backtop

Hardware VIA Announces 1-Watt x86 CPU
Anh Tuan Huynh (Blog) - August 23, 2007 11:13 AM
Print E-mail del.icio.us 56 comment(s) - last by tcsenter.. on Aug 27 at 8:52 PM

VIA Eden ULV processor  (Source: VIA)
New 500 MHz VIA Eden ULV is the most power efficient x86 processor in the world

VIA today announced the world’s most power efficient x86 processor, in a mere 1-Watt envelope, the 500 MHz VIA Eden ULV processor. The new 1-watt processor operates at 500 MHz and optimized for fan-less operation. At idle, the 500 MHz VIA Eden ULV processor only requires 0.1-watt of power.

VIA targets the 500 MHz VIA Eden ULV processor for industrial, commercial and ultra mobile applications. The 500 MHz VIA Eden ULV is a small processor, measuring in at 21 mm x 21 mm and packaged in a NanoBGA2 package.

The processor operates on VIA’s V4 bus clocked at 400 MHz. It has 16-stage pipeline with 128KB of full-speed L2 cache. VIA also integrates the PadLock Security Engine for hardware encryption algorithms such as AES, SHA-1, SHA-256 and Montgomery Multiplier.

“Giving our customers the building blocks to create innovative systems and driving PC technology into new markets defines our ‘Small is Beautiful’ strategy,” said Richard Brown, Vice President of Corporate Marketing, VIA Technologies, Inc. “With its performance, energy efficiency and compact size, our new VIA Eden ULV processor provides a way for embedded developers to add real value to their systems and push the market forward.”

VIA recommends pairing the 500 MHz VIA Eden ULV with the CX700/M chipset for a system that only draws 3.5 Watts, at a maximum, storage devices not included. VIA expects a system powered by the 500 MHz VIA Eden ULV and CX700/M with storage and connectivity options to consume no more than 10-watts of power.

Expect the 500 MHz VIA Eden ULV processor to show up in embedded systems.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
Cool
By jak3676 on 8/23/2007 12:04:42 PM , Rating: 5
As much as I don't like a lot of the x86 baggage, It'll be nice when there's a single architecture for a large range of items. Something like this will be great for PDA's and smart phones, etc. Definitely makes my life easier as a software developer.




RE: Cool
By grenableu on 8/23/2007 12:22:37 PM , Rating: 3
that cpu would make a great start for my next home file server.


RE: Cool
By mindless1 on 8/23/2007 5:40:51 PM , Rating: 3
Not really, the performance is too low for peak gigabit throughput and you'd be fairly limited in having to use a hardware raid if you wanted redundancy (or again, a performance bottleneck).

Via has several other processors that are more appropriate for a desktop system if you merely want the processor passively cooled, let alone the option to underclock one of those to your required performance level instead of having to try to overclock this to get there (and if history is any indicator, it won't overclock very well).


RE: Cool
By MrFluffo on 8/23/2007 12:26:39 PM , Rating: 2
Having a processor like this is any type of mobile device would be nice due to its low power usage. I wonder what the battery life would be on simple products such as mp3 players etc.


RE: Cool
By TomZ on 8/23/2007 12:31:31 PM , Rating: 4
The article states that systems based on this chip would take 10W. That would use up the battery in an MP3 player in a few minutes. :o)

Seriously though, this processor is not designed for that type of application. Think of bigger devices.


RE: Cool
By mindless1 on 8/23/2007 5:42:15 PM , Rating: 2
Agreed, if nothing else the packaging (carrier) is too large.


RE: Cool
By Torched on 8/27/2007 10:54:36 AM , Rating: 2
The nanoBGA2(21mm x 21mm) is very tiny for a x86 CPU if you ask me. The whole eden-N line uses the same packaging.


RE: Cool
By Torched on 8/27/2007 10:48:52 AM , Rating: 2
This processor is aimed to compete with the already ~1W AMD Geode LX800 SOC. These chips are used in ATM, POS, & other industrial applications. My company has used the LX800 for our products for the past year now. The nearest competitors for the LX800 were the VIA Luke & Mark dual die SOC & ULV Celeron M. Even though this new chip is not a SOC it is low enough power to compete.


RE: Cool
By tcsenter on 8/27/2007 8:52:21 PM , Rating: 2
quote:
These chips are used in ATM, POS, & other industrial applications. My company has used the LX800 for our products for the past year now.
Precisely. These aren't mobile battery-powered devices, but near always plugged.

