AMD's New Low-Power CPUs

illumina+us writes "AMD has released a new family of CPUs targeted at the portable computing market. The new CPUs, collectively named Alchemy, consume less than 1Watt of power. The CPUs have already been named the CPU of choice for Tivo's new Tivo-To-Go technology and are powerful enugh to run DivX, WMV9, and MPEG. The AU1550 consumes just 0.5 Watts at 400 MHz and the AU1100 consumes 0.25 at the same clock speed. These processors consume so little energy they don't even need a heatsink."

Comments

imagine...

[Anonymous Coward]

A Beowulf cluster of these...

Hey, at least the power bill wouldn't kill you.

Re:imagine...

[Anonymous Coward]

And yet, you're still here. Amazing!

Re:imagine...

[mattyrobinson69]

you stopped reading slashdot eh? When you returned, were you shunned and labeled anonymous coward?

Re:imagine...

[Anonymous Coward]

Already done: http://www.orionmulti.com/ 12-node cluster built on Transmeta chips, feeds on only 200 Watts altogether. A little larger than a desktop case in size.

Re:imagine...

[cheezfreek]

Stopped reading /. except for what you read in order to post the above comment, eh?

"She turned me into a newt...I got better."

Re:imagine...

[Max Romantschuk]

Actually low power chips are gaining popularity in super computer environments. A major cost of operating a huge cluster is power and getting rid of all the heat.

Low power chips are therefore much cheaper to operate, and can be packed more densely as they require less cooling. The future of computing lies in massively distributed low power solutions, it simply makes much more sense than the alternatives.

Re:imagine...

[Anonymous Coward]

tee hee...fag

Re:imagine...

[3Suns]

Actually, one of my big disappointments with Transmeta chips is that they don't support SMP at all. Clustering is one thing, but SMP is better for many applications. Imagine a laptop with 4 CPUs, or a 1U blade with 8? An 8-way desktop machine would rule, especially in terms of responsiveness; MP3 player, web browser, window manager, filesharing, and word processor all could run on separate CPUs without ever having to interrupt each other.

I wonder if anyone will produce MP boards for these new AMD chips?

Re:imagine...

[Anonymous Coward]

Actually low power chips are gaining popularity in super computer environments. A major cost of operating a huge cluster is power and getting rid of all the heat.

Speaking as a nonexpert, I'd think it depends somewhat on the kind of computing you're doing.

For my work (analysis of physics events) the work is neatly divided into independent chunks, and so low power CPUs and a simple cooling solution (e.g., low-conductivity water cooled racks), are surely the way to go.

But for numerical integration of a big hairy equation, it may better to have (clusters of) more powerful processors in closer communication, even at the cost of a higher electric bill and more complicated power + cooling system.

Re:imagine...

[eh2o]

It seems to me that its about high efficiency, not low power. Merely having a low power chip does not help a supercomputer if you need that many more of them to get the same performance.

Re:imagine...

[IHateSlashDot]

The future of computing lies in massively distributed low power solutions

Thanks Nostradamus. I think I'll file this opinion in the same file as the classic "Computers will never need more thean 64kb of memory".

Re:imagine...

[smittyoneeach]

U R teh st00p3d. Everyone knows that the real upper bound is 640k. Running Emacs, obviously.

PDA's

[SlongNY]

Wonder if these will pop up on PDA's and stuff soon..

Re:PDA's

[cnettel]

OTOH, the current crop of Pocket PCs are able to decode DivX in 640x480 with ARM-based chips from Intel, even without video acceleration. It's not like this kind of performance is a revolutionary breakthrough.

Re:PDA's

[Oculus Habent]

The 1100 is aimed at that market. It seems most of the line is special use - routers & various appliances.

The 1100 Development Board looks like fun, though. Is it bad when you have the urge to say, "Ooohh, preeeeetty" even if you have no real use (or skill to work with) it?

Re:PDA's

[tmasssey]

Given this formfactor, I'm really suprised they didn't create a Mini-ITX format board. Then they'd have a whole slew of ready-made periphials: cases, power supplies, etc.

Re:PDA's

[yope]

I doubt they'll become very popular.
They Alchemy family is not that new actually. AFAIR they have been around for at least 2 or 3 years now, and have barely gotten some attention from the embedded developer world.

AMD has made inroads in the embedded processor bussiness before, with their Elan and embedded-K6 processors. Those have been moderately popular by those seeking x86 compatibility, since the Elan is a mocked-up 486 with chipset functionality and some periferals in one chip: Expensive, extremely power-hungry, slow and very modest on-chip periferals, but x86 compatible. They are mostly forgotten now.

The Alchemy on the other hand is based on a 32-bit MIPS core (remeber SGI? Guess where their chip developers went?). That makes the Alchemy more powerful, less power-hungry, cheaper and able to include some more amount of periferals on-chip, but they are not x86 compatible.

That leaves them pretty much out in the cold, because there are IMHO far more attractive alternatives of non-x86 embedded processors, like those based on the ARM family of cores, built by Samsung, Atmel, Philips, TI, Cirrus-Logic, Intel and many more, as well as the PowerPC based embedded processors from Motorola and IBM. Specially the Power-QUICC I and II families from Motorola cover an impressive price and performance range, offer modest to very high processing power, and unprecedented flexibility due to their second integrated RISC based communications processor and programmable bus controller.
Those are the two most popular embedded processor platforms around these days. If you need power-efficiency, there's no better than ARM. If you need high computing performance or high-bandwith data processing, go for PowerPC. AMD's Alchemy is somewhere in the middle, but until now they only cover a narrow range of applications.

Re:PDA's

[lion2]

That's the first thing I thought. However I dont think any Palm or PocketPC PDA is going to use it, because they all use ARM based processors and this AMD chip is a MIPS32 processor. Microsoft no longer makes the PocketPC OS for MIPS so Alchemy is out of the question.

Im surprised that AMD didnt make this CPU an ARM processor. It would be nice if Intel had more competition in the PDA market.

Re:PDA's

[LuSiDe]

I was thinking: Perhaps its good for competition (and hence price)?

Last time i checked, the Sharp Zaurus was pretty expensive, especially the newer models. What was it again, like $700 for the newer model in Japan. I know PDAs sell like baked bread in Japan but still, why are these toys so expensive? A friend of mine has a CL 5500 and while its a very nice PDA it was also pretty expensive when he bought it. And, newer versions are more convenient.

I hope to see more competition / prices going more down since i'd love to see the GSM (as in, a telephone) integrated with the PDA (as in, basic simple office work, e-mail synced, WiFi/UMTS) with a multimedia player (preferably also Vorbis-compatible). I want 1 device which is small enough but also gives me the features i want to have without using a beast such as a laptop and without having 1 device like e.g. the GSM which can phone (what i actually want), SMS (expensive!), play retro games (boring!) without the features i also would like to have.

Re:PDA's

[tonywong]

Actually this has been in the works for a while... AMD did not originally design this but the design team was bought lock stock and barrell from a fabless startup called alchemy semiconductor. Just off the top of my head I recall that these were the original strongarm design team guys that took off when digital got bought out or something like that.

check out the archive.org links for more about them...
http://web.archive.org/web/*/http://alche mysemi.co m

too lazy to use html, or even caps, but oh well cut and paste...

Re:PDA's

[gadget junkie]

" i'd love to see the GSM (as in, a telephone) integrated with the PDA (as in, basic simple office work, e-mail synced, WiFi/UMTS) with a multimedia player (preferably also Vorbis-compatible). I want 1 device which is small enough but also gives me the features i want to have without using a beast such as a laptop and without having 1 device like e.g. the GSM which can phone (what i actually want), SMS (expensive!), play retro games (boring!) without the features i also would like to have."

I think that, while what you say is valid, form factor is a big issue here. I just went for a motorola mpx 200 smartphone, just because I wanted a compact pda/phone file repository without photocamera.
It does have media player installed, but the screen real estate is so small that I do not see myself using its multimedia features anytime soon.

AFAIK, the consumer market is going after the blackberry/ Ipaq/ Ipod form factor, which is in my view rather unwieldy as a phone. It all depends on prevailing use, i.e. mail and multimedia or voice.

Re:PDA's

[ssimontis]

It could be nice for a laptop too. It would meet my needs, and I could sit in the living room and watch TV and use my computer at the same time!

Re:PDA's

[iamhassi]

"It's not like this kind of performance is a revolutionary breakthrough."

No, it's not, but the fact that AMD is creeping into a market that Intel currently dominates, and AMD has already declared dominance in the gaming and server microprocessor market in 2004, so this could cause serious problems for Intel if the AMD chips turn out to perform better with less power than Intel's current offerings. Sure the processors are running at slower mhz speeds but as we all know a a slower mhz AMD processor can perform at the same level as a much faster mhz Intel processor

Re:PDA's

[Anonymous Coward]

Yes, that would be excellent. Especially since we wouldn't need to use VIA's ass chipsets and CPU's.

Re:PDA's

[smartdreamer]

with ARM-based chips from Intel

Correct me if I'm wrong, but isn't ARM Intel's competitor? It is a completely different architechture licensed to many companies. Also interesting is that they are RISC processors (that's where the R in ARM comes from).

Re:PDA's

[ckaminski]

You can have it today:

Kyocera 7135 Color Palm.

http://www.kyocerawireless.com/

Re:PDA's

[Anonymous Coward]

Actually, Alchemy had nothing to do with SGI apart from making a MIPS32-implementing processor. Alchemy was founded in 1999 mostly by DIGITAL engineers (the ones who had developed the StrongARM chip, yes the one that powered the infamous Apple Newton and went up to 206 MHz when other embedded designs were just approaching 30 MHz...), after the latter was sold to Intel and reworked to become their XScale line. Apparently, the DIGITAL guys didn't like the atmosphere at Intel and started their own firm, which was acquired by AMD in 2002.

I guess, Intel learned their lesson and made sure they brown-nosed the engineers on the Alpha project sufficiently to make them not go away - although from what we hear lately, maybe Intel is screwing that one up, too (for the benefit of AMD, who's Dirk Meyer I'm sure welcomes the Intel defects in person ;-) )

Re:PDA's

[drinkypoo]

I know they're in third place, but there are an awful lot of MIPS cores out there. For that matter, the Playstation uses a pretty standard R3000 processor and the PS2's Emotion Engine is also built from MIPS cores.

Re:PDA's

[ShieldW0lf]

Sure the processors are running at slower mhz speeds but as we all know a a slower mhz AMD processor can perform at the same level as a much faster mhz Intel processor

That comparison only really holds water when you're comparing AMDs chips with the P4 architecture. Intels mobile market focus is evolving the P3 architecture, P4 is for desktops. The P3 architecture gets a lot more done per cycle than the P4, and is pretty comperable if you look at the Athlons vs Coppermine/Tualatin

Wonder how fast they'd be if Intel hadn't dropped them for the P4s... that was such a dumb move.

Re:PDA's

[WhyCause]

Correct me if I'm wrong, but isn't ARM Intel's competitor?

Sort of. ARM designs and licenses processor architectures to manufacturers, specifically, the processor cores, around which a manufacturer adds the cache, etc., needed for the specific application. From the "ARM Corporate Background" PDF on their website:

When ARM pioneered the concept of openly-licensable IP for the development of 32-bit RISC microprocessor-based SoCs in the early 1990s, it changed the dynamics of the semiconductor industry forever. By licensing, rather than manufacturing and selling its chip technology, the Company established a new business model that has redefined the way microprocessors are designed, produced and sold.

I believe that the Intel PXA-series processors found in many Pocket PCs are derived from the StrongARM design that DEC (I believe) co-developed with ARM. Then, after a few subsequent buyouts (DEC -> Compaq -> Intel), Intel ended up with a very successful architecture that they did not (initially) develop.

So, while they compete at the architecture level, Intel is more interested (perhaps to their detriment) in selling a lot of silicon, while ARM sells ideas. ARM makes money by thinking in its ivory tower, Intel (and Samsung and Toshiba, and IBM) makes money by bringing that thinking down to the filthy masses.

Re:PDA's

[yope]

You are right. NEC also makes embedded processors based on the MIPS core, but I think before the Alchemy's there is still a third place for Motorola Coldfire in popularity. A time back I saw a chart comparing market-share of all these embedded processors, but I can't remember where.

