First Menlow Board Released

nerdyH writes "German board vendor Lippert has unveiled what it claims to be the first motherboard based on Intel's 'Menlow' chipset for ultra-mobile PCs. The CoreExpress-Menlow is smaller than a credit card, yet clocks to 1.5GHz, has 1GB of RAM soldered onboard, has multiple PCI Express lanes, USB 2.0, HD audio, an IDE interface, and a digital LVDS video interface. The board is the first in a proposed 'CoreExpress' standard motherboard form-factor measuring 2.6 by 2.3 inches (65 x 58 mm)."

Comments

I for one...

[sykopomp]

...kinda like the idea of not having a massive box on my desk. This is more than welcome.. small form factor is nice.

Re:I for one...

[Jeff DeMaagd]

There are more compact desktops. Some even use notebook computers as a desktop.

Re:I for one...

[sumdumass]

This would be more like using a palm trio as a desktop. 2.5 by 2.3 inches is just a little bigger then your first three finger on one hand side by side and just a little longer then the width. Depending on the size of your hands that is. But it beats most laptop sixes with a stick then laughs at is. Imagine a slimline DVD burner like one you would find in a laptop. Now imagine a desktop powerful computer using on of those and about the total size of a DVD burner you would put in a 5 1/2 inch drive bay minus the video. That is now or soon to be possible.

Re:I for one...

[khellendros1984]

You mean...just slightly smaller than the Wii sitting on my desk =)

Re:I for one...

[sumdumass]

Well, yea. Except they need a way to power it. But it could be powererd externaly with a laptop like charger. Run a USB cable to a hardrive drive and plug the monitor in and your good to go.

And it's probably going to be damn expensive

[Enleth]

Which is a pity - it would be perfect for a SLAM-capable robot project I've got on hold since half a year because even the crappiest embedded motherboards out there are damn expensive when you want to buy just one or two of them and are a student...

Re:And it's probably going to be damn expensive

[TheRaven64]

It's probably slow too. The Ars Technica analysis implied that it would be slower, clock-for-clock, and use more power than most ARM variants. It's a shame, considering that ARM is the most widely deployed architecture that there aren't many boards that are cheap in lots smaller than a thousand. Something like this would be really nice to play with.

Re:And it's probably going to be damn expensive

[Enleth]

But it would be a lot more than enough to handle the data analysis and mapping, especially with this amount of memory. AVRs, which I'm using now, have their limits, not in the clock speed (16MHz IS a lot for controlling quite a bit of a relatively simple robot's subsystems, a few times that is enough for the whole thing) but in the memory, a whopping 4-16KB of which is build into a typical one. Just no room for any serious data processing - and chasing the robot with a laptop and an RS232 cable attached to it is no fun.

Re:And it's probably going to be damn expensive

[TheRaven64]

It uses Silverthorne CPU. Wikipedia says Silverthorne's performance @2 GHz is equivalent to a first-generation Pentium M Don't trust Wikipedia for this kind of thing unless it includes citations. The Pentium-M was out-of-order, the Silverthorne is in-order (unlike any other Intel chip since the 486). Take a look at Ars' coverage for more information. At 2W, it's competing in the same performance bracket as the ARM Cortex A9 at 250mW. Don't treat an XScale as representative of ARM chips - Intel's versions always had an appalling IPC compared to other ARM variants (400MHz chips from other manufacturers regularly outperformed the 625MHz Intel version). Also, comparing MIPS across architectures is completely meaningless, but for reference the A9 gets 4 instructions per clock, giving 4,000 MIPS at 1GHz and supports multicore configurations. Most ARM licensees put other interesting things on the die as well (TI, for example, put very nice DSPs on their ARM variants).

Re:how much does one cost, ya think?

[neomunk]

Software guys can become half-assed circuit guys pretty easily with modern microcontrollers. Learn a couple of (very) simple circuits, hell, you don't even have to learn to solder with the PIC kits available today. After you do the monkey's job of putting part A into slot A everything else turns into a software problem... Well, everything AFTER you figure out how to turn a hit on the dartboard into a properly formatted digital signal. At least when looking for a PIC kit you should be able to find one with a switch demultiplexer or something.

Anyways, I think this is serious overkill for what you're talking about, so much so that I'd imagine it would be MORE complicated to use one of these as opposed to a PIC kit, but that's just my opinion.

Applications barrier to entry

[tepples]

The Ars Technica analysis implied that it would be slower, clock-for-clock, and use more power than most ARM variants. I think the point of this board that is that proprietary applications for Windows Mobile, which are compiled for ARM, tend to have less functionality than the corresponding proprietary apps for Windows XP, which are compiled for i686. People want full-size versions of familiar apps on a pocket-size device, and only a processor with the i686 instruction set can deliver this.

Re:Applications barrier to entry

[poot_rootbeer]

I think the point of this board that is that proprietary applications for Windows Mobile, which are compiled for ARM, tend to have less functionality than the corresponding proprietary apps for Windows XP, which are compiled for i686. People want full-size versions of familiar apps on a pocket-size device, and only a processor with the i686 instruction set can deliver this.

Riiiiiiiiiight.

It was very smart of Intel's engineers to design and implement the CABPWAPF (Clear 'Artificial Barrier to Proprietary Windows App Performance' Flag) instruction that's in the i686 instruction set.

Re:Applications barrier to entry

[Hal_Porter]

It's not about instruction sets per se. The CE version of Office was built from scratch (as far as I know) for CE back when a CE device had a few megs of memory, a tiny screen and a slow processor. It's not the same codebase as the desktop Office and it is much more limited. And OpenOffice doesn't work on CE either. So the fact is that an x86 device in a handheld does make sense because you get run desktop apps. Now there's no reasom why you couldn't run Desktop MS Office or OpenOffice on a handheld these days, but unfortunately neither is ported to CE. Oddly enough, MS Office does run on Macs and did back when they had both a different processor and a different OS.

Re:Applications barrier to entry

[wild_berry]

People want full-size versions of familiar apps on a pocket-size device, and only a processor with the i686 instruction set can deliver this.

I hear of a new technology that Von Neumann and Turing are playing around with, called the 'universal computing machine'. Maybe it'll offer the ability to algorithmize and implement computing tasks independent of the engine performing those tasks. I hear that a group of compiler engineers hope to use this 'universal computing machine' to create something called a 'cross-architecture compiler', allowing one to compile source code on a machine of one architecture to run on a machine of a different architecture.

I could be completely wrong in the potential outcome of these 'new' technologies.

(Once you've got over the sarcasm in that -- BTW: The prospect of 'years of x86 software freely available to a portable x86 device' doesn't overcome the problem that full-sized programs don't have good interfaces on small handheld devices. The interface problem is why the iPod is so succesful. I'll accept the argument that x86/Windows programs never had nice interfaces -- while true this is an unrelated point. Finally, we haven't addressed the issue that x86 is typically focused upon raw-power computing, where ARM is focused on reasonable performance for exceptional power efficiency.)

Re:Applications barrier to entry

[try_anything]

Your point is valid, but I think the overriding risk with ARM is that a full version of Windows might never ship on ARM, and if it did, there might not be many applications available. If it's a lot of work to get Windows Vista (or XP) and normal Windows apps running on ARM, Microsoft might prefer to write off ARM entirely rather than overhaul its entire codebase. Microsoft has already decided once before that it's cheaper to stay with x86-compatible architectures, even in areas where x86 offers inferior price/power/performance tradeoffs. If they can afford to stick with x86 and still be competitive, they will do it. Microsoft and x86 have traditionally survived together while technologically superior solutions have fallen by the wayside.

God does my RAZR ever look old and dowdy

[spazdor]

So, does it run Android?

Re:put a screen with it...

[Gideon Fubar]

screw that.. i wanna attach a 3g module and use it as a phone. ;)

Re:how much does one cost, ya think?

[spazdor]

You could probably accomplish all this with an HC-11 microcontroller. No need to throw anything nearly that fast at it.

Power

[corychristison]

Well, the article is skimping on details, but does anybody know where I can find more info on the power consumption? I doubt it's anywhere comparable to ARM... speaking of ARM. Anybody know where I can get a small ARM based board? I've been searching and searching, can't seem to find anything that isn't mass-order.

Re:Power

[lenski]

There are several inexpensive ARM boards available in small quantities, look in any recent Linux Journal. I am currently running several systems built by Gumstix which work well for me and for my customers.

I have no financial relationship with Gumstix. I have several of their systems. Cheap, effective, and supported by an exceptionally active, friendly and supportive mailing list.

Your post doesn't describe required specs or desired features, so it's not easy to know what you're looking for.

Re:Power

[topham]

buy an iPod touch.

or; buy my Axim x50v

Re:Power

[RingDev]

There's always VIA and the Nano-ITX systems. They are slightly larger than this (10cm x 10cm) but they are available, inexpensive, well tested, and have all the stats published. Check out http://www.mini-itx.com/ for more info and cool cases.

-Rick

Re:Power

[corychristison]

... they use much more than a small ARM based system. The best ITX board I can find is the AMD based ones which are about 5W.

Arm is generally under 1W.

Emulation costs power

[tepples]

I think the point of this board that is that proprietary applications for Windows Mobile, which are compiled for ARM, tend to have less functionality than the corresponding proprietary apps for Windows XP, which are compiled for i686. People want full-size versions of familiar apps on a pocket-size device, and only a processor with the i686 instruction set can deliver this. Riiiiiiiiiight.