21mm x 21mm may be tiny for an x86 processor, but its not nearly efficient enough for the kind of mobile battery-powered applications some have suggested as suitable for Eden ULV.

Eden ULV would be great for numerous plugged machines; kiosk PCs (internet, ticketing and reservation, electronic catalog and ordering), ATMs, thin clients, small industrial PCs, et. al.


RE: Cool
By TomZ on 8/23/07, Rating: 0
RE: Cool
By grenableu on 8/23/2007 12:32:31 PM , Rating: 2
Um, "X86 baggage" has nothing to do with USB ports. Its the instruction set of the cpu. And sure you can do WinCE development under ARM, but you still have to rewrite everything else. X86 compatibility is an extremely important feature for certain embedded markets.


RE: Cool
By psychobriggsy on 8/23/2007 12:36:52 PM , Rating: 2
Anything requiring hand-written assembly would have to be written especially for the x86 (SSE3/4 etc) or the ARM (whatever co-processor is attached), except that the 1W x86 variant will probably not implement the features, or they'll be really slow compared to the co-processor on a 0.2W ARM SoC.

Sure, embedded markets can use x86, but mobile ones can't, and won't for a while. ARM's performance is ever increasing too - ARM Cortex for example whilst keeping power use low. Industrial apps like x86, especially no-moving-parts x86, so this will be great for them (no fans, flash hard drive, ...).


RE: Cool
By TomZ on 8/23/2007 12:47:44 PM , Rating: 1
quote:
Um, "X86 baggage" has nothing to do with USB ports.

It depends on what you're talking about. For example, the system described in this article has a chipset that helps it implement and support PC-compatible hardware.

Sure, you could run an X86 stand-alone without a PC chipset, but what would be the point of that? You might as well run a smaller, cheaper, more power efficient ARM instead.


RE: Cool
By grenableu on 8/23/2007 12:53:03 PM , Rating: 3
The "recommended chipset" here is just that. A recommendation. There are other chipsets, with other featurs.

ARM and X86 are different markets plain and simple. Neither one is 'better' than the other, they're each better suited for certain applications.


RE: Cool
By DallasTexas on 8/23/2007 1:48:26 PM , Rating: 2
As usual, you will be proven wrong. X86 moving down to smartphone and similar products is going to happen faster than you think. It will first happen with UMPC's and MID's as these devices feature larger displays (5-7inch diagonal) and will tolerate the higher energy penalty of x86 vs ARM.

In one years time or so this discrepancy in energy will be irrelevant at the platform level. The benefits of x86 as the architecture of choice for Web 2.0 development is very important for smarthone & UMPC's moving forward.

The customization of applications like YouTube for things like iPhone is a fantasy and a short term "fix" to run internet and other social networking apps.

And as someone correctly said, the peripheral set has ZERO to do with x86. Even so, the PC peripheral set is hardly the "baggage" it once arguably could have been.


RE: Cool
By TomZ on 8/23/2007 2:01:15 PM , Rating: 5
I've heard the same argument for 5 years already, and it still hasn't happened. The fact of the matter is that ARM and other low-power architectures are also moving forward. So both become more efficient, but at any point in time, if you compare the two for a particular product design, ARM is always more efficient.

The only reason to have mobile X86 is to run Windows XP or Vista. For smaller devices where the OS choice is less critical (e.g., PDAs and smartphones), ARM seems to get all those design wins.
quote:
As usual, you will be proven wrong.

I'll try not to take that personally. :o)


RE: Cool
By omnicronx on 8/23/2007 2:10:16 PM , Rating: 3
You keep count of the times that he 'will' be wrong? Do you have some 'you will be wrong' software to keep track? Written for an x86 smartphone maybe?.. without the baggage of course..


RE: Cool
By DallasTexas on 8/24/2007 9:24:21 AM , Rating: 2
Yes, I do. I merely click on the persons name and out comes a record of their posts from DailyTech servers. I import it all to a spreadsheet, pick out the outrageous ones and put on the big screen for a good laugh with friends.


RE: Cool
By mindless1 on 8/23/2007 5:53:05 PM , Rating: 1
This "discrepancy in energy" is going the wrong direction. We don't need x86 to have functionality and who really cares if a developer actually has to do some work instead of the easy bloated way out?