One thing I forgot to mention is the new series of Goede processors from AMD which are the result of a buyout from National Semiconductor. Those are quite powerful x86 processors partly aimed at the embedded market. They are also very power-efficient for a x86 core. The problem is that AMD has yet to demostrate to the industry that they are taking their x86 embedded business serious again, after abandoning the Elan series.

Anyway, don't get me wrong, I still think AMD is a cool company, making cool chips, not only for the PC market.

Re:PDA's

[yope]

Yes, they do. AFAIK a few chip developers from SGI went to AMD after SGI's chip division went belly-up. As a result they helped creating the Alchemy series of embedded processors. I don't know if some of the ex-DEC engineers also had something to do with it or not.

What everybody does know, I believe, is that Digital developed the DEC Alpha processor, the most powerful processor at that time (64-bit). Also some Digital engineers who worked on the Alpha joined AMD after DEC went boom. The result of this is the AMD Athlon and specially it's outstanding local bus system.

Re:PDA's

[Anonymous Coward]

The Alchemy on the other hand is based on a 32-bit MIPS core (remeber SGI? Guess where their chip developers went?). That makes the Alchemy more powerful, less power-hungry, cheaper and able to include some more amount of periferals on-chip, but they are not x86 compatible.
Actually, a lot of the Alchemy team came from DEC's StrongARM group. The group left DEC rather than join Intel, and moved to Cadence for a while. They then spun off into a startup called Alchemy, which AMD acquired a few years ago.

Obviously not for Canadians

[Anonymous Coward]

These new chips are obviously not aimed at the Canadian market, or any market that has severe winters. We use our computers to heat buildings, fry eggs and cook bacon.

And when there's no electricity, we burn them for heat.

Re:Obviously not for Canadians

[ackthpt]

These new chips are obviously not aimed at the Canadian market, or any market that has severe winters. We use our computers to heat buildings, fry eggs and cook bacon. And when there's no electricity, we burn them for heat.

Oh... I'm sure you can find a way to overclock them, and carry a car battery around for power, that should keep you nice and warm and fit in the GWN, eh!

Re:Obviously not for Canadians

[menkhaura]

Car battery! Damn funny, thanks for the laugh!

Re:Obviously not for Canadians

[hlygrail]

Uhh... don't you mean back-bacon?? :)

Re:Obviously not for Canadians

[hlygrail]

Uhh... don't you mean back-bacon, eh?? :)

Re:Obviously not for Canadians

[DrEldarion]

I live near Chicago, and my room used to be the coldest in the house. During the winter, frost would form on the walls.

Then I got an Athlon-650 computer in there and a 19" CRT monitor, and from then on, it was the warmest room in the house, even in the winter.

Re:Obviously not for Canadians

[snorklewacker]

Like, uhh, don't you mean back-bacon, eh hoser?

Re:Obviously not for Canadians

[hlygrail]

Take off, eh !!

Re:Obviously not for Canadians

[__aamkky7574]

Even in Ireland (far more temperate than Canada) the bedroom is a lot more toasty in the morning if I've left the PC on downloading movies^H^H^H^H^H^H^Hperforming important calculations all night. I'd love to know how it works out cost-wise using a PC to heat a room than an actual heater.

P.

Re:Obviously not for Canadians

[hfolkers]

In winter times my amd-650 runs stable at 750mhz with normal cooling, in summer times it isn't possible with the same cooling.

Re:Obviously not for Canadians

[Anonymous Coward]

Well, yeah, cause you started spending all your time in it playing Everquest!

Re:Obviously not for Canadians

[Chicks_Hate_Me]

I did a comparison of my electricity bill and leaving my old Athlon 1GHz on all the time increased the bill only $1.50 more. Not too shabby.

Re:Obviously not for Canadians

[mrbcs]

I had about 4 computers running in my basement. The best one (for heat) was a p-pro quad 200 with 12 9 gig scsi drives and 3 power supplies ;-)

After I took it out, I actually needed my gas fireplace. It also seems that power is cheaper than gas. I saved 25 bucks a month on my power bill... but my last gas bill was really high.. though it was very cold last month.

Re:Obviously not for Canadians

[InvalidError]

If your primary heat is electric, the only cost difference is in computer repairs since computers are far more likely to fail than an electric radiator.

During summer though, I keep only the PCs I need turned on and I take a break from SETI&all.

Re:Obviously not for Canadians

[Anonymous Coward]

Actually, there are a lot more differences than that. For one thing, any electric heating system is going to have a thermostat, so it's not going to be running 24/7. Electric heaters also tend to reach much higher temperatures more quickly, in order to make sure they heat up the room as fast and efficiently as possible. Unfortunately, while the heater is going, it's doing absolutely zilch productive work with all those electrons buzzing around, basically using them for their resistance--but no resistive element can be 100% effective and still carry a current. That's one of the reasons why electric heating turns out to be so much worse than gas. Your computer, on the other hand, is using the current for its own purposes, and producing heat as an unwanted byproduct. Thus, as long as you're using your computer for something anyway, it's probably going to be a win to leave it running all the time. Also, it's probably more efficient to have a low level heating element (your CPU + GPU, bwahaha) running all the time than having a high level heating element turn on and off all the time. I know modern gas heaters are supposed to prefer the former method.

Re:Obviously not for Canadians

[honestbob]

Well, excepting the cost of buying and maintaining the hardware, the pc should consume just as much power as an electric heater that consumes the same power. This is a thermodynamic inevitability. I have always understood that electric heat is something like 2-4 times as expensive as gas heat. This makes sense, since oil-fired coal plants operate somewhere in the 40-50% efficiency range (whereas, if you burned it in your heater you'd only give up the small amount of heat in your exhaust gas). Admittedly, natural gas is more expensive than the oil that power plants burn, but you could probably offset that with the losses due to the transmission on electrical power lines.

Re:Obviously not for Canadians

[Anonymous Coward]

Try living in south florida... 2 LCDs and a Athlon 2800

AC keeps the house at 75 degrees or so i think

my den is ALWAYS at least 82.

so, yeah. pumpin out BTUs :)

Re:Obviously not for Canadians

[damiam]

the pc should consume just as much power as an electric heater that consumes the same power.

Wow. Who woulda thought?

Re:Obviously not for Canadians

[julie-h]

And some say p2p is not good for anything.

Re:Obviously not for Canadians

[jsight]

All of this assumes that your electric heater being used for comparison is using heat strips.

Heat pumps are usually far more efficient (depending upon ambient temperature).

Re:Obviously not for Canadians

[Anonymous Coward]

I must be cursed.. I live in Edinburgh, Scotland, and my shitty council flat is utterly cold (less than 10c), all the damn time, yet the interior temperature of my athlon 1800 sits around 67c, with the case temp around 40, despite all the fans I put in to spread this toasty goodness around..

Maybe I should ditch the case and just screw the motherboard onto the wall...

Re:Obviously not for Canadians

[InvalidError]

Heat pumps for heating make sense mostly for central HVAC setups. My second sentence clearly mentionned "electric radiator" which clearly indicates what sort of electric heat I was implying in the first one.

Sure, heat pumps are ~10X more efficient than radiators and can usually do double-duty as air-conditionners.

One problem with heat pumps in Canada and other places higher up North is that their efficiency drops significantly below -20C and you would be fundamentally screwed if it is the only heat source available. For apartments where central heating is not possible and only one type of heating is the practical limit, the choices are either electric radiators or some variation of closed-circuit water (or sometimes even steam) heating.

Re:Obviously not for Canadians

[daft_one]

Regarding your last gas bill being really high... I (and my local friends) had an abnormally large bill last month as well, not entirely due to actual usage. It seems Montana Power was allowed to substantially increase "carrier fees" and such for the month, in order to pay their taxes. Inspect your bill for such mentions in the subatomic script at the end, might clear things up ;-)

Re:Obviously not for Canadians

[drinkypoo]

My retail box XP 2500+ 333 bus overheats in spring weather with the included cooler. I had to get a big thermaltake copper cooler to keep my system from overheating.

Re:Obviously not for Canadians

[InvalidError]

Under limit conditions, one gets:
Computer: ~200 Watts 24/24.
Heater: 2kWh ~6 minutes per hour.
(The above limit condition presumes the environment/room is leaking just enough heat to maintain steady temperature with the PC as the only heat source - human adults add about 80-120W apiece while idle, are we made on Intel's 65nm process?)

In either case, 17.28 megajoules worth of energy got injected in your work environment, the main difference being that the computer produced other useful work (and more noise) along the way.

Re:Obviously not for Canadians

[http101]

I live in Hell, er, I mean, Texas... down here, temperatures reach upward of 110ºF and computing seems to take place more at night when the temperature dips to 86ºF at 1000% humidity.

During the day, my Athlon XP 3000 (400Mhz FSB) with my 1.0Gb of Corsair PC3200 DDR (Dual Channel), sATA RAID, 550W High-Efficiency Power Supply, and 21" NEC Multisync XP21 monitor heat the upstairs computer lab/room/den to a toasty 92ºF with the A/C on!

It's no wonder I sit around in my boxers all day on Saturdays. I'm thinking about installing a rotisserie casemod with WinBraise v1.1 so I can monitor by food without having to open up the broiler/computer every 30 minutes.

Not x86 processors

[ackthpt]

These are based upon the MIPS32 RISC processor. Remember the R4000? These are 32 bit, how about that MIPS64 ;-)

upgrade from your SGI workstation to a tablet today!

Re:Not x86 processors

[OverlordQ]

If you want x86 processors AMD already realesed their Geode line for low-power mobile solutions.

Re:Not x86 processors

[doofusclam]

If you want x86 processors AMD already realesed their Geode [amd.com] line for low-power mobile solutions.

Have you ever built embedded or other low-power devices with a Geode? They're horrible. The problems are down to be b0rked DMA implementation which locks up at random, interrupts don't happen when you expect them to or in a decent timeframe, a crappy framebuffer which compresses the screen (with associated time penalty) whenever it changes, I could go on.

Oh, and I know this is specific to the board I used, but the 'hardware watchdog' was actually software emulation (much like the Geode sound 'hardware') which is no good when the board hangs and therefore so does the watchdog.

Nasty pieces of kit. Don't bother.

Re:Not x86 processors

[orionpi]

These have been avaliable for a few years now, I found them a while back when looking for drop in upgrades for an SGI Indy. The idea of pocket IRIX is sort neat, would just need a 1280x1024 touchscreen on your PDA.

oh no.....

[kidoman]

there goes my idea of a portable egg frying machine.

One catch

[Quasar1999]

I bet they aren't x86 compatable... cuz if they were... HOLY CRAP!

Re:One catch

[dead sun]

Who cares? Linux runs on MIPS, and if I remember correctly, so does WinCE. If you want small fanless x86 try the lower speed VIA C3/Eden offerings.

Re:One catch

[Zutfen]

Realistically, it doesn't matter much that they aren't x86. To me at least. I'm sure Linux will be running on it in no time.

Besides, they're aiming for the PDA market, which doesn't have x86 compatibilty as it stands anyhow.

I wonder if this technology will be adapted to the PC market in any way, shape or form. With such low power consumption, they are a fanless CPU, and a fanless power supply would probably be feasible I imagine. True silent computing sounds good to me... or is that doesn't sound...?

In any case, it is very cool tech, literally, and figuratively speaking.

Re:One catch

[Anonymous Coward]

You're thinking about the Geode processors. Those are < 1W, around 500Mhz, and x86 compatible.

Re:One catch

[Anonymous Coward]

Fanless PSUs exist now.

As far as silent computing, the gumdrop iMac came very close, but you could always still hear the hard drive. It was very quiet though.

Re:One catch

[truesaer]

They aren't. However, AMD does have a line of x86 embedded processors known as Geode. For example, I see here a 400Mhz Geode consumes 1.1 watts of power. This is part of their more general x86 everywhere plan...with x86 chips as cheap as $1.

Re:One catch

[Anonymous Coward]

the geode is one craptastic piece of glass...

Re:One catch

[Anonymous Coward]

I've already got silent computing: Mac mini. This machine is bliss after switching of my hoover-PC. :-)

Re:One catch

[bill_mcgonigle]

Who cares? Linux runs on MIPS

And it runs well. I stopped by the Gentoo booth at LinuxWorld (super great bunch of guys) and they had it running on damn near everything and not just plodding along like you might have seen MkLinux do.

I hear a persistent meme that Linux is inexorably optimized for little endian but I can't a reference to that actually being true.

