AMD Updates Alchemy PDA Prototype

[Janak Parekh]

<div class='os_post_top_link'><a href='http://story.news.yahoo.com/news?tmpl=story&cid=1093&ncid=1093&e=5&u=/pcworld/20030929/tc_pcworld/112668' target='_blank'>http://story.news.yahoo.com/news?tm..._pcworld/112668</a><br /><br /></div>AMD is still working at getting a crack in the PDA market, despite <a href="http://www.pocketpcthoughts.com/forums/viewtopic.php?t=16300">lukewarm reception</a> of their Alchemy processor. They've updated their PDA prototype to, among other things, do full-motion video with no artifacts.<br /><br />"Based on AMD's 400-MHz Alchemy 1100 processor, an early prototype of the PDA reference design was demonstrated in August by the company at the LinuxWorld Conference &amp; Expo in San Francisco. Since then, AMD has advanced the prototype's design with the addition of full-screen video capabilities and has completed work on the device's battery charger, said Phil Pompa, vice president of marketing for AMD's Personal Connectivity Solutions group, in an interview on the sidelines of the Computex exhibition in Taipei last week. 'It's certainly at the point where we can go hand to this to an OEM,' Pompa said, adding that AMD is currently working on improvements to the design's power management capabilities."<br /><br />What I find interesting is that of all the Linux PDAs coming on the market, only the Sharp units seem to have generated considerable interest. It'll be interesting to see how this market evolves in comparison to more-established handheld OSes, like Windows CE.

[Foo Fighter]

Mobile Linux OS on a MIPS processor? Two useless technologies rolled up into one. :roll:

[jage]

The reference design is able to run Windows CE in any case. Makes me just wonder if Microsoft's decision to drop MIPS was a bit too early one... AMD's offering could be pretty powerful CPU.

Alchemy does have some niceties, like that nice integer division instruction which helps sometimes quite a bit (data compression, texture mapping, etc.). AMD also mentions fast memory.

Might be interesting to see this thing running PPC2000 to compare.

[jage]

Mobile Linux OS on a MIPS processor? Two useless technologies rolled up into one. :roll:

What makes MIPS CPUs or mobile linux useless. Please do tell!

[Foo Fighter]

MIPS is dead. ARM is quickly becoming the standard. Linux on mobile devices has gone nowhere. Sharp has had very little success with its Zaurus line. Linux has a bright future ahead of it in servers, and perhaps the desktop, but mobile devices? Not going to happen anytime soon.

[David Prahl]

http://www.mobilemag.com/content/images/2089_large.jpg

:roll:

[Janak Parekh]

MIPS is dead. ARM is quickly becoming the standard. Linux on mobile devices has gone nowhere. Sharp has had very little success with its Zaurus line. Linux has a bright future ahead of it in servers, and perhaps the desktop, but mobile devices? Not going to happen anytime soon.
As a consumer platform, perhaps. However, ARM does have its limits too (no floating-point, among other things). While ARM has its advantages, we can't assume it will have permanent dominance in the consumer handheld industry.

--janak

[Foo Fighter]

While ARM has its advantages, we can't assume it will have permanent dominance

Never said it would have permanent dominance. But MIPS is not contender. I'd still like to see a COMPLETELY new mobile processor designed from the ground up for Mobile devices. We are getting closer to that with "System on a chip" technologies (integrated CPU, GPU + wireless).

[jage]

MIPS is dead. ARM is quickly becoming the standard. Linux on mobile devices has gone nowhere. Sharp has had very little success with its Zaurus line. Linux has a bright future ahead of it in servers, and perhaps the desktop, but mobile devices? Not going to happen anytime soon.

MIPS is dead? Good joke. Tons of devices from printers to games consoles have MIPS CPU inside. Think PS2. Think about your printer. Think about your digital camera. MIPS is a very sound microarchitecture and there's nothing dead in it. Some implementations might be bad, but the architecture itself is good. MIPS could make a comeback in the form of AMD Alchemy. Alchemy sounds like it could have potential to wipe floors performance-wise with ANY current or near future ARM-based processor, including Intel's X-Scale. There's absolutely no reason to dismiss MIPS yet.

