorized Pandora 2, which configuration can run Fallout: yours, running on some Arm chip, or mine, running on an x86? How 'bout Starcraft? There are literally tens of thousands of games where the answer is "only mine".

My personal usage for the Pandora would not be emulation, and certainly not emulation of old Windows games. For a small portable device, I want small games that don't last forever such as arcade and console games. Also where are the tens of thousands of interesting games on Windows? I guess you can bring that down to at max a thousand, and this is about the number of interesting games that run on consoles and arcades that could be emulated on the Pandora.

Yeah that'd be great to go around bragging about my small portable system running a full-blown RTS. Except I would *never* play, for instance, any RTS on a small screen. I play them at 1920x1200.

(Note these are all personal point of views .)

Anyway my interest in the Pandora is not in emulation first, it's that I like its chip and would love hacking it And I have no fun in programming any x86 at assembly level.

QUOTE
I do not think that a unified platform kills innovation. I can't think of any evidence to support that. I can certainly think of examples where unique limitations have spurred innovation, but I don't think that those things are equivalent.

If AMD hadn't put out its Athlon do you think we would have these wonderful pieces of engineering called Core? And if MIPS, Alpha and other RISC chips hadn't have some success do you think the x86 would be where it is now?

I agree that unification has some benefits from a consumer point of view in the short term, but I still stand to my claim that too much unification slows down or kills innovation in the middle and longer term. Let's all embrace OOXML and x86, and see what happens.

QUOTE
I'm not suggesting that we stick to x86 forever though, I'm just suggesting that we be a bit more organized about it. Should be treating this more like a standard. Apple, I think, handled their switch to x86 fairly smoothly and in a manner that should really be exemplary to the industry as a whole.

Apple could handle their switch because they keep the control over all of their platform at the hardware and at the software level. Apple machines are nice, but they don't make me dream as a hacker. They are nice in some market segments, not in mine.

QUOTE
It's the fragmentation of the market that prevents this sort of unified move and sticks us with the same architecture that we've been using for the last thirty years. Not that I don't realize that what I'm proposing wouldn't be an intense bureaucratic mess, but at least it would happen. At this rate x86 is never going to go away.

I am afraid I don't get your point. What exactly are you proposing?

QUOTE
Anyway, to sum up, here are my options:

1. There's a computing emergency. This software needs to be run, and now, but there isn't a desktop or laptop in sight!

"I'll do it," I say confidently, pulling my Pandora out of my pocket, knowing that whatever it is, I can probably run it and save the day. And I can. And I do. And here come the ladies, oh so grateful.

2. There's a computing emergency, etc.

"Hey," I say. "Is the source available? Is this configured for Arm? We have time to get everything recompiled, right?" I think you can see where this is headed.

Hum, I guess thousands of packages already compiled for ARM don't count.
Also problem resolution on Windows machines is very easy: reinstall it or let it rot.

QUOTE
As it turns out, much as I may have talked it up, the ladies just weren't impressed by my slightly better performance per watt.

Were they more impressed by your UMPC running Starcraft?

 

[Firefox]

I think that there's a widespread public misconception that x86 is good because the Intel Core* and AMD Athlon processors are so fast. In fact, those processors are fast because:

* They have staggering quantities of cache memory built into them.

* They contain a large amount of circuitry dedicated to guessing what the program they're running is trying to do.

x86 instructions aren't a very good notation for expressing programs, so you have to take huge gobs of them, analyse what they're doing, come up with a more efficient way of doing the same thing on the fly, do that, then record the results in such a way that they're consistant with what the original x86 instructions would have done. That takes a lot of clever circuitry, which consumes a lot of electricity.

That is why fast x86 processors suck up as much juice as a large TFT monitor and need a hefty heatsink and fan bolting to them to stop them melting.

The Intel Atom has low-ish power consumption because they've thrown away most of this performance- enhancing circuitry. That not only means that they won't be anywhere near as fast as a Core 2 Duo running at the same clock speed, it means they won't be anywhere near as fast as a Pentium-M running at the same clock speed.

RISC processors like the ARM have a very powerful, expressive instruction set carefully designed from scratch that allows the programmer or compiler to explicitly tell the processor what to do in an efficient manner without needing all of this "magic guessing circuitry" to make the program run fast.

That's why RISC processors fit into such a small amount of circuitry, which in turn means they're cheap to manufacture and consume teensy amounts of power, which in turn means you don't have to cool them. Because they're so small you can build them into a chip that contains the whole chipset as well as the processor, which saves even more expense and electricity.