Re:One catch

[Wraithlyn]

I want solid state computing... zero moving parts. No fans, and a flash-based "hard drive". I don't care about noise, I care about having something that sucks very little juice and will essentially never break down.

I'm happy right now with my GamePark32... open ARM architecture, variable speed up to 166MHz. Great homebrew community. It can play 320x176 24fps DivX/XVid (if you know your way around encoding), has emulators for everything up to an SNES, and even has a port of Linux. (You can get a keyboard for it) Runs on SmartMedia cards and 2 AA's.

Give me one of these new AMD units, and a device that runs off CompactFlash (SmartMedia is limited to 128MB) and I'd be set.

Re:One catch

[iminplaya]

With such low power consumption, they are a fanless CPU, and a fanless power supply would probably be feasible I imagine. True silent computing sounds good to me... or is that doesn't sound...?

Will this do?(check it fast. The link won't last) The price is right. If you want quiet, don't try to print.

Bah

[Anonymous Coward]

I like high powered processors.

Re:Bah

[oKtosiTe]

Like this one?

Obligatory Simpsons quote

[illumin8]

I, for one, welcome our new low-power overlords...

Re:Obligatory Simpsons quote

[Anonymous Coward]

Ya know what. It's not obligatory, it's stupid now.

It was funny for the first 500 stories someone did it on, now it's just dumb. As is saying something is obligatory, give it up people... it's not funny anymore.

Re:Obligatory Simpsons quote

[RoceKiller]

It's only obligatory when some karmawhore says it's obligatory. Stop it!

Re:Obligatory Simpsons quote

[Anonymous Coward]

In Soviet Union, Simpsons quote YOU!

FLOPS per Watt?

[DumbSwede]

It'd be nice to see how these stack up in FLOPS per Watt.

Perhaps these are the chips Supercomputer manufactures should be building machines with. Sounds to be low in cost to build AND low in cost to run.

Re:FLOPS per Watt?

[Anonymous Coward]

I was just thinking that you could more than make up for the slower speed on them by going MPC.

They don't currently use the x86 instruction set, but slap 24 of them into the same box and with an okay-ish emulation layer it could probably do a pretty good job at pretending to be either a single x86 or a PowerPC chip while consuming very little power.

Re:FLOPS per Watt?

[Anonymous Coward]

FLOPS per Watt? That's a bit personal.

Re:FLOPS per Watt?

[Ewan]

bluegene is based on that concept, 700mhz per cpu, 12 watts, i believe it achieves something like 6watts per gflop.

Re:FLOPS per Watt?

[ivan256]

Probably zero.

There's no evidence there that the MIPS32 core they used implements the (optional) floating point instructions. Of course you have to sign up for details so I can't say for sure...

Since the video capabilities are handled by an accessory processing unit, and since they were trying to cut power consumption, I'd be surprised if there was an FPU in the general purpose core.

Re:FLOPS per Watt?

[Duncan3]

Oh, then you want the Clearspeed chip, it's 25 GFLOPS @ 5 Watts, so 5 GFLOPS/W.

Re:FLOPS per Watt?

[lakeland]

For comparison, my favourite chip for this sort of thing (Motorola's G4) uses ~21W and gets 376 MFLOPS/W. That gives roughly twice as much power consumed as your cpu, for four times as much work. Depends on how much processing power you need I guess.

Re:FLOPS per Watt?

[abigor]

You're entirely correct. There's no FPU in these chips, right up to and including the au1550. If you want floating point, you have to do it in software. It's gross and slow, by the way.

Re:FLOPS per Watt?

[ratboy666]

Not so good -- most of these are SOC, and contain LCD controllers, bus controllers, net controllers, etc. on the same die.

You want MIPS core on a chip, without the extras. I don't think that this is (commonly) available; the MIPS core is generally dropped into other designs.

Ratboy.

Re:FLOPS per Watt?

[Carewolf]

You have no real use for floatint point in this type of applications anyway. The only place FP is used here is where programmers are too lazy or stupid to used fixed point.

Not powerful enugh

[Anonymous Coward]

Not powerful enugh for a spell checker though.

How about laptops

[I_am_Rambi]

with processors that create that little heat. That would be nice if they could clock them faster also.

Re:How about laptops

[advocate_one]

why make them faster??? why not just put more processors on the same board...

Re:How about laptops

[Inkieminstrel]

Yeah, should just be a matter of running more power to them, right? Oh...

Re:How about laptops

[Entropius]

The processors in modern laptops are responsible for only a small part of the power consumption during operations that are not CPU-intensive (i.e. almost everything). Most of the power draw comes from all the other gadgets that go into a computer, especially the screen.

When under heavy load (running seti@home, for instance), laptop processors do suck up a lot of juice... but for 90% of tasks, the processor consumes very little power.

This is true for the most recent offerings of both AMD and Intel.

How fast are they really?

[jandrese]

The writeup says you can run DivX, etc... but they don't say at what resolution or framerate. I've got PXA-255 based PDAs that can run DivX/WMV...as long as it is no larger than a postage stamp and encoded at more than 15 fps. The processor is still dog slow at stuff like compiling though. The writeup nor the articles give a good impression of exactly how fast these guys are, and that's a little worrysome. I don't mind energy efficent processors, but the last thing I want is something underpowered in my media center (oh, it can't handle 640x480 DivX, yay!).

Re:How fast are they really?

[Anonymous Coward]

Big fat RTFA for you!

Re:How fast are they really?

[cavemanf16]

Who cares about compile times? Sounds like you don't even need drivers written to support streaming media because that's already built-in. In other words, all you would have to worry about is assembling the parts, getting some graphics guys to build the GUI for your PVR, and then you throw just enough assembly code to run the GUI for your PVR, filesystem stuff, and a few other low level memory and subsystem drivers (like TCP/IP) and you'd be all set. And 400MHz can EASILY handle assembly coded drivers and apps that provide a wonderful multimedia experience to Joe Sixpack.

Re:How fast are they really?

[GuyRiley]

On the subject of PDA's, I have an iPAQ 2210 (PXA255/400MHz), and it can run XviD videos of varying bitrates at a full 320x240 @ 29.97fps. I imagine the same would be true of DivX. But for some reason, my PDA will not get more than 15fps out of a WMV encoded at any bitrate or resolution. I think that's more a result of shoddy coding than processing power though.

Re:How fast are they really?

[mobby_6kl]

There must be something wrong with your device. I have an older PXA-250 PocketPC, and it plays 320x240 divx video fullscreen just fine. I just ran a benchmark and without the speed limiter it ran at 141% speed, 35.2 fps.

Re:How fast are they really?

[Anonymous Coward]

From my experirence, Au1500 (400 MHz) it is probably 10 times slowed than a 1GHz Athlon.

The gotcha' is that it does not have a FPU and all floating point operations are emulated in an interrupt handler. So any audio or video encoding floats will take forever (way longer than real time). Porting it to fixed point will make things moving.

BTW, Monta Vista has a decent Linux 2.4 port for this setup: http://www.mvista.com/products/boards.html#mips. And no, I do not work them and I don't sell any of the above products.

Re:How fast are they really?

[Anonymous Coward]

Please tell me you're not running Gentoo on that!

how long....

[no reason to be here]

until we can get that kind of low power consumption on desktop chips? is there something inherent in desktop applications that prevent some chip maker from making a really low-power, high-performance (~1GHz) processor?

Re:how long....

[oringo]

Don't overlook the potential of these 400MHz chips. Mind you that the fastest computer today (IBM Blue/Gene) is made from 30,000 700Mhz cores. Of course they got heck lot of I/O capabilities.

Re:how long....

[Ironsides]

Please look at the Crusoe Processor They consume under 2 watts and I believe they have broken the Ghz barrier.

Re:how long....

[Anonymous Coward]

Market factors dictate against low-power cpu's.

The male-dominated market likes its computers to be just like its cars and women - hot and fast.

Irrational but true.

Re:how long....

[Junks Jerzey]

until we can get that kind of low power consumption on desktop chips? is there something inherent in desktop applications that prevent some chip maker from making a really low-power, high-performance (~1GHz) processor?

Super low power consumption and ultra high speed are inherently at odds with each other. It's like the memory/speed tradeoff that programmers have to deal with. (Crusoe is up around 1GHz, but they're already at twice the wattage of these chips.)

Remember, all CPUs had this kind of power consumption back in the day. You never saw heatsinks on CPUs until the mid 1990s. And processors in 8-bit home computers used milliwatts of power.

Re:how long....

[mjm1231]

You mean like these?

Re:how long....

[SlayerofGods]

Because it's a desktop which means there will be a nice little hole in the wall nearby that can provide you with all the power you need.
It's easier to make a fast hot processor and just keep it cool with heatsinks/fans then to create a fast processor that runs cool.

Re:how long....

[corngrower]

I'm not sure you'll be able to get down to 1 watt on a good desktop type cpu, but I'ld bet that you could get down to two or three watts.

From what I understand, the leakage current of the 90 nm chips is noticably higer than that of 130 nm chips, so first you'ld probably start by designing your chip to use 130 nm technology. (Maybe you could used something a bit smaller, depends on how much leakage curren you want to tolerate).

Secondly you'ld want to use an instruction set that's more power efficient to decode than that of x86 architecture, meaning it will probably be an ARM based architecture as it is quite efficient this way.

Thirdly you'ld probably get elimanate things like superscaler architecture and much branch prediction logic. Branch predicition logic takes quite a bit of room and would likely use a fair amount of power, however you're going to loose some performance.

Generally you'ld just keep subtracting things off until you got the power down to where you want it.

Take a looksie at Sharp's ARM based chips. These may be close to what you're looking for. They're under 1 watt, but are not quite a general purpose CPU.

Re:how long....

[corngrower]

It's like the memory/speed tradeoff that programmers have to deal with

Memory - speed trade off? Hockey pucks. Because cache memory is so much faster than going to ram, programs these days are optimized for speed BY optimizing for size!

Re:how long....

[LWATCDR]

The X86 ISA is another issues for low power on the desktop. That and the clock speed wars. If you want it to be PC compatable and fast you are going to get hot. People seem to be happy to have loud hot computers on there desktop.

Think of it this way. This chip uses what 1 watt of power. A P4 uses 120+ I think.

So you could have 100+ of these chips for the same power as one P4. Think of it as a cluster on your desktop. It would be kind of cool.

Re:how long....

[Junks Jerzey]

Memory - speed trade off? Hockey pucks. Because cache memory is so much faster than going to ram, programs these days are optimized for speed BY optimizing for size!

For some small kinds of problems, yes, but the classic tradeoffs still apply in a huge way. There are so many examples of this that it seems silly to even give 'em. Let's say you're writing a 3D physics package. Which is faster: (1) compute the convex hull of each object on the fly for each potential collision, or (2) precompute the convex hulls and store them in memory? There's no way the former is going to be faster for more than trivial models, but it sure is lighter on the memory. And so on and so on.

Re:how long....

[Jeff DeMaagd]

Take a look at the ultra low volt Dothans. You can get them such that they run low single-digit-watts. The ULV series runs from 800MHz to 1.2 GHZ, I think. This is a PIII-based chip with up to 2MB of on-die cache.

Re:how long....

[Tim C]

That's great in theory, but there are a few problems with your "desktop cluster" idea:

1) not all tasks are paralellisable (in fact most aren't), so for the majority of desktop users it would be wasted
2) it'd be very much more expensive than a traditional PC
3) it'd be *huge*

Don't get me wrong, I think that at the moment at least, multi-CPU machines are increasingly looking to be the way to go for those of us who will take as much processing power as we can get. A cluster that size, however, isn't the way to go. I don't want a PC the size of a fridge-freezer or wardrobe.

Re:how long....

[Anonymous Coward]

Bullshit. X86 are hot because they are the fastest thing around.

Re:how long....

[Anonymous Coward]

> And processors in 8-bit home computers used milliwatts of power.

Your newfangled 3GHz Processor still uses milliwatts of power - lots of them !

Re:how long....

[Queer Boy]

is there something inherent in desktop applications that prevent some chip maker from making a really low-power, high-performance (~1GHz) processor?

Well, that depends on what you consider to be low-power. Clearly these are super low but also have relatively low MIPS compared to current processors. PowerPC has always had low power consumption. The G5 is probably the most power hungry and consumes 42 Watts at 1.8 GHz, the G4 at 1 GHz consumes 30 Watts. Compare that with x86 processors coming from Intel and AMD and they are low power by comparison (a 2.8 GHz P4 consumes 68.4 Watts). Look at the MIPS to see the advantage.