This is the kind of thinking that is ruining consumer devices, trying to make them all do more than the buyer needs, thus leaving the owner with no way to get an acceptible runtime per battery charge and unnecessarily higher expense, and probably shorter product lifespan as well.


RE: Cool
By tcsenter on 8/24/2007 5:24:19 AM , Rating: 2
quote:
As much as I don't like a lot of the x86 baggage, It'll be nice when there's a single architecture for a large range of items. Something like this will be great for PDA's and smart phones, etc. Definitely makes my life easier as a software developer.
Not quite ideal for PDAs or smart phones, at 440mm² for the CPU alone. The same 21mm x 21mm footprint will get you a complete ARM/PPC System on Chip (SoC) that integrates a RISC CPU running at 400MHz ~ 533MHz, 128KB L2 cache (full-speed), hardware encryption engine, and a complete core logic solution including 2D/3D graphics accelerator, all for not more than 500mW MAX TDP (~250mW is typical).

Check-out the Freescale i.MX31 SoC, which brings all of this (and more) on an even smaller 14mm x 14mm BGA package:

http://www.freescale.com/files/32bit/doc/prod_brie...

http://www.freescale.com/files/32bit/doc/fact_shee...

That chip is iPhone material and only costs $21 in 10K quantities. Honestly, I don't know what VIA's market is here other than small budget prototyping with off-the-shelf components, custom integration, and hobbyist DIY. Real developers of set-top boxes, PDAs, gaming portables, video entertainment, navigation systems, and the like have access to vastly superior and less expensive embedded components that would have a fraction of the power requirements.


Heat
By dm0r on 8/23/2007 12:07:58 PM , Rating: 2
Wow, only 1 Watt...I wonder if it requires a heatsink




RE: Heat
By GoodBytes on 8/23/2007 12:24:11 PM , Rating: 1
I don't think so, as the article mentions it won't require one.

Maybe they can put a basic tiny heat sink, like on some sound cards, for the "just in case" situation.


RE: Heat
By TomZ on 8/23/2007 12:29:15 PM , Rating: 4
The article states that it doesn't require a fan; it doesn't say whether a heatsink is required. In any case, the choice of whether or not to use a heatsink depends on the overall system (product) thermal design. Probably some designs will require a heatsink while others won't.


RE: Heat
By omnicronx on 8/23/2007 1:33:03 PM , Rating: 1
but just thinking in perspective, i86-i486 cpus did not require a heat sync or fan, and they were definitely rated at more than 1w, so you would think theres almost no reason for a heat sync on a 1w chip, what heat is there to dissipate from 1 watt.


RE: Heat
By TomZ on 8/23/2007 1:52:54 PM , Rating: 4
To figure this out, you have to multiply the 1W times the chip's thermal resistance to ambient, which is in ºC/W - that gives you the temperature rise of the die. Then, take the maximum internal temperature of your product (which is higher than its overall maximum temperature rating) and add this delta, and compare it to the maximum rating for the chip.

So you see, there are a number of variables that enter into it. In the case of 486, you had PCs that were basically designed to be run at room temperature (maybe 40-50°C max). This chip is designed for embedded applications that might be +85°C or +125°C, so you can see there is some difference.

BTW, the 486 @ 100MHz has a TDP of about 4W. ftp://download.intel.com/design/intarch/datashts/2...


RE: Heat
By peldor on 8/23/2007 4:37:25 PM , Rating: 2
You've also got to consider that the die for this Eden chip is a mere 30 mm^2 where a 486 die is much larger (165 or 84 mm^2 depending on process AFAICT).


RE: Heat
By mindless1 on 8/23/2007 5:57:26 PM , Rating: 2
Agreed, it's the thermal density that matters, and the packaging. The question is how well the BGA 'sinks away heat, as being BGA could be either a virtue or a liability here.

Frankly I see no reason not to use a heatsink if it fits in the available space. A tiny circa-'90 northbridge style 'sink can't help but cool this flipchip far better than the BGA could (if designed to).


RE: Heat
By Anonymous Freak on 8/23/2007 6:56:56 PM , Rating: 2
And the 100 MHz 486 almost always used a small heatsink. Even my 66 MHz 486 had a small heatsink on it. (Smaller than the Northbridge heatsink on my current motherboard.)

386 and earlier generally had no heatsink.

But, one big difference is that those chips had a full packaging. The die wasn't exposed to the air, so the packaging would help dissipate the heat. This has its die exposed directly, so it may need some small heatsink to help in its cooling. (An Intel-style heatspreader would probably be enough.)