It was very smart of Intel's engineers to design and implement the CABPWAPF (Clear 'Artificial Barrier to Proprietary Windows App Performance' Flag) instruction that's in the i686 instruction set. This artificial barrier is the i686 instruction set itself. ARM CPUs cannot run applications that have been compiled with x86 instructions without an x86 emulator, and I don't see how an emulator would outperform this native x86 board and use less power.

Re:put a screen with it...

[mikael]

It has a USB port interface, and you can get USB wireless modems. Maybe a micro-short USB cable would help?

Re:But does it have crypto instructions?

[JSBiff]

I gotta agree here. I'm very puzzled why Via is the only CPU maker that sees the value in adding hardware acceleration for crypto into their CPUs. On the mobile side of things, pretty much every business and government organization is mandating encryption on laptops. On the Server side of things, a lot of network communication is being encrypted, and it wouldn't surprise me if companies have decided there is value in encrypting data even in a 'secure' data center. Plus, there's stuff like encrypting data to backup tapes. On the desktop side of things, again a lot of network communications are being encrypted.

It just makes *so much* sense to put some crypto hardware on CPU's. Yet via has a corner on that market. So, without the support in hardware, that means that my laptop, server, or desktop is going to run just a little bit slower. Sure, most CPU's are fast enough to do software-based encryption pretty fast these day, but it would still be better to do it with optimized hardware. (And, I bet the optimized crypto hardware uses less power to do it than doing it in software on the general purpose CPU).

Re:But does it have crypto instructions?

[Eric Smith]

Perhaps because being based in Taiwan, they don't have problems dealing with the idiotic US crypto export regulations the way AMD and Intel do.

Re:But does it have crypto instructions?

[Score Whore]

I'm very puzzled why Via is the only CPU maker that sees the value in adding hardware acceleration for crypto into their CPUs.

Can you expand on what you think the added value is? In this form factor? Seriously. For a processor that is intended to power servers (ssl or vpn or the like) then sure, there is value. And Sun has added it to the Niagra line of processors. But for some little doodad that is itty bitty. What exactly are you planning on doing on this thing that it's basic math primitives won't be more than sufficient?

Re:But does it have crypto instructions?

[arivanov]

You are wrong. Intel has it on the future roadmap for its ultramobile offerings.

As people say - imitation is the sincerest form of flattery.

You missed one more factor. One of the biggest battery eaters for me in a laptop is the VPN. On via it costs me nothing. On an Intel my battery life is often 20% less as a result (AESing at 256 access to a 4.5G+ IMAP store is expensive).

Re:But does it have crypto instructions?

[TheRaven64]

In a mobile device, encryption is very important. If you have any important data on your local disk, you are going to want to encrypt it - ideally encrypting the entire volume, which means you need encryption for any I/O, including swapping. You also need encryption for pretty much any WiFi usage and for a lot of general network stuff (e.g. IMAPS, SMTPS).

Re:But does it have crypto instructions?

[Score Whore]

I understand the need for encryption, but please explain why dedicated primitives are relevant?

Menlow isn't a chipset...

[illusoryphoenix]

it's the platform for the MID/UMPC/etc. The chipset is Poulsbo and the CPU is Silverthorne. Just setting the record straight on the submitter's summary. :) By the way, you have to love how every component of the system gets a codename as well as the final system.

Re:But does it have crypto instructions?

[TheRaven64]

It all comes down to battery life. Dedicated silicon almost always uses less power than trying to run the same algorithms on general-purpose silicon (assuming both were fabricated with the same technology). Think how much power your CPU would take trying to emulate a GPU, and then think how little power a GPU of equivalent speed would take.

The other aspect is, of course, performance. Ultra-mobile chips are underpowered by the standards of desktops or even laptops. While you can do full-disk encryption with a Core 2 without much slowdown, trying to do it with a much slower chip is likely to make things crawl, and network encryption is even worse. When I connect my MacBook Pro to my PowerBook with a FireWire cable and try to scp a file, the bottleneck is the PowerBook's 1.5GHz G4 CPU. 802.11n promises to be three quarters of the speed for FireWire 400, and even if it's only one quarter then it's still going to place a big load on this kind of CPU.

The big killer for battery life is clock speed. Power consumption goes up a lot faster than clock speed, which is why it's important to reduce the clock of mobile CPUs when they are not fully loaded to get the best battery life. If you have some dedicated silicon for a commonly-used algorithm, you can get the same overall performance by running the CPU at a lower speed, which can give a huge improvement to battery life.

Re:put a screen with it...

[Gideon Fubar]

There are 3G USB modems that are becoming increasingly common around where i live. Doubt it'd take much effort to make that into a viable phone platform..