I am assuming that you think desktop chips are just x86.

Re:how long....

[Enigma_Man]

1) Agreed, not all tasks benefit from parallel processing, but those that do would love running on 100 CPUs.

2) Not necessarily. Low powered embedded-style chips like these can cost on the order of a dollar or two apiece. Consider that the new P4 chips are $$$$, and the price isn't that different.

3) With some clever arrangement (multiple dies on one physical package, etc) you could easily fit 100 CPUs into a normal desktop case. Add into that the fact that you wouldn't need any heatsinks (just some air moving over each core).

-Jesse

Re:how long....

[Anonymous Coward]

I use an Athlon XP-M 2400+ (1.8ghz) on my desktop, which is rated at 45W max with the stock 1.45V setting. It easily overclocks to 3000+ (2.25ghz), but I usually underclock and undervolt it.
At 1.8ghz it runs completely stable undervolted to 1.25V to consume around 30W. I usually have it undervolted to 1.125V and underclocked to 1700+ (1.35ghz) where it consumes less than 20W and runs completely fanless with a big heatsink (never seen it go over 45C). So I don't find these PPC numbers very impressive at all.

Re:how long....

[CPUGuy]

Yeah, and much is that fat CPU increasing your monthly electric bill?

Re:how long....

[SlayerofGods]

Not much.
It uses what? 100 watts at the most?
There's ~720 hours a month. So running 24/7 you'd use 72,000 watt hours a month @ 8.6 cents per kilowatt hour that's like $6.19 a month....

Re:how long....

[magarity]

You never saw heatsinks on CPUs until the mid 1990s

The heck you say; my 386-20 in 1989 had to have a 2 inch tall heatsink and fan on it in order to run stably at 33Mhz.

And who claims overclocking kills? I used that rig in a variety of roles til the motherboard went out just a few of years ago. It just wasn't cost effective to buy a 386 motherboard in 2002...

Re:how long....

[The Master Control P]

When the power supply and dissipation of the average desktop become excessively expensive. Your desktop can draw all the power a single- or dual-cpu machine requires in essentially limitless quantities from the wall for pennies to nickels per kilowatt-hour. It can then dump the heat it generates into the case and the surrounding air with no risk of scalding your hands or testicles.

With that in mind, it's certainly *possible* to create a low-power, good-performance system and use it as a desktop, but it's not (given the power use of current-generation high-power CPUs) yet economical.

Re:how long....

[ball-lightning]

Ever use one of those? At 700mhz they feel a lot slower than my P2 300mhz... (plugged in!) They're pretty nice power-consumption wise, but powerhouses they are not.

Re:how long....

[drinkypoo]

Most tasks are parallelizable given enough bus bandwidth or enough acceptable latency. Things like file compression, for example, demand a certain high level of throughput. Video encoding, on the other hand, can be broken up into chunks between keyframes and stitched back together. Elements of video games could be farmed out to cluster machines or broken out into threads to run on other local processors, but they aren't because we don't have machines like that.

If you could really have an assload of CPUs in a single desktop system, let's say between 64 and 128 of them, we would start seeing very aggressively multithreaded applications become the norm. It would be no big deal to have several hundred threads floating around, because you would be able to handle so many context switches. Threads would gradually migrate from processor to processor and everything would work out just fine.

Most threads that aren't parallelizable don't require much power to begin with. If you could get some hardware in there to do floating point math it would be perfectly acceptable to use a boatload of these alchemy processors and some fairly parallelized programs. Perhaps you could come up with a PCI-Express x16 video card that does double-precision fp math, hell maybe a couple of them. It wouldn't leave you much fp in the grand scheme of things, though.

The only problem I can see with the whole thing is that the crossbar logic becomes a pain in the ass. However, if you had a bus like HyperTransport, you could make a pretty convincing machine. Processors would come in modules with some RAM (Probably 64-256MB) and you'd just stack them up and put a crossbar at the top. The only questions remaining to be answered are how to make it cheap and how to make it so jimbob can plug it in. I'd like to to see fiberoptic communications and power provided with its own protection device, and I'd like to see processors fail gracefully such that if they lost power they would revert to being a jumper for the processors (or whatever) surrounding it. That sounds kind of expensive, but the fiber can wait.

Re:how long....

[evilviper]

is there something inherent in desktop applications that prevent some chip maker from making a really low-power, high-performance (~1GHz) processor?

HAHA!

What you obviously don't realize, is that Intel's LV P3 933MHz (1.15v) processor was very low power. Under 12watts, the lowest spec chip since they made 486s. Even if you can't find that specific chip, P3s and Celerons of about that speed are only a few watts more.

If you are an AMD fan, the Sempron 1GHz proc is only 20watts, and there are plenty of AMD64 procs with nearly as-low power specs.

In addition, run fvcool on your AMD system to get the S2K support, and the processor will be using a fraction of that power when it is idling. Intel has the advantage of not needing stupid hacks like S2K bus disconnect, and doesn't require certain special motherboards. But if you have the right motherboard, AMD is just as good.

So, what the hell are you complaining about?

What I want to know is when motherboard manufacturers are going to stop keeping their specs secret. After buying a couple very high-power motherboards, I've tried to get that information from all the major manufacturers like MSI, Asus, Gigabyte, etc., and they all either promise it they'll send me those (but never do), or they ignore my request completely. I'd buy from PCChips at this point, if I could just get the power specs from their site for each motherboard.

Re:how long....

[CPUGuy]

I have 4 computers @ home, plus, $6.19 is $6.19. Thats $74.28/year. Why is it that when we have a $.30 coupon for something we jump to use it, but when something is just another $6.19 a month to your >$200 power bill, it's nothing at all.

HTPC

[OverlordQ]

What socket are these going to be? Will they fit into existing Mobo's? Sounds like they'd be a good chip to put into a HTPC.

Re:HTPC

[Anonymous Coward]

They're MIPS, therefore: they won't be 'socket' anything, and no, they won't fit in existing motherboards. (Except maybe for some existing MIPS boards. I don't know enough about that hardware architecture to know if the various versions were pin-compatible or not.)

Re:HTPC

[Anonymous Coward]

These are not any "socket". They are intended for embedded applications, and integrate memory controller, I/O controller, graphics controller and video acceleration all into one chip.

Hmmm...

[Stanistani]

*scribbles info on back of napkin*

This looks like a possible solution to something I'm actually working on...

Especially this (from the AMD link):
AMD Alchemy(TM) Au1500(TM) Processor and AMD Alchemy(TM) DBAu1500(TM) Development Board

The AMD Alchemy(TM) Au1500(TM) processor provides a high-performance, low-power system-on-a-chip (SOC) for the Internet Edge Device market. These devices include thin clients, Web servers, routers, printers, wired and wireless gateways, Wireless Access Points (WAPs), handheld computing, Web pads, set top boxes, and multimedia applications.

Features of the Au1500 processor include the following:
Optimum performance at very low power
Highly integrated features: on-chip memory controllers, a 66MHz PCI 2.2 controller, and Internet access peripherals
Runs a variety of operating systems, including Windows® CE.NET, Linux, and VxWorks
Integration of peripherals with this unique, very high-performance MIPS®-compatible core can provide lower system costs, smaller form factors, lower system power requirements, simpler designs at multiple performance points, and shorter design cycles
The AMD Alchemy(TM) DBAu1500(TM) development board allows you to evaluate the operation and performance of the Au1500 processor. This highly versatile system serves multiple purposes:
Chip evaluation
Software development
Example of hardware systems design using the Au1500 processor

Re:Hmmm...

[Humorously_Inept]

Looks like a lot of embedded system potential! The built in controllers should save designers a lot of time.

for the record:

[drinkypoo]

These processors are new but the Alchemy name is not. AMD has been selling alchemy processors for at least a year.

Mini-ITX replacement...

[B5_geek]

While I love the size and versatile nature of Via's Mini-ITX line, I have found their systems to be very unreliable.

These AMD systems would be perfect for many linux applications;

firewall, file servers, dumb-terminals, HTPC boxes, hell make a cluster out of 100 of them and they still waste less energy then a P4!

It would be cool to see how a cluster like that could handle mpeg4 encoding/decoding.

You also have CarPC's and many other options.
I want some, can ya tell?

Re:Mini-ITX replacement...

[2nd Post!]

Have you taken a look at the Mac mini, for your needs then?

They use a CPU that scores in the 1.3V range, and I wouldn't be surprised if that meant it only drew 15-20W for a 1.25GHz CPU.

That, and it's only 6.5x6.5x2 in dimensions

Re:Mini-ITX replacement...

[Donny Smith]

> It would be cool to see how a cluster like that could handle mpeg4 encoding/decoding.

Did you see that it barely (30fps) does proper playback?

I think in a cluster it'd run like shit because this CPU is not suitable for that purpose (or, you can say that line of CPUs wasn't made for HPC - AMD64 would do much much better).

Generally speaking Intel processors are more suitable for such tasks.

Re:Mini-ITX replacement...

[B5_geek]

No, I don't know anything about Mac's. Does the software exist (free like linny) that does everything linux can do?

Re:Mini-ITX replacement...

[Joseph Vigneau]

Don't hold your breath for a PC that has this processor in it. Besides, if you really want a MIPS box running Linux, get a Playstation2...

Re:Mini-ITX replacement...

[B5_geek]

Just kind of "thinking-out-loud" but, why not have it setup so that each system in the cluster only worked on 25% (if there are say 5 systems + 1 traffic cop)

And add to the fact that Most TV's display less then 600x400 and 30fps is enough for video playback, it just might be beefy enough.

Re:Mini-ITX replacement...

[Anonymous Coward]

You can install, errr, "linny" on it if you wish...

Re:Mini-ITX replacement...

[aristotle-dude]

There is the fink http://fink.sourceforge.net/ project, Gentoo for Mac OS X http://www.metadistribution.org/macos/, Darwin ports http://darwinports.opendarwin.org/ or you could just install linux.

Also check out versiontracker.com for a lot of Open source software with pretty GUIs.

Re:Mini-ITX replacement...

[Anonymous Coward]

Via's mini-itx stuff is unreliable? Could you go into detail? I'm thinking of buying a M10000 mobo.

Re:Mini-ITX replacement...

[garethw]

Yeah, I was curious about that comment, too. I've had a M10000 (Nehemiah) for about a year, and more recently I purchased an MII 6000E fanless model. I run both of them diskless, booting over NFS, keeping them cool and quiet (absolutely silent in the case of the MII 6000E).

The fan of the M10000 has become a little loud recently - it seems to rattle, but that's about the only issue I have, and the fan could be replaced.

Re:Mini-ITX replacement...

[Anonymous Coward]

These things are not x86 compatible dude.

I do agree with you though, VIA sucks ass. Their chipsets and their CPU's. Blech.

Re:Mini-ITX replacement...

[n9fzx]

I was wondering about this comment myself. I've been running an Eden system pretty much continuously for the past three years, never a problem; an Epia/800 system for about the same time in an industrial environment, again no problem. Have an M6000 compiling KDE on my bench at the moment, running without a hitch. I use laptop drives to keep the heat down, and because they're generally more reliable.

Re:Mini-ITX replacement...

[Anonymous Coward]

I've had an Epia/800 running constantly for almost two years, been rebooted 4 times for kernel upgrades only. Chugging along as our cvs repository, web-server, firewall and mail. Constantly connected to internet via a 10mbit line. Btw, It's actually running only with the supplied heatsink (original fan removed due to noise) The fan of the (standard) power supply seems to provide enough air circulation in the micro-atx Aopen box.

Good chips are not the problem

[LittleLebowskiUrbanA]

Getting these things into some decent laptops is. I only issue IBMs at my company and for good reason: the Stinkpads are built like tanks.

Re:Good chips are not the problem

[Anonymous Coward]

MIPS instruction set is not supported on WinXP, so would limit general laptop use.