I wouldn't be so quick to dismiss MIPS or Linux, regardless of the current market situation. Linux as a kernel is definitely way more advanced than Windows CE is and has also been demonstrated to run on embedded devices, why would PDAs be any different if someone puts enough bucks in the development?

[jage]

Never said it would have permanent dominance. But MIPS is not contender. I'd still like to see a COMPLETELY new mobile processor designed from the ground up for Mobile devices. We are getting closer to that with "System on a chip" technologies (integrated CPU, GPU + wireless).

Ah you mean like Samsung's ARM, X-Scale or Alchemy? :) I'd like to see GPUs integrated, but doesn't look like it's going to happen yet, maybe in 1-2 years. There's nothing in MIPS or ARM architecture that prevents you from integrating GPU and wireless in them. For example Intel's PXA-26x series is pretty integrated already.

If you want to see a completely new microarchitecture, I doubt it will happen.

[Foo Fighter]

I'm talking about handhelds, not embedded devices? How many developers write MIPS applications for printers? Sorry, but MIPS has no future in handheld computing.

[Foo Fighter]

Alchemy sounds like it could have potential to wipe floors performance-wise with ANY current or near future ARM-based processor,

Yes it could. Unfortunately that doesn't matter. As long as PalmOS and Pocket PC do not support this architecture it won't make any difference.

why would PDAs be any different if someone puts enough bucks in the development?

Because PalmOS and Pocket PC are the standard in handheld computing. Developers aren't writing for Linux (other than open-source advocates), and every iteration of Linux for mobile device has proven to be very poor quality. Look at QTopia, it's crap. Slow, buggy, and very poorly designed. Until we see major software firms back Linux with real support and innovation, it won't go anywhere.

[jage]

I'm talking about handhelds, not embedded devices? How many developers write MIPS applications for printers? Sorry, but MIPS has no future in handheld computing.

All it needs is high computational power (check), low power consumption (check) and low price (check). If the OS is same, 95% of the apps (not counting games) are a simple recompile, and the rest (ARM-specific games) require modifying maybe up to 5% of the code, probably less.

In fact PPC2000 was running MIPS -- there's no inherent reason new versions couldn't support MIPS again. Limiting ourselves to one architecture might not be a good idea.


Yes it could. Unfortunately that doesn't matter. As long as PalmOS and Pocket PC do not support this architecture it won't make any difference.


The point is there's no reason they couldn't support. The previous MIPS implementations weren't good enough -- looks like AMD's is. Nothing is written in the stone yet. Windows CE is portable and I bet also Palm learnt it's lesson and makes new PalmOS versions easily portable between microarchitectures. It's not like handheld market is like x86 where there's a _huge_ installed base and recompiling cannot be the answer.

I get your point that Symbian, Palm and Microsoft embracing ARM means BIG roadblocks for a MIPS design to gain foothold -- but to say that MIPS or mobile Linux is dead is just plain wrong and totally premature.

[Janak Parekh]

I get your point that Symbian, Palm and Microsoft embracing ARM means BIG roadblocks for a MIPS design to gain foothold
I think that's ultimately Foo's point. For the foreseeable future, ARM has clear sailing -- both PalmSource and Microsoft have endorsed it, and I do agree it will stay for at least one or two more generations without any reevaluation of architecture, so other architectures are "dead" for the time being. I'm glad to see someone is keeping up with embedded MIPS solutions, though, and who knows what'll happen in the future... :)

Re multiplatform: MS dropped it because of the support headaches back in the PPC2k days. There were a lot of developers who had architecture-specific bugs, or didn't make versions of their code available for other platforms for various reasons. Standardizing a broad consumer platform on one architecture has its advantages, which is why Microsoft dropped support for other platforms on PPC2k2. Of course, the underlying Windows CE codebase is totally portable, and will remain so, due to the need for embedded solutions -- so if for some reason Microsoft wants to switch back in the future, it can -- but that isn't happening anytime soon for Pocket PC.

--janak