 

[ldesnogu]

Here is a very instructive read about "new" features to come in next generations of Intel chips: http://www.theinquirer.net/gb/inquirer/new...vx-instructions

That will perhaps open the eyes of the younger who don't get how unification has slowed down the evolution of the computer industry.

 

[ledow]

To those posting about the age of the x86 architecture:

You do know that ARM processors originally came from the company that made Acorn computers, back in the early 80's? Yes the x86 (although it wasn't called that back then because it started with older chips like the 8088, which was supposed to be a Z80-replacment/competitor, it even has a limited "z80 compatibitlity" mode) came first but ARM have been around since the 286/386 were the chips of choice, by which time most of what we call x86 architecture was only beginning to be standardised. That's WHY they were called ARM in the early days - the original A stood for Acorn, although now it stands for Advanced.

The age of the architecture has nothing to do with it, it's the *design* of the architecture. ARM was supposed to be a RISC chip from the ground-up which makes it scale well, and means the design is much simpler, much more compact and (usually) less energy hungry. x86 *wasn't*. Compilers have to jump through a few more hoops to make things work on ARM but, in general, ARM is just as competitive, if not more, in terms of performance per pound than most major chip makers. ARM just suits itself well to embedded design, x86 doesn't (it's for big desktops, really). Even the Intel "mobile" CPU's are horribly power-hungry for simple tasks.

Just consider, for a moment, your GP2X - 2 x 200MHz chips running off a set of double-A's for several hours, and that's with (let's be honest) very little effort put into power-saving in the software. How many pieces of GP2X software ramp the CPU speed *down* when they aren't doing much? How much power-saving does the machine actually TRY to do, other than what the chip does itself? x86 chips basically rely on the hope that they can switch a lot of stuff off when it's not being used, that's how "power saving" works on x86, but at "full speed", they literally suck power.

Slightly off-topic, modern desktop/laptop CPU's are a disgrace, just like the software that they need to run. You need GHz and GB to write a letter and print it out. ARM basically never forget that you can work wonders with just a few MHz if you put the effort in. I have an Amstrad Notepad sitting next to me here - A 4.77MHz Z80 that can boot instantly, last for hours on a set of *ordinary* AA's (it can last weeks on the 2700mAh I use in my GP2x), take megabytes of notes, and print it if I need to over serial or parallel ports. What the hell happened to those days? And now people are trying to "re-capture" that sort of performance (instant booting, low-power etc.) and claim that it's somehow new and revolutionary when they don't even come close.

 

[ldesnogu]

And now people are trying to "re-capture" that sort of performance (instant booting, low-power etc.) and claim that it's somehow new and revolutionary when they don't even come close.

Pandora is trying that and should be able to reach that goal (though I doubt it will be as power efficient as your old Amstrad note pad)

 

[StealthBagel]

X86 will eventually die and MS will either die (by putting all of their resources into that sinking ship that is Vista)

Quantum leaps rarely happen in this industry. Things slowly change and software evolves with it. It's also perfectly possible for one microarchitecture to interpret instructions meant for another, albeit more slowly and/or with a great portion of the die devoted to support hardware. The idea of Microsoft, the only company making a marketable desktop operating system for standard PCs, going under because their new OS is a bloated mess is outrageous.

In reality, a great deal of updates and service packs later, Vista will be (slowly) accepted. It'll be tough since it doesn't bring much useful features to the table compared to the 98SE transition to 2K/XP, but eventually it'll be the standard. XP was pretty bad at first too.

edit: x86 also doesn't create limitations that would stunt the further development of the microprocessor. The need for many layers of compatibility with older instruction sets means it's less ideal for a low-power simple design you'd expect in a mobile SOC, but these limitations don't exist on the desktop. The real limiting factors are the basic technologies upon which modern CPUs are built, and the common enemies aren't being based on the 8086, they're things like heat, power consumption and current leakage while attempting to maintain regular die shrinks. People really overblow this "X86 is the devil" nonsense-- it may not be the most elegant CPU technology. But they're cost-effective, so it's no wonder most of the processing power in the world is provided by x86 parts.

 

[nubie]

Alright, y'all can quit it with the "go buy an Eee" nonsense. Obviously, I'm here because I like something that the Pandora has to offer. If I'd wanted an Eee, I'd have one.

Given that much of the focus of the Pandora is directed at emulation (i.e.: cross platform compatibility) I'm a bit surprised at the dedication to a less prominent architecture.

Um, since when is ARM a "less prominent architecture" in the hand-held market? (I know it is "shaped" like a laptop, but it really isn't one), as far as I know ARM is in a lot of integrated applications. Gameboy advance anyone?