Re:Good chips are not the problem

[Anonymous Coward]

No, these are built like tanks: http://http//www.panasonic.com/computer/toughbook/ video2.asp?videoname=ThisIsToughbook

Re:Good chips are not the problem

[Anonymous Coward]

Stupid me, screwed up, even with preview! No, these are built like tanks: http://www.panasonic.com/computer/toughbook/video2 .asp?videoname=ThisIsToughbook

Re:Good chips are not the problem

[Anonymous Coward]

http://catalog2.panasonic.com/webapp/wcs/stores/se rvlet/ModelDetail?displayTab=O&storeId=11201&catal ogId=13051&itemId=68953&catGroupId=16916&modelNo=T oughbook-29&surfModel=Toughbook-29

Video won't play for some reason, but dig on that.

Re:Good chips are not the problem

[Anonymous Coward]

MIPS is supported by linux so you have your email / web / general network support, open office runs on linux so you have word processing, spreadsheets, presentation software etc. etc. Just why would MIPS limit general laptop use?

Re:Good chips are not the problem

[Anonymous Coward]

Because 90% of the world uses Windows.

From the Product Briefs linked to in the article:

[grommit]

# Support for MPEG1, 2, 4, and WMV9 scaled up to 1024x768
# MPEG2 main profile/main level (720x480, 10Mbps, 30fps)
# MPEG4 advanced simple profile/level 5 (720x480, 8Mbps, 30fps)
# WMV9 main profile/medium level (720x480, 2Mbps, 30fps)

Doesn't look too bad to me. This was for the Au1200 btw.

Old News

[Anonymous Coward]

I was looking at these for a project two years ago. Must be a slow news day. They are very nice for what they are though.

transmeta

[Anonymouse Cownerd]

dare i say it? this makes transmeta all the more irrelevant. once a cool company with innovative technology, now they are no more than third place runner-up in the processor company race, and falling behind fast.

Re:transmeta

[Anonymous Coward]

Actually, these have little to do with Transmeta. Alchemy chips aren't x86--Transmeta chips are--so they're not competing for exactly the same market segment.

Now AMD GEODE processors on the other hand, THEY compete with Transmeta (and Via). Too bad the available motherboards don't seem very exciting, the chips themselves don't seem bad.

Re:transmeta

[DataPath]

One difference - these Alchemy chips are NOT x86 architecture. The transmeta chips are.

Re:transmeta

[NaruVonWilkins]

I doubt that. Several companies (especially Sharp) have laptops running Transmetas that are doing very well in the Japanese market. Sharp still has prime corner real estate in Akihabara for a reason.

Yeah, but...

[Anonymous Coward]

powerful enough to ruu DivX, WMV9, and MPEG

Why on earth would TiVo be running this on the main CPU? I had thought the direction for DVR's was to offload most of the encoding/decoding to the video card/cards, no?

It's not terribly impressive to say "can run MPEG-2 for video encoding!" when the main CPU's not doing the actual work...

Sharp has something similar

[corngrower]

These appear to be MIPS based chips. Sharp has some ARM based chips that are geard for similar applications and provide similar functionality. These also, IIRC, run with fairly low power and have 32 bit cores. The high power requirements of x86 architectures is due to their very high clock speeds and all the complexities of a general purpose CPU chip (mmu, floating point, cache, fancy branch prediciton logic, etc).

http://www.sharpsma.com/sma/products/mcu_soc/LH7A4 04_splash.htm

Re:Sharp has something similar

[cnettel]

And, of course, fairly high performance per cycle. I think that even a Prescott will crunch out more IPC than a Cyrix C3. IPW (instructions per watt) is of course a different matter...

And, just before anyone asks me to do an ARM-to-x86 comparison of IPC, that won't work. If you have a RISC architecture, of course you'll get high IPC. Make each instruction easy and atomic and you get a lot of benefits, including bogus IPC numbers for comparisons. That's why we need MIPS (not the architecture) and MFLOPS for architecture comparisons, and it's still a matter of lying, statistics and... *shudder* benchmarking.

Wow feal the burn

[Mr. Falco]

Wow these processors are going to fuel greater expansion in pc development. Alchemy has truly come of age. From an old science to modern masterpiece. wow.

Well done, AMD

[masterOfTheObivous]

This might very well be AMD's next big thing. The Athlon 64 garnered a lot of attention for them, but now they've entered a new market- competing with Via's Mini-ITX series. SFF's that need the power for MPEG-4 decoding so they can be a good home theater PC would do well to be equipped with one of these. In fact, they even mention:

"AMD Alchemy(TM) Au1200(TM) Processor - is a low-power, high-performance processor solution for Personal Media Player (PMP), automotive and Digital Media Adapter (DMA) applications.

The implications of a low-power, low-heat solution with a lot power go beyond the home theater. The idea of "ubiquitous computing" (IMHO an awful blanket term that gets thrown around far too often) might become possible with a small but still powerful processor.

The one last innovation that caught my eye was the on-processor AES encryption/decryption. Anyone have any ideas of practical applications for this?

Re:Well done, AMD

[Lisandro]

I was thinking the same - they quote power consumption between 400mW and 1W. *ONE* watt. At 400MHz. It consumes the same as older processors like the Z80 (while smoking it in processing power)! Even Transmeta couldn't get below 5W with their crusoe line - yet, the Crusoe is x86.

If AMD markets this thing right and performs as promised, they will make a killing out of it. There's a lot of money in the embeeded systems market.

Re:Well done, AMD

[SpongeBobLinuxPants]

Why do we need low-power cpu's in a home theater system when an hdtv takes 300 watts or more to have on? What's an extra couple of watts in the cpu going to save?

Re:Well done, AMD

[AddressException]

Get a better TV then, you Mac mini sponging scum.
Also, learn how to use the apostrophe.

Thanks.

Re:Well done, AMD

[Temsi]

need the power for MPEG-4 decoding so they can be a good home theater PC
Since most people don't really need a second computer dedicated to TV watching, this could be a great for a WIFI/Ethernet box that reads videos from the main computer in the house, or even from a firewire drive.
Small enough to put in the living room, and since it has no fan, is completely silent (provided it's diskless as well).

Having wasted my money on the 933Mhz Mini-ITX that could barely decode mpeg2, this sounds infinitely better.
Granted, HDTV support would be great, but I think it's only a matter of time before AMD goes in that direction.
A fanless multimedia front end capable of processing WM9,MPEG4,MPEG2 decoding at full 1920x1080 resolutions would go straight to the top of my list of must haves.

Re:Well done, AMD

[SpongeBobLinuxPants]

Accord'n to cdtv:

Typical Power Consumption - HDTV
Plasma TV - 50 in. 650 Watts
CRT Proj. TV - 50 in 350 Watts
CRT TV - 32 in 250 Watts
DLP TV - 50 in. 200 Watts
LCD TV - 40 in. 180 Watts
STB Cable or Satellite 60 Watts
Audio system 200 Watts

Also a hittachi plasma consumes 530 watts. So, in actual'ty my crappy CRT has lower power consumpt'n than a "better" plasma, although, not as low as an LCD. Even the power hogging P4's and old Athlon XP's use less watts than the TV.

Re:Well done, AMD

[antispam_ben]

It consumes the same as older processors like the Z80 (while smoking it in processing power)!

I'm glad I'm not the only one who remembers when processors didn't need heat sinks. My specialty was the 6502.

If AMD markets this thing right and performs as promised, they will make a killing out of it. There's a lot of money in the embeeded systems market.

Maybe. One big question in the embedded market is how long a part is expected to be available. A couple of years is plenty of time in the PC market where a product lifetime is short, and no one would think of selling a product based on two-year-old technology. Embedded manufacturers like their products to have a longer lifetime, and likewise for the components, so that a year or two down the road production doesn't get stopped for silly things like new PC board layout for the "new, improved" replacement chip.

Re:Well done, AMD

[Anonymous Coward]

What's an extra couple of watts in the cpu going to save?

Heat disipation mainly.

Re:Well done, AMD

[Lisandro]

Ohhh, the venerable 6502. The Z80, 6502 and other 8-bit microprocessors are still arround, live and kicking - they're cheap, low power and quite flexible for all sort of designs.

Of course, you know this. But now imagine what would happen if AMD just priced this processor right - not giving it away, but cheap enough so it becomes a serious consideration for embeeded design. You'd have performance, consumption AND price; what keeps you from using it instead of a 6502 other than nostalgia value?

Maybe. One big question in the embedded market is how long a part is expected to be available

Great point, i haven't tought of it. I imagine if the part becomes popular enough it's existance will be more or less assured. If there's a market, there will be a seller. Hence, marketing is key - there's a lot of great processor designs that never made it (Crusoe, for example).

Re:Well done, AMD

[Lisandro]

Reiability. The moment you part off heatsinks, coolers and such, you ensure that your lovely new media center won't die overnight because you left it running 24/7 for a year and a cooler broke. Also it becomes smaller and quieter, which is a bless.

Re:Well done, AMD

[fish waffle]

You'd have performance, consumption AND price; what keeps you from using it instead of a 6502 other than nostalgia value?

An enormous existing investment. That includes employee knowledge, tools, techniques, inventory, related components, purchasing deals, ...

Slightly OT: More efficient CPUs are long overdue

[bach37]

Thunderbird or outlook express starts just as fast on a 400Mhz machine as a 3Ghz machine. I can understand hard core gamers or pro studios/labs who need the beefy cpu, but I hope this turns the trend to making more efficient processors, as we are hitting the 4Ghz barrier now. If we just crank it more, won't cpus soon be putting out microwaves soon? Not to mention need water cooling.

/my rant for the day

Blackfin

[arsenix]

The Analog devices blackfin is a good chip for such devices as well (it is classed a "hybrid DSP"). It consumes 280mw at 600 Mhz, and also comes in a dual core. uCLinux runs on it as well. Cool chip with similar capabilities, but definetely a different marketing angle.

Re:Blackfin

[Kiryat Malachi]

AD likes to charge *bank* for their parts, though. I love a lot of their chips, they make some great stuff (the SSM series has some fun toys, especially), but they do charge for them.

overclockers

[frankmu]

i'm waiting to see if the modders will overclock these chips and put major heatsinks on them.

that'll teach AMD!

Transmeta technology?

[WindBourne]

One of the best thing coming from Transmeta was the low power. I wonder if this part of what was sold.

embedded processor supercomputer

[flaming-opus]

That's EXACTLY what blue gene is. It happens to use a derivative of IBM's embedded processor, rather than AMD's, but that's the design. Each node is pretty wimpy, but make up for that by using 65,000 of them.

It works pretty well for problems that can be cut up into 65,000 pieces. However, a lot of problems don't easily divide into that many tiles, or the work of parallelizing the problem exceedes the benefit at some point.

Re:embedded processor supercomputer

[bhima]

That's Blue Gene /L, Blue Gene /C uses hotter node CPUs and Blue Gene /C uses interesting network topology.

I wonder when they'll make Blue Gene /Cell.

Yeah

[Anonymous Coward]

Then we'd be able to comfortably have our laptops on our laps without killing off our future offspring.

These are not new

[Weaselmancer]

I've had an Alchemy Au1100 devboard on my desk for over a year. The disk that came with the devboard is dated 1-27-2003.

There is already a very complete Linux port mostly done by Montavista.

These are not exactly new

[Anonymous Coward]

The A1x00 family has been around for a couple of years. Also it should be noted that the Au1200 is the only chip in the family intended for the Portable Media Player (PMP) market with its built in Media Acceleration Engine (MAE). The MAE is what allows the Au1200 to decode MPEG2/MPEG4/WMV9 at D1 resolution. The Au1550 is meant for network devices like gateways and routers.

Re:These are not exactly new

[webmind]

indeed, had my Au1500 powered meshcube for atleast 3/4 of a year now..
and this thing already runs linux by default and has wireless and ethernet and usb..
and yes it's damn low power (around 6watts of power usage with Wlan afaik)

Micro-ITX

[colmore]

When these fall in price enough, and when there's a board out in Micro-ITX spec for them...

helloooooo dashboard computer

Desktops don't usually run on batteries.

[cbreaker]

It's the truth. I'd rather have a 20 watt CPU running fast then a 1 watt cpu that runs much slower, when I am plugged into an outlet.

While I do feel as though some power conservation could be in order; if only to reduce heat and thus fan power. But I don't care about the power usage itself - if it's silent and uses 100 watts, I'm okay with that. The wall can provide enough power.

For laptops, PDA's, phones, etc.. it's a different story. You want the batteries to last for as long as possible.

Re:Desktops don't usually run on batteries.

[Anonymous Coward]

The wall can provide enough power.

heh, wait til you plug in the video card.

Re:Desktops don't usually run on batteries.

[bob65]

if it's silent and uses 100 watts, I'm okay with that.

Yeah but how likely is it going to be silent at 100 watts, unless you use some wacky sort of water cooling (and even then, there's the pump...)

Re:Desktops don't usually run on batteries.

[Anonymous Coward]

Besides,

how much did PCs consume electricity in California in 2001? Around 12%? I bet that number has gone up a bit. If everyone used a low-power PC you/we wouldn't have to worry about the Kyoto protocol. Oh wait, the US electorate and the government doesn't worry about that.

no heat?

[AviLazar]

and on my budget (computer or heat the home) how am i supposed to heat my home?

Re:no heat?

[Misroi]

I doubt the 1/2 watt the cpu consumes will hurt your electrical bill... you'll be able to have not only your computer but also heat!

Why Tivo would run this on the main CPU...

[Moryath]

Two words:

NO HEATSINK.

If you can get a video board that works with only a passive heatsink, and then run this thing with a minimal heatsink, you lower your heat problems.

Lower them enough, and you can get a smaller fan to cool the entire unit, or even get away without a fan entirely (though given how long a TIVO has to stay turned on, it's likely you need some minimal level of guaranteed airflow to avoid overheating the unit the same way you used to be able to overheat an NES).

But the smaller, and fewer, fans you have to put into it, the quieter it is. And living-room appliances want to be as quiet as possible, to avoid interfering with the quiet moments inside of a game/movie/TV show.

Re:Why Tivo would run this on the main CPU...

[Anonymous Coward]

also, fans will break down after a few years of running so I wouldn't want any fans in my living-room appliances

Re:Why Tivo would run this on the main CPU...

[mikefe]

Actually, larger fans that run at lower RPMs are quieter than smaller fans that run at higher RPMs.

You typically have smaller fans on processsor/chipset/GPU chips, so if you can replace those with passive heat sinks, you can let a larger case fan do the work and have less noise.

Not x86

[Anonymous Coward]

That'd be a great idea, I'm sure NOONE at AMD thought of that one smart guy.

By the way, these are not x86 cpus.

What about *MORE* power?

[SmokeHalo]

C'mon, we've done fried eggs on a CPU, but that's not enough. I wanna cook a T-bone on that sucker!

Re:What about *MORE* power?

[strelitsa]

Well, here's your chance.

Now? Last year? Two years ago?

[2nd Post!]

What do you think Apple puts into their Mac mini?

They use a 1.3V CPU laptop style CPU in a desktop unit.

They've been using these for years; see the iMacs, for example.

Big problem

[emil]

Is some sort of non-executible stack possible on this architecture?

If I remember correctly, OpenBSD does not implement W^X on the MIPS architecture because the design of the unified data/instruction cache makes it impossible to mark pages as non-executible.

If NX/PaX/W^X will not work on this CPU, I'd be less inclined to attach it to the internet.

Re:Big problem

[Anonymous Coward]

I don't believe NX existed on MIPS32, but only on MIPS64.

We use these (they are smokin!)

[Anonymous Coward]

We have ported eCos RTOS to Au1x00 and have used Alchemy CPUs in two embedded products. They have outstanding performance, good GNU tool support, and easy-to-understand MIPS risc goodness.

One thing to watch for: The onboard peripherals are geared more to PDAs (no real watchdog, limited-feature timers, etc). You would want to check your embedded application requirements. On the plus side, the JTAG TAP makes board support and debugging a snap.

Re:We use these (they are smokin!)

[scatterbrained]

I wish you guys would have submitted your Au1x00
patches back to ecos! I wanted to use redboot
on something we're doing, but wound up using
u-boot instead, just because I didn't have time
to get all the scaffolding in place for redboot.

Re:We use these (they are smokin!)

[Anonymous Coward]

We are going to release the port back to the eCos community when we have cleaned up the code a bit and when we better understand the eCos way of doing things (configuration items and packaging).

Re:These are not new - PS

[Weaselmancer]

Here is an already-in-the-market device that uses the Au1100. It's an automotive scantool by Chrysler. Note the article is dated May 5, 2004. They were in production at the time.

Daimler-Chrysler scantool link.

AMD mucking around in other fields

[oboylet]

I'm always pleasantly surprised with what AMD cooks up in addition to their x86 business.

AMD is a much more interesting company that we geeks often realize. Too often we think, AMD=Athlon/Opteron, but I find their gadgety endeavors really interesting.

Apple's Airport (and maybe extreme/express, dunno) has a tiny AMD processor , and as the parent points out, now their playing with MIPS archs. A friend of mine worked at the fab in Dresden and said that a third of their operations had to do with flash.

Call me a fanboy, but I sure do like the AMD kool aid. They make neato products and deserve mucho respect.

Re:AMD mucking around in other fields

[SIGPUNKT]

And let's not forget that it was the bit-slice AMD 2901 at the heart of the VAX-11/780. Still the nicest machine I've ever worked on, I'm looking forward to getting an emulator running (at probably 10x the speed of the original).

Re:AMD mucking around in other fields

[iabervon]

Call me a fanboy, but I sure do like the AMD kool aid.

Wouldn't it be more appropriate this time to call you a nofanboy?

Re:AMD mucking around in other fields

[Anonymous Coward]

Yeah, like all those "neato" CPU's that will kill themselves and all your expensive equipment connected to it (ie. motherboard, cards, etc) if you remove the heatsink. It's a wonderful design for throwing away money.

The Opteron is the only one with thermal protection (versus Intel where every CPU since the P3 is protected). You're Athlon-64 will be more than happy to eat itself if your CPU fan dies.

Re:AMD mucking around in other fields

[Ben Hutchings]

The original Alchemy chips were developed by a separate company, Alchemy Semiconductors, which AMD bought 3 years ago; they're not entirely an AMD invention.

Re:AMD mucking around in other fields

[JimmehAH]

AFAIK most motherboards for AMD processors have thermal protection for the processor. Not as good as on-chip stuff but still nice to have.

Re:AMD mucking around in other fields

[Anonymous Coward]

fanlessboy?

Re:AMD mucking around in other fields

[Anonymous Coward]

Uh, sorta. It's controlled by the BIOS and sometimes needs OS support.

It's a "soft" solution to a hard problem. Motherboard sensors are notoriously inaccurate. Besides, they are too slow at detecting temperature changes. Your stuff will be dead before it reduces the clock.

AMD are morons for not putting in a $5 protection diode and circuit.

Re:AMD mucking around in other fields

[MrResistor]

Networking chips, also. And I've seen their name on other stuff too.

I'd guess they have a finger in at least every pie Intel does.

Re:AMD mucking around in other fields

[Anonymous Coward]

Kool Aid guy bust through wall
Kool Aid Guy: Oh Yeah!!!

Re:AMD mucking around in other fields

[Anonymous Coward]

Apple's Airport (and maybe extreme/express, dunno) has a tiny AMD processor , and as the parent points out, now their playing with MIPS archs.
That "tiny AMD processor" WAS one of the MIPS chips. The story submitter doesn't seem to be aware that the Alchemy line has existed for years.

The Transmeta Factor

[The-Perl-CD-Bookshel]

Perhaps this technology comes from their strategic partnership with Transmeta whom has always put great emphasis on energy conservation. Decent Google search here.

Re:The Transmeta Factor

[scatterbrained]

AFAIK, Alchemy semiconductor developed this stuff
themselves - it had nothing to do with Transmeta. AMD bought Alchemy a few years ago.

Alchemy sold the Au1000 and I believe 1100 and 1500
themselves. The 1550 came out roughly a year ago,
and the 1200 is brand spanking new.

Re:Beowulf

[Anonymous Coward]

Its a lame phrase that should have died years ago.
Check the second post faggot

Very nice, but...

[Anonymous Coward]

...will it run Linux?

Re:Very nice, but...

[scatterbrained]

yes

http://www.linux-mips.org/cvsweb/linux/arch/mips /a u1000/

VIA Eden Still PC-Compatible King?

[BaldingByMicrosoft]

So, for a PC-compatible CPU, it looks like the VIA Eden series is still on top?

VIA Eden ESP 4000 (4.0 x 100MHz)@1.05V: 1.7W typical, 3.0W max
VIA Eden ESP 10000 (5.0 x 200MHz)@1.05V: 6.1W typical, 7.0W max

Unlike a previous poster, I've been running a VIA EPIA M 10000, 1.0GHz (Nehemiah) on a workstation, and a VIA EPIA V, Eden 533MHz on a server with no issues.

Re:VIA Eden Still PC-Compatible King?

[MarcQuadra]

I can EMULATE x86 on a PowerPC 750 and get better performance/watt than VIA's EDEN chips. Really, VIA chips have always been seriously weak sauce, just because they can't clock them well and they use not-much juice doesn't make them superior in any way.

The Pentium-M and AMD GEODE NX CPUs produce MUCH more horsepower per watt than VIA's chips. VIA's only advantage is that they mass-produce Mini-ITX boards and sell them to distributors. If someone mass-produced a GEODE NX bMini-ITX board, it would wipe the floor with the EPIA, fanlessly.

on-processor AES

[thrashbluegrass]

A low-power firewall acting as a VPN concentrator could certainly take advantage of crypto hardware.

Or, for the pathologically paranoid (join with me, my Pathanoid kin!), quick swap encryption sounds pretty tasty.

Implementations

[SumDog]

Does anyone have links to implementations of this MIPS arch? Are there embeded versions of Gentoo or Debian that have been showen to work on this chip? Are there any media players like mplayer that are designed to support its instruction sets. What type of embeded boad solutions are there? Has anyone tried this and what are your experiences?

Re:Implementations

[10Brett-T]

Gentoo MIPS -- not complete, not very active, but at least started.

Re:Implementations

[Anonymous Coward]

http://http.us.debian.org/debian/dists/woody/main/ disks-mips/current/r4k-ip22/images-1.44/
~omi

VIA

[phorm]

Looks like VIA will have some competition. I've got a few Epia boards which I mainly use due to the low profile/power consumption.

Once I see how well the AMD products (including the motherboards for the chip, etc) work with linux I may consider a switch. At 400Mhz equivilent they could do nicely for servers and the video capabilities would make them decent enough for small media units. Wonder how well they would handle DVD, etc playback and TV out... as my M10000 does quite nicely for that with fairly low CPU consumption.

Re:VIA

[Gogo Dodo]

AMD's Alchemy processors are not x86 compatible so you can't directly compare it to your C3/Eden setup.

FLASH

[bsd4me]

AMD also makes FLASH devices, and they have for a long time. Given how many things have FLASH in them, I suspect that it is their cash cow.

Re:FLASH

[corngrower]

If you've been paying attn to the stock market, last quarter AMD did not make nearly as much as they were expecting on flash sales because Intel was dumping flash memory in order to depress prices of flash. Thoughts are that this was done specifically to hurt AMD's bottom line.

Re:FLASH

[mkarpinski]

Actually, AMD is losing money with their flash business. They may be looking at dumping those products...

AMD Without the flash?

So no HDTV?

[glrotate]

Hmm. 1998 is calling, it wants its CPU back.

Re:So no HDTV?

[Anonymous Coward]

1998 just called to thank us for the 0.5W CPU revolution we sent them. They promised to use the technology to send us Pentium 4s that don't accidentally fuse hydrogen when overclocked.

Re:So no HDTV?

[Datafage]

Yes, because if it can't run HDTV while sitting in my palm it's useless.

20 watts

[SWTP_OS9]

Just saw Toms Hardware with a review on the Mac Mini. Stated it only eats 20 watts and 28 watts for DVD playback. With a PC eating 160 watts.

Not bad for a little 1.2~1.4Ghz box!

They also had a review on SFF units. Still noize! Looked like a hardware hack to the Mini.

Re:20 watts

[schleyfox]

Crap, My friend's uber computer + monitors uses precisely one kilowatt!

The OCer says...

[lateralus_1024]

Can't wait for ThermalTake to make a Copper Brick heatsink for this bad boy. I'll break the 2Ghz barrier with this. So what if my PDA weighs 15lbs, can yours run AutoCad?

Re:The OCer says...

[drinkypoo]

What version of Autocad? I have a gridpad 1910 with dr-dos and pc-geos, 2MB RAM, and a 20 meg hard drive... With the original graffiti handwriting recognition system from Palm Computing (back in the Zoomer days) it's not a bad PDA, either. I do need to get the rest of the built-in applications from my gridpad 2390, though.

Re:I remember the R4k

[winkydink]

It was the chip that killed MIPS as an independent company. The R6000 that preceeded it was late. The R4000 was late as well. The lateness of the R4000, coupled with Blob Miller, the CEO, being more focused on writing his own "Soul of a New Machine" than the business, killed the company. As a matter of fact, the running joke at the time was that MIPS stood for "Miller Inserts Penis in Shareholders".

low power x86 solution.

[CdBee]

If what you need is a cheap, low-current x86 workstation, consider putting a Slot-1 Pentium 3 processor into a Pentium-2 motherboard and underclocking it by say 30%

Add a bootable PCI ATA133/SATA controller (P2 mobo's have 66/100mhz drive controllers) and a USB2.0 or firewire card and away you go. Instant cheapo mediabox or server.

Total cost would be ridiculously low - probably less than £100 (GBP - my currency) or say $160 US?

Re:low power x86 solution.

[beyondtheblack]

The answer to all of these problems is the Pentium M. This should be the de facto desktop chip for all office desktops, because except for gaming and floating point deals, the PentM kills on productivity and halves (at least) power consumption. What gives? Toss the P4 into the river. On a large scale, companies could save thousands on power alone with these babies. Sure, initial investment is more now...but ramp up production and we're there. Plus...new 915s with dual-channel DDR...nice.

Re:low power x86 solution.

[Anonymous Coward]

Actually if they go with Pentium Ms they will have to pay more for heating, my LAPTOP is a P4 and lets just say it could neuter you if you put it in your lap.

Re:low power x86 solution.

[Entropius]

Sitting here with an Athlon 64 laptop on my lap, with shorts on -- it's barely warm, and the system fan isn't running.

Since the mobile A64's hit the market, is there any reason to get a P4 laptop?

man

[sakura the mc]

when is intel going to get with the program???

i have never used an amd product before, but ive been tempted to try them out. intel better get back in the game if they dont want to lose my patronage.

x86

[eddievalentine]

Due to the fact that x86 isn't RISC.

Re:x86

[kryptkpr]

From what I've been told in my CPU architecture classes, the x86 is a CISC, but it's implemented in the hardware as RISC.

Cool, but does it fit in a wristwatch?

[davidwr]

Wristwatch Linux - cool.
Wristwatch Linux that can boot Knoppix-on-thumbdrive - cooler.

Two things not mentioned yet...

[Anonymous Coward]

1) Lower power = no heatsink needed (that was mentioned). Laptops were mentioned. But, perhaps, one of the sweetest applications for this would be in the field of intrisnically-safe (ie: explosion-proof) laptops, like the GoBooks. Traditionally hampered by the lack of a fast processor in a completely sealed environment, this could be another real step forward in indestructable laptops.

2) Does anyone else see the irony in AMD making such a low-wattage chip? The Kings of Cook, all cooled off. (and yes, I do use my dual Athlon to warm the room it's in. Really!)

late to the party

[Sebastopol]

Sorry kids, this is essentially an Embedded CPU.

Intel has had their XScale running at >1GHz and under 1W for over FOUR YEARS.

AMD is finally playing catchup, but at 400 MHz, good luck.

Re:late to the party

[Anonymous Coward]

It's really the other way around.

The Alchemy silicon was out in 2000 / 2001.

Re:late to the party

[scatterbrained]

Hogwash.

Compare the parts based on market segment - i.e.
the Intel PXA series parts (not the network
proce$$or stuff) and the Alchemy. IIRC, Intel
tops out at 600 MHz and AMD at 500 right now.
The peripheral mixes are similar, although AMD
supports DDR on their newer parts and Intel
doesn't. Power consumption at a given MHz
favored AMD last time I compared.

As a small customer AMD is a lot easier to deal
with than Intel - a bunch of stuff Intel has
you can't get unless you're Nokia or Samsung.

WOW!

[jav1231]

An AMD CPU without a heatsink? Now that is news! Barring the old chips they used to supply to NIC manufacturers, this is practically a first for them.

AMD's Geode

[dfj225]

Also of interest would be AMD's Geode line of processors, found here: http://www.amd.com/us-en/ConnectivitySolutions/Pro ductInformation/0,,50_2330_9863,00.html. It looks like these guys run on about 1 Watt and are x86 compatible.

These are great for $100 PCs

[davidwr]

One of these, an cheap 800x600 video chipset, a cheap modem and/or LAN chipset, and a few bootable usb ports, and you've got yourselves the beginnings of a Really Cheap And Quiet Computer. Perfect for libraries, schools, 3rd world users, and poor Americans.

Just add a keyboard and mouse (under $30 for set, including USB/PS2 adapter), a cheap monitor (under $50), cheap printer (under $50), and Knoppix-on-USB, and you are good to go.

An AMD without a heatsink?

[jacklebot]

Am I the only one who thinks it's really funny to see AMD and "no heatsink" in the same sentence?

Re:An AMD without a heatsink?

[DeathByDuke]

it'd be a lot more funnier if it was Intel and 'no heatsink' in the same sentence....

Re:An AMD without a heatsink?

[Neuroelectronic]

yes. Now, Prescott and "no heatsink" now that is funny.

Great use

[Anonymous Coward]

I could see a great use for these in a 24/7 entertainment/security box. The electric bill won't have to cost you an arm and a leg.

What about AMD64 Winchester?

[Ized]

You can also underclock the AMD64 Winchester processor, so it uses extremely low wattage, but still provides a 1.0Ghz+ performance.

I read on one forum someone had tested running the winchester (3200+ model) up to 1.2Ghz, with only a hint over 1.0v. Without a heatsink.

I'm currently running same cpu, but with CNQ (Cool'N Quiet active). At the moment it's automatically adjusted itself to 1.0Ghz with core at 1.08v. One could go to the BIOS and set those values there manually and run this machine without heatsink.

On a sidenote, I've looked at a number of "embedded"/small form factor boards for HTPC use and my only gripe with these new systems is that: they are _expensive_! I once even called a manufacturer to get a quote on some pentium-m almost-ITX sized board and it cost over 400eur.

Re:Wow...

[Anonymous Coward]

You've got a pretty creative use of the comma.

Why the hell arent they doing this to desktops?

[GatesGhost]

i know the power requirements might be much higher, but still, cant they try to make cooler, low power processors for the desktop? well, at least they're not burning up like the pentiums....

Cell Phones?

[Anonymous Coward]

Is AMD already in the cell phone market? I remember that the market is mainly dominated by the ARM series of processros (ARM7, ARM9, StrongARM). These processors run at about 185 to 250MHz@0.25 to 0.8mW/MHz. For a ARM940T using 0.18 micron process, it runs at 185MHz@0.8mW/MHz. This is in constrast to AU1550's 1.25mW/MHz and AU1100's 0.625mW/MHz. Granted, I don't quite know what ARM processor is "best" for the cell phone market, and I'm comparing with the "worst" mW/MHz ARM chip.

All in all, these Alchemy chips do seem to be geared towards cell phones also, which is probably bigger than than (individually, not all together) "thin clients, Web servers, routers, printers, wired and wireless gateways, Wireless Access Points (WAPs), handheld computing, Web pads, set top boxes, and multimedia applications" markets (then again, I may be totally off). IIRC, this cell phone technology issue was also mentioned in that "leaked" Intel President blog, linked from /. yesterday.

So I've heard...

[Anonymous Coward]

... the Stinkpads are built like tanks.

The Norwegian army bought 10,000 Stinkpads to beef up their tank fleet.

Re:Slightly OT: More efficient CPUs are long overd

[peragrin]

The big problem is that Intel can't get off of x86.

x86 is a poor(relatively) design. New avenues need to be opened up.

There were once two wars in computers.

Cisc / Risc and Microsoft / everybody else

cisc & MSFT won those early battles. And on one had we should thank them for mass producing compuers to the general Population.

On the other hand sometimes I wish the past decade didn't happen and the better technologies came out on top, instead of the cheap ones.

ex DECcies strike again

[Pemdas]

This comes out of AMD's aquisition of Rich Witek's startup (named Alchemy). Rich Witek was one of the original guys working on the Alpha chip (among other projects). Alchemy originally targetted PDA's with their low power MIPS32 processors and on-chip peripheral support.

Interestingly, Dan Dobberpuhl, another Digital alumnus who was influential in the Alpha project, also founded his own company to make MIPS based processers, though for a slightly different target market. That company was SiByte, and was acquired by Broadcom in 2000 or 2001. He has since moved on to start PASemi, which seems to be in the same general business.

Digital may be gone, but it's engineers are still making waves!

Alchemy is old news... like 2002...

[thpr]

...is when AMD purchased Alchemy Semiconductor....

Finally I can plug in two Radeon Y-NOT-8G cards ..

[RedLaggedTeut]

So this means that finally I can plug in two Radeon Y-NOT-8M cards consuming 100 Watts each without buying a new special power supply. So who says there ain't no progress and that history is bound to repeat itself?

My Centrino based laptop goes slow

[Anonymous Coward]

My laptop runs noticeably slower when running on battery power. I can't find any power management options that allow me to burn through the battery as fast as I can work, so everything takes me three times as long to do when on bettery power. Is there something inherent in my "mobile" Centrino chip that reduces the power available when running on betteries?

Re:These are great for $100 PCs

[corngrower]

See Sharp's LH7A404. This puppy includes * an 800x600 LCD controller * A Codec for a modem * A 100 Mbs Ethernet controller * USB ports ( both sides ) * Touch screen controller. * PS/2 Mouse and keyboard port * Serial ports I believe that's everything you asked for, but all on a single chip. Plus its ARM based, so I'm sure that's Knoppix is easily portable.

Re:Mod down that post! Transmeta is the be-any Arc

[Anonymous Coward]

I got an laptop w/ a P4 3.06 Ghz processor in it and it can be up for 13 hours easily, though it needs to be plugged in and proper airflow underneath, and even then the BIOS starts freaking out and speed stepping it down to some ungodly slow speed (even a terminal lags at that speed, it claims '25%' but my 400 Mhz p2 is FASTER than it at that speed.

Next laptop I get I plan on it being a Transmeta laptop, they seem to be friendly towards linux (dell sucks towards everyone, all the power management features other than freak out and kill the mhz don't work in the slightest ANYWHERE). Its gonna be cool trying to find a new laptop in 1 1/2 years.

Now I understand why companies started calling them 'notebooks' and not 'laptops'.

Not actually a joke

[EmbeddedJanitor]

Someone put together an idea of using a bunch of ARM "systems on a DIMM" in a laptop format using clustering. Quite a neat idea... Some benefits include modularity and power saving. Just turn on enough CPUs to support your current processing load and sleep the rest. Just buy the ones you need.

Made the last chips in the Gravis sound card too

[neibwe]

To bad it didn't do very well. I tore out a lot of hair setting that thing up... although those old demos sure were amazing in the older days before these new 3D cards came to power.

http://slashdot.org/article.pl?sid=03/01/09/189218 (Slashdot link to Demo Scene animations on DVD and more. Anyone else remember these things? =)

Recomendations for experimenting?

[captwheeler]

Any recomendations on how to get just one for fun -- montivista for linux, but any hints for the rest (boxen/components/...), what vendor to call, what setup...

Also, a guess at the cost, if you would.

Re:Recomendations for experimenting?

[Weaselmancer]

Not really sure how you'd go about hobbying around with one. We bought 3 devboards (DbAu1100) directly from AMD, and IIRC they were over $1000 each. But that was over a year ago. Might have changed by now.

It's a shame there's not something like the LART StrongArm board for the Alchemy yet. They're really great processors. I had one running an Xorg X server and the IFS screensaver (which takes the CPU to 100%) all day, and it was still cooler than a cup of coffee. Barely even warm, really.

Orion Multisystems! 12 CPUs, 220W

[toby]

Actually low power chips are gaining popularity in super computer environments

Nobody ever seems to mention Orion Multisystems in this context, so I am going to. Visionary product.

Even lower than this..

[adeyadey]

If we can get power consumption figures for these chips lower by, ooh say 1 watt or so, we can use beowulf clusters of these generate power & to solve our power problems - no more greenhouse gases!

low power chips often better tradeoff

[idlake]

Generally, the relationship between compute power and power consumption for a single chip is super-linear. So, for well-parallelizable problems, using more chips that are individually less powerful helps you with overall power consumption.