RE: Heat
By psychobriggsy on 8/23/2007 12:33:04 PM , Rating: 2
Well the die is pretty small, so maybe a small heat spreader would be useful, but 1W isn't much energy to dissipate, even for that size die.

Given the die is so small, why don't VIA integrate a full (for the application) northbridge onto the die as well, and save some space, and possibly improve overall power use?


RE: Heat
By mindless1 on 8/23/2007 6:00:09 PM , Rating: 2
Generally with a TO220 sized package, it's assumed up to 1W can be acceptible if the ambient conditions are right - this with a metal tab on it that is larger than the core on this processor. Put a heatsink on that TO220 device @ 1W and the 'sink will feel pretty hot unless it's fairly large. Since many processors seem instable past about 70-80C, it wouldn't be surprising to find it needs a small 'sink in many applications.


RE: Heat
By pnyffeler on 8/23/07, Rating: -1
RE: Heat
By TomZ on 8/23/2007 12:52:18 PM , Rating: 5
No, requiring one watt to run is the same as producing one watt of heat, in the case of processors. Practically all the power you dump into the processor is dissipated as heat.


RE: Heat
By Dactyl on 8/23/2007 2:23:39 PM , Rating: 3
Step 1: Steal Underwear
Step 2: For every 1W in, get 90W out
Step 3: Profit!


RE: Heat
By retrospooty on 8/23/2007 4:50:32 PM , Rating: 4
"I wonder if it requires a heatsink"

It will if we are to overclock it !


RE: Heat
By wordsworm on 8/25/2007 8:27:31 AM , Rating: 2
quote:
"I wonder if it requires a heatsink"
It will if we are to overclock it !

Damn, you beat me to it!


By nothingtoseehere on 8/23/2007 3:03:27 PM , Rating: 2
That's kind of a bold statement, especially when AMD has had the LX800, also a 500MHz processor, for quite a while now...

The power specs of the LX800 and this new Eden ULV are very close. Benchmarks are needed to prove which is more power efficient...




By omnicronx on 8/23/2007 3:49:46 PM , Rating: 2
quote:
AMD Geode LX 700@0.8W consumes a maximum power of 3.1W (TDP)


From how i read it, the Eden runs at 0.1w with a maximum power of 1W


By nothingtoseehere on 8/23/2007 3:58:27 PM , Rating: 2
I saw a 3.5W number for that eden too. They are both not very clear about the details what is included and what is not (at those power levels, the chipset power matters).


By omnicronx on 8/23/2007 4:10:25 PM , Rating: 2
quote:
VIA recommends pairing the 500 MHz VIA Eden ULV with the CX700/M chipset for a system that only draws 3.5 Watts, at a maximum, storage devices not included.

3.5w includes the chipset


By nothingtoseehere on 8/23/2007 4:18:56 PM , Rating: 2
In the press release, the 3.5W is just the chipset, without the CPU...


meh
By Treckin on 8/23/2007 6:23:50 PM , Rating: 1
people on DT tend to get lost in their own arbitrary arguments, and loose sight of the broader scope of the article, almost always in favor of faggoty nitpicking of some typo or discrepancy or fanboi BS.

I could see very useful apps for this... think AA battery powered universal remote w/ lcd color screen that DOESNT suck ass...

One thing that VIA should do is shrink the carrier... the core is tiny, its a shame to see its uses limited by the physical size of the pin anchors...




RE: meh
By TomZ on 8/23/2007 6:34:19 PM , Rating: 3
How long do you think you could run this 10W system off of a AA?

Also, how big is your remote going to be?

LOL, you might want to read the comments a tiny bit more. You seemed to have read them enough to be critical, but not enough to have picked up any potentially useful information.

In short, this chip is not designed for the kind of application you're describing.


RE: meh
By tastyratz on 8/23/07, Rating: -1
RE: meh
By TomZ on 8/24/2007 5:54:54 PM , Rating: 2
1. So what are you going to connect that 21x21mm chip to? You'll probably want some RAM? You'll probably also want some flash, too, right? Probably some IO, maybe a power regulator, probably some crystals? Seems like your "remote" just grew quite a bit.

2. Yes, the typical remote control is 2xAA's. Yes, you may not require 10W for your remote control. Let's say 2W just to have some figure. OK... (punches some buttons on calculator)... from fresh batteries to dead batteries, that's about 1 hour.