QUOTE

Anyway, to sum up, here are my options:


1. There's a computing emergency. This software needs to be run, and now, but there isn't a desktop or laptop in sight!

"I'll do it," I say confidently, pulling my Pandora out of my pocket, knowing that whatever it is, I can probably run it and save the day. And I can. And I do. And here come the ladies, oh so grateful.


2. There's a computing emergency, etc.

"Hey," I say. "Is the source available? Is this configured for Arm? We have time to get everything recompiled, right?" I think you can see where this is headed.

As it turns out, much as I may have talked it up, the ladies just weren't impressed by my slightly better performance per watt.


The future will bring these x86 "MIDs", although I am highly skeptical about them supporting XP or Win2k, or even win95/98 for games.

Why win 95/98/2k/XP? Because if you are running games you need to run the OS that they came with (I know about Wine, but it is stupid to use it instead of just having windows on the device, it is hundreds of times more compatible).

Will there be win95 drivers? will there be XP drivers? Will they support games?

If you are just asking if the Pandora crew will investigate an x86 platform designed for gaming, I can't speak for them, but I will say that if you expect to play games written around 1998, and the integrated "SoC" Atom is out in the next 4 years, you will still be paying a premium of $200-300 over a comparable ARM system, not to mention that for hand-held the Pandora will probably still be able to work for most of what you want to do.

 

[nubie]

X86 will eventually die and MS will either die (by putting all of their resources into that sinking ship that is Vista)

Quantum leaps rarely happen in this industry. Things slowly change and software evolves with it. It's also perfectly possible for one microarchitecture to interpret instructions meant for another, albeit more slowly and/or with a great portion of the die devoted to support hardware. The idea of Microsoft, the only company making a marketable desktop operating system for standard PCs, going under because their new OS is a bloated mess is outrageous.

In reality, a great deal of updates and service packs later, Vista will be (slowly) accepted. It'll be tough since it doesn't bring much useful features to the table compared to the 98SE transition to 2K/XP, but eventually it'll be the standard. XP was pretty bad at first too.

edit: x86 also doesn't create limitations that would stunt the further development of the microprocessor. The need for many layers of compatibility with older instruction sets means it's less ideal for a low-power simple design you'd expect in a mobile SOC, but these limitations don't exist on the desktop. The real limiting factors are the basic technologies upon which modern CPUs are built, and the common enemies aren't being based on the 8086, they're things like heat, power consumption and current leakage while attempting to maintain regular die shrinks. People really overblow this "X86 is the devil" nonsense-- it may not be the most elegant CPU technology. But they're cost-effective, so it's no wonder most of the processing power in the world is provided by x86 parts.


The trouble with Vista is that it is to XP what ME is to 98. There is no "expanded functionality" over XP, except that which is artificially introduced by "DX10", which has no technical or logical reason to require Vista. (note too that MS has flooded the market with "Windows Mobile" mp3 players, and that these refuse to function without Windows Media Player 11, and that will not install on W2K, how long until the basic mp3 players require Vista for Windows Media Player 15 and you can't listen to music on them? And it isn't Just the Zune, Creative and others are drinking the kool-aide)

 

[ldesnogu]

Um, since when is ARM a "less prominent architecture" in the hand-held market? (I know it is "shaped" like a laptop, but it really isn't one), as far as I know ARM is in a lot of integrated applications. Gameboy advance anyone?

The GBA is nothing: ARM has already sold 10 billion of processors, and currently the rate is around 2 billion a year.
In terms of units the x86 is nothing

 

[Firefox]

Here is a very instructive read about "new" features to come in next generations of Intel chips: http://www.theinquirer.net/gb/inquirer/new...vx-instructions

That will perhaps open the eyes of the younger who don't get how unification has slowed down the evolution of the computer industry.

Too bad we all know what happen to the Alpha and its original owner anyway, and that's not gonna change unless, say, Nvidia discovers that Alpha also had the world's best real-time X86 code translator for Windoze, the FX!32, making it the only remaining high-end candidate for the firm's CPU in the absence of AMD buyout. Is Nvidia willing to do a fabless EV9 feeder for their Geforces?

How cool would that be? Alpha is a beautiful architecture, I could really get behind an Alpha + Geforce system. B)

 

[ldesnogu]

How cool would that be? Alpha is a beautiful architecture, I could really get behind an Alpha + Geforce system. B)

Yay, it was really great
But now Alpha engineers have found jobs in most other CPU companies (AMD, Intel, Apple/PA semi, ...) so it would be difficult to restart that :lol:

 

[javaJake]

In reality, a great deal of updates and service packs later, Vista will be (slowly) accepted. It'll be tough since it doesn't bring much useful features to the table compared to the 98SE transition to 2K/XP, but eventually it'll be the standard. XP was pretty bad at first too.

Finally, some sense. I've heard nothing but pure Vista hatred as if Vista was nothing but junk (when in reality all people do is focus on the bad parts). Someone besides me has finally stepped up and taken a stand in the truth of the matter.

I can rest in peace at last.


Oh, and, uh, go ARM!

The trouble with Vista is that it is to XP what ME is to 98. There is no "expanded functionality" over XP, except that which is artificially introduced by "DX10", which has no technical or logical reason to require Vista.

*sigh*... I really want to avoid this debate, but I can't help but make a last comment.

We all know that the "there is no new functionality" comment is false. Security, for one. Sure, maybe a little inconvenient, but at least you can give Microsoft credit for working on it! And then there's the added visual experience and interface which I always find enjoyable and an upgrade from XP. Yes, it does require extra requirements, but I think/hope Microsoft will get smart and start streamlining it. Oh, and don't forget the reworked internals! Microsoft is striving for more stability and security with an upgraded structure. Driver incompatibilities are, as far as I can see, mostly because companies never paid any mind to Vista until their customers started complaining.

OK, I'm done.

 

[sindbad]

(when in reality all people do is focus on the bad parts).
People expect new things to be improved. There are things that became worse in this new thing and it's natural for them to be annoyed.

Sure, maybe a little inconvenient, but at least you can give Microsoft credit for working on it!
All those annoying dialogs that jump around when you try do something don't increase security one bit. After 10min I stopped reading them and I suspect this is the case with most users.
No one gets points for trying in the real world.

But this is offtopic.

 

[sephiroth111]

look at the X86 arch. they JUST removed the 16 bit subsystem with the switch from the 9x kernel to the NT kernel. for YEARS people have bitched about how they cant run their "good old dos software" but there really isn't any software that i can think of that is useful when it was written over 10 years ago for dos. (linux is excluded. While i think X.org should be rewritten from the ground up using completely 32/64bit code and take advantage of new graphical/ CPU developments because the X libs are a mess, that's beside the point.)

People don't want to abandon their x86 software, and x86-64 is really just a tack on of 64 bits in an environment that has "legacy mode" which is 8bit, 16 bit, and 32 bit. until someone decides that they need a new arch that removes all native support from the cpu itself (maybe emulated with a coproc) we cant advance any further. basically what i'm proposing is a removal of all "legacy modes" stripping the chip down to pure 64 bit mode, and refining the feature set, opcodes, and making it a bit more graceful.

and vista is more the windows ME in the NT line of kernels. wikipedia Windows Seven. vista is complete and total crap, and it's going to mostly stay that way.

Considering most companies are excercising "downgrade rights" until june of next year, everything's gonna be alright.

We all know that the "there is no new functionality" comment is false. Security, for one. Sure, maybe a little inconvenient, but at least you can give Microsoft credit for working on it! And then there's the added visual experience and interface which I always find enjoyable and an upgrade from XP. Yes, it does require extra requirements, but I think/hope Microsoft will get smart and start streamlining it. Oh, and don't forget the reworked internals! Microsoft is striving for more stability and security with an upgraded structure. Driver incompatibilities are, as far as I can see, mostly because companies never paid any mind to Vista until their customers started complaining.

OK, I'm done.

Were you paid to say that stuff? damn. Security wise? i remember there being something like 15k bugs in the initial release, most of which were fixed. UAE was executed in a pain in the ass method. the BIT LOCKER? a joke. Aero is nice, and i admit the start bar improvements were nice, HOWEVER it doesn't justify the ram jump from 512mb for a fully loaded XP system, to 768mb for a fresh install of vista. DX10? still a joke. they have to PAY people to use it. let me know if i forgot anything.

 

[Blah]

QUOTE(sephiroth111 (paraphrased for clarity) @ Apr 29 2008, 06:33 PM)

Linux is part 16bit and runs on DOS.



What the hell are you smokin', boy?

 

[ldesnogu]

(linux is excluded. While i think X.org should be rewritten from the ground up using completely 32/64bit code and take advantage of new graphical/ CPU developments because the X libs are a mess, that's beside the point.)

Where exactly is there non 32/64 bit clean code in X?
I used X on a 64 bit Alpha workstation 15 years ago and it worked fine.

X is perfectly adequate for 2d window management. Perhaps not for embedded devices