Re:low power chips often better tradeoff

[adolf]

Bzzt.

If what you say is true in the context of getting things done, I'd think we'd be seeing a -lot- more Slashdot articles about i8086, i80286, and M68000-based supercomputers. They're all low-power, and you can buy a bushel of them for next to nothing.

After all, for well-parallelizable problems, using more chips that are individually less powerful helps you with overall power consumption. No-brainer, that. But it's only one factor in the equation.

I think you've forgotten that the single reason people enjoy pissing away millions on supercomputers is that they'd like to, well, compute. And they'd like to do it fast.

Efficiency rules.

HTH

Re:low power chips often better tradeoff

[idlake]

There is a current technology frontier of power/performance tradeoffs. If you buy chips from that frontier, then for parallelizable problems, over a wide range of tradeoffs (namely, where the curve is convex), you are better off with more low powered chips. If you buy chips below the technology frontier (either because they are poorly designed or because they are obsolete), your parallel super computer ain't going to be so super.

But it's only one factor in the equation.

There are actually two equations:

total-perf = #proc * perf-per-proc
total-power = #proc * power-per-proc

You get to choose different (perf-per-proc,power-per-proc) pairs.

Go work it out yourself; you can do it.

Bzzt.

That must have been the sound of a thought misfiring in your head.

Re:low power chips often better tradeoff

[adolf]

total-perf = #proc * perf-per-proc
total-power = #proc * power-per-proc

Isn't that just a fancy way of saying that more efficient chips are more practical in this context, and thus reinforcing my singular point?

Are you arguing for me, or against me? Do you even know?

I think you all forgot to switch on your sarcasm detectors this morning...

Re:low power chips often better tradeoff

[Anonymous Coward]

Isn't that just a fancy way of saying that more efficient chips are more practical in this context, and thus reinforcing my singular point?

I don't even know what you mean by "efficient". In any case, the two equations by themselves don't tell you anything, you need to plug in the available choices of (perf-per-proc, power-per-proc), and the best available processors at each power level happen to form a convex curve, which gives rise to the claimed result.

I'm not sure how to explain it better than I did in my post, so you'll really have to think about it yourself.

Are you arguing for me, or against me? Do you even know?

I'm simply arguing that using the fastest chips in your cluster does not give you the overall best performance per watt: using more lower powered chips is a better tradeoff if you are concerned with maximizing performance for a given power budget.

Re:low power chips often better tradeoff

[adolf]

Oh, for fuck's sake:

Main Entry: efficiency
Pronunciation: i-'fi-sh&n-sE
Function: noun
Inflected Form(s): plural -cies
1 : the quality or degree of being efficient
2 a : efficient operation b (1) : effective operation as measured by a comparison of production with cost (as in energy, time, and money) (2) : the ratio of the useful energy delivered by a dynamic system to the energy supplied to it

A more efficient CPU is one which does more work for a given amount of power, just as a more efficient automobile engine gets more miles per gallon, and a more efficient lightbulb produces more useful light per unit of energy.

It's just English, and fairly common English at that. What's so hard to understand?

If long-term operating cost is a concern when building a cluster or a supercomputer or a dishwasher or a jetplane or anything else, you'd better be using efficient parts. These are not necessarily low-power parts, but are merely those which produce the most work ("useful energy") for a given amount of energy input.

If there's any other way to look at it, please let Merriam and Webster know so that they might be able to include your views in the next New Collegiate. Thank you.

alchemy processor, eh?

[MyKarmaSucksEggs]

so that's what powers all the metal appendages that Ed and Al use. In fact it powers Ed!

So, How do you overclock them?

[Anonymous Coward]

Will it need liquid cooling then?

Buzzword Bingo! Thanks!

[RalphBinaca]

The new CPUs, collectively named Alchemy, consume less than 1Watt of power.

Thanks! With that Alchemy link I just won the office Privoxy "Buzzword Bingo" contest! ;-)

not too difficult

[KZigurs]

Simple oil radiator consumes about 1500 - 2500W. Your computer - 400-700W (computer, monitor, any peripherals, the broadband box so you can down^H^H^H^H support your important calculations).

So, if the computer is hot enough, it's definetly cheaper than heater :)

Re:not too difficult

[PurpleFloyd]

Actually, the cost is exactly the same, whether you heat your home with computer waste heat or electric heat. 1 kWh costs the same whether it goes into a computer or a heater, and thermodynamics ensures that a computer consuming 400W will eventually output every last bit of that energy; almost all of it is turned directly into heat (the second law, of course, says that it will all eventually end up as heat). You might get more benefit out of putting a kWh into your computer, rather than a heater, but you pay the same amount.

The only "free" heat is reverse-cycle air conditioning, which heats by using an air conditioner to cool the outside air, and even that incurs energy costs to run the compressor and pumps. You most certainly pay for the heat coming out of your computer!

See now if it's me demonstrating these.....

[geofforius]

It's a case of "These processors consume so little energy they don't even need a heatsiiiiii OH MY LORDY IT'S ON FIRE SOMEBODY DO SOMETHING"

Let the AMD fanboy

[Anonymous Coward]

circlejerk and Intel flogging commence.

Heating a Room with a PC

[Ride-My-Rocket]

Actually, it's quite possible -- all you need to do is insulate the room with Aerogel, and you'll be good to go.

Stealing articles, even if online, is ok?

[tier 3 geek]

It's amazing to look at the similarity between what illumina+us wrote and Maximum PC's magazine article in it's March 2005 Issue on page 14. The only difference between the two is the opening sentence, and that the magazine's article came first and had more information.

Re:transmeta? Let's do the math...

[Moekandu]

Orion Multisystems makes "cluster workstations" using the Transmeta Efficeon CPU's. You can get them in 12 CPU and 96 CPU flavors.

The 96 CPU C/W draws a maximum of 1500W of power and you can cram three of these things into a 42U rack, for a maximum of 4500W power draw and 288 CPU's.

The 1.4Ghz Efficion can crunch a theoretical 3GFlops.

288 CPU's x 3 GFlops = 864 GFlops(theo/max)

A 2.0Ghz G5 can cruch a theoretical 6GFlops.

A dual 2.0Ghz 1U XServer draws a maximum of 400W. You can cram 42 of them into a 42U rack for a total of 84 CPU's.

84 CPU's x 6 GFlops = 504 GFlops(theo/max)

Except that you're sucking 16,800W, more than three times the power consumption for 2/3 of the processing power. To get the same amount of crunch power, you would need another 60 CPU's and 12,000W.

4500W, hell, I can run that in my apartment! Assuming, of course, that I unplug my washer and dryer.

28.8KW? Fuhgeddaboutit.

This is why Transmeta kicks ass.

Better than VIA

[tim256]

Via's website says they have the lowest power CPUs on their website. I think they need an update:

AMD Geode(TM) GX 533@1.1W processor
VIA Eden-N 533MHz@2.5 watts

Via, which took over Cyrix, is building low voltage, low power CPUs. Both CPUs support x86 instructions. VIA Eden-N AMD Geode(TM) GX

Re:Better than VIA

[tim256]

Via has more cache: 64 KB L1, 64 KB L2 and AMD's has just 32 KB L1. That probably makes a big difference. Via's also has a cryptography engine and supports SSE instead of 3DNOW. However, I think their website is still wrong.

What's the deal with Canadian Bacon?

[Anonymous Coward]

Come on now, we all know it's just HAM!!!!!

Re:Mod down that post! Transmeta is the be-any Arc

[crimson30]

Next laptop I get I plan on it being a Transmeta laptop, they seem to be friendly towards linux

Just a tip

Re:transmeta? Let's do the math...

[ddustan]

OK... I'm a long time listener first time caller.

I have built a share of Beowulf clusters, exclusively with AMD MP CPUs (in 2001-2002).

Looking back, the main liniting factor for "growing" these clusters was power consumption and heat production of the completed product. Transmeta breaks through these two limitations.

I worked for a company that put together a 48 node cluster, including the head node. These were dual AMD MP processor systems so we had 96 processors total. In order to build this cluster and test it we had to run it between 3 power sources, Two separated services in the building, and a generator "out back." Our service panel was extremely hot to the touch, to the point that there was much concern that we were going to burn the workplace down.

Moekandu has done the math and I concur. Power is the most limiting factor for larger clusters. *Anyone*, hobbyists, small businesses, and even corporate IT departments will benefit greatly from being able to produce more CPU cycles from far less power!

Heat dispensed by individual CPUs is another factor that seems to go hand-in-hand with power consumption.

Space is another factor that can be considered, I guess... I had 96 processors in 48 2U rack mounted pizza boxes. This neat little project fit into 3 packed racks including 3 KVM switches and two 48 port switches. Now I see that in that space with the Orion MultiSystems option, I can pack in 96 nodes in a fraction of one cabinet (1/3) and I have more room for much more computing fun. Also, no more need to keep a fire extinguisher on hand...

The Transmeta products show themselves to be an option that is sure to be considered more and more in the next year. It is truly in a different class and opens up many opportunities.

David

You too?

[Grendel Drago]

I have two Athlon T-birds, a 750 and a 1000. My room requires an air conditioner well into the fall, and even though the AC is still in the wall, so head leaks out constantly, the room is only cool when the outside temperature drops well below freezing.

And I don't even have a CRT, just a nineteen-inch flat-panel which doesn't put out that much heat.

--grendel drago

Congrats are in order here

[Simonetta]

Congratulations to the engineering team that produced such powerful processors with low electrical power requirements.

Now let's take a brief microsecond to consider the impact that this technology will have on our lives. And specifically what types of conflicts will it create.

In very general terms, every advance in digital technology increases the level of conflict between the people who 'own' ideas, recordings, works-of-art, and other ethereal cultural commodities that can be now more easily distributed by the new digital technology.
Creating a machine (a microprocessor in this case) that removes video processing from a fixed location and makes it handheld portable will invoke a backlash by the content owners. And this backlash will be a renewed and more focused effort to get laws passed to detain, fine, and imprison people for viewing, manipulating, and distributing video content that is primarily, at the present time, pre-recorded and copyrighted 'property' of the media corporations.

In another application, I see these handhelds being used the lower the cost of advanced language translators by at least an order of magnitude.
I can see governments in 10-20 years claiming that they own their national language and either attempting to ban these translators outright, or demanding a huge tax on them 'in the interests of preserving the national government's control of the nation's language and culture'. I can see places like Singapore banning handheld language translators when schoolchildren start refusing to memorize how to write thousands of Chinese characters when the handheld translator can do the interpretation of the written character into the spoken word (and vice versa).

In an era of rapid technological change such as this one, it is important to take a little time to at least try to predict the disruptive impact that any new type of technology will have.

Typo!

[ModemJunki]

It's the simple things... the header for the product page indicates "AMD Connectivity Soultions..." I wonder how long it will take for them to fix..

Re:From the Product Briefs linked to in the articl

[Kjella]

MPEG4 advanced simple profile/level 5 (720x480, 8Mbps, 30fps)

Is it just me, or does that seem awfully high? Your average divx rip is around 1Mbps, 8Mbps should be enough to carry a HDTV stream. (1920x1080 being 6x 720x480).

Kjella

This was bound to happen

[melted]

Just three days ago, I bought this VIA C3 based system (processor dissipates 7W at idle), and they roll out sub 1W chips. Oh, well. Hard drive is gonna take 9W of power anyway.

Comparing chips.

[Grendel Drago]

I think he was comparing high-power chips from 2005 to low-power chips from 2005, not to low-power chips from 1981.

--grendel drago

Without a heatsink? No way!!

[mjh49746]

I tried a similar attempt after reading your post, and there's no way I would trust my system to run like that. Then again, I have a NewCastle 3500+. Anyways, I've underclocked this sucker all the way down to 800MHz, adjusted the juice down to 0.9Vcore, unplugged the fan, left the heatsink on, and entered the hardware monitor through the BIOS just to watch the temperature slowly run away. Once it got close to 55c after less than about 30 seconds, I killed the power and plugged the CPU fan back in. So nope, I can't do that with my setup.

Re:Without a heatsink? No way!!

[Ized]

Sorry, I meant to say - without fan :D.. Glad you didn't fry your cpu.

Also Newcastle core is different than Winchester. Winchester's default core is 1.4v so it's able to run on lower voltages, while maintaining ok speed.