Does that seem viable to you?

I appreciate your defending the OP, but really this part is not designed for small, mobile applications. It's physically too large and consumes too much power, as I've said in my other posts. I've shown you a couple of reasons why, I hope you can appreciate that.


RE: meh
By psychobriggsy on 8/23/2007 7:54:44 PM , Rating: 2
That's a task an ARM cpu drawing a tenth of the power would excel at.

Look at the 480x320 user interface on the iPhone to see what can be done in a fraction of what the x86 + support chips could do.

Indeed there was more than adequate software for old Palm handhelds for universal remote control, and they were running on 20MHz 68k derivatives (about the equivalent of a 10MHz ARM if that).


Left out of the story
By peldor on 8/23/2007 12:39:29 PM , Rating: 2
This is actually just a really-low-end release of the Eden ULV.

In January 2006 they launched a 1GHz/3.5W version and a 1.5 GHz 7W version of the same CPU.




RE: Left out of the story
By mindless1 on 8/23/2007 6:04:26 PM , Rating: 3
Good point, people are acting like this opens up new opportunities in devices that aren't limited to 1W heat nor trying to run from battery power.

Everyone reading this article right now could've been running a Via processor but they weren't. Why? Their pimped up Core2Duo has multiple times as much performance. So it'll be with this new Via CPU, devices that would run it would be benchmarked and found to be low performance = bad to many people based on past history of performance-at-all-costs mindset so commonly seen.


potential
By omnicronx on 8/23/2007 12:46:16 PM , Rating: 2
Does anyone else see the potential of this in small devices? PSP that actually has good battery life anyone?




RE: potential
By tastyratz on 8/23/07, Rating: 0
RE: potential
By tcsenter on 8/25/2007 10:17:59 PM , Rating: 2
quote:
Does anyone else see the potential of this in small devices? PSP that actually has good battery life anyone?
Sure, if by 'good' battery life, you actually mean reducing it by about 85%.

Sony's custom PSP system processor has a MAX TDP of approx. 500mW. This includes dual MIPS R4000 32-bit cores, a 2D/3D graphics processor with hardware H.264 (AVC) decode, programmable DSP, 4MB eDRAM, mobile DDR IF, USB 2.0 and Memory Stick controller...all in one chip that is smaller than VIA's Eden ULV @ 17mm x 17mm.

I don't think some people realize how HUGE the VIA Eden ULV is at 21mm x 21mm, at least for embedded applications, given that it doesn't bring anything but the processor. Companies have been cramming entire systems on a chip this size with 1/2 to 1/5 the power consumption for years now.


pardon my ignorance but...
By thartist on 8/23/2007 5:26:38 PM , Rating: 2
what could be about making a 2W 1000Mhz processor? or a dual-Eden?

wouldn't that be fantastic, and at the same time trendy? ;)




RE: pardon my ignorance but...
By mindless1 on 8/23/2007 6:06:32 PM , Rating: 2
It won't happen given current tech, to ramp up to 1GHz will require higher voltage, exponentially more power/heat. If you want 1GHz version you can already buy the equivalent Via processor today.


500Mhz... or 400?
By Screwballl on 8/23/2007 12:30:17 PM , Rating: 1
how many times can you say 500MHz in a single story?

I counted 3 in the first paragraph and 8 in the entire story... plus one mention of 400 instead of 500MHz...

Rather than saying "the 500 MHz VIA Eden ULV" 8 times, why not just say "the Eden ULV" or "the Eden"

As a direct comment on the story, it will be good to see this used in some handheld devices that can actually make use of it...




RE: 500Mhz... or 400?
By Screwballl on 8/23/2007 12:33:15 PM , Rating: 1
correction.... I see they said 400MHz bus speed, not cpu speed

my bad, sorry


By Roy2001 on 8/23/2007 12:44:22 PM , Rating: 2
Via never did fundamental change to that chip.




Perf per watt
By Durrr on 8/23/2007 6:39:27 PM , Rating: 2
How would 120 of these stack up against a X5355? The Cooling reqs would be close to the same, and usage might be a bit higher, but would be offset by non-use of FB-DIMMs.




"The whole principle [of censorship] is wrong. It's like demanding that grown men live on skim milk because the baby can't have steak." -- Robert Heinlein









botimage
Copyright 2009 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki