ARMed ARMADA Sails

This is the Year of ARM. Check out Marvell’s latest CPU. 1.6gHz X4 with 64bit ECC RAM capability. Servery for sure. This would make a fine thick client too if you don’t do too much floating point maths. It uses only 10 watts for 16600 DMIPS! x86-64, Watch Out!

The thing to watch for is multiples of these in a box so each CPU has only to handle a few processes to give huge throughput.

About Robert Pogson

I am a retired teacher in Canada. I taught in the subject areas where I have worked for almost forty years: maths, physics, chemistry and computers. I love hunting, fishing, picking berries and mushrooms, too.
This entry was posted in Uncategorized. Bookmark the permalink.

13 Responses to ARMed ARMADA Sails

  1. There are plenty of users around who just play music/video/browse. They don’t need a lot of fancy apps. I have one lady still using XP who had not used her x free runs of Office in a year. She uses her browser to type e-mails and prints from that… I installed OpenOffice.org for her.

    Android has a huge base of apps but so do the GNU/Linux distros like Debian/Ubuntu. It would take very little to move them onto ARM. They already build almost everything for ARV v7.

    see ftp://ftp.debian.org/debian/dists/squeeze/main/binary-armel/Packages.bz2

    The package list for ARM is 6MB compressed.

  2. oldman says:

    “ARM CPUs would not be the bottleneck especially with SSD these days”

    The spinning disk bottleneck will be going away real soon on all computers. SSD’s are becoming cheap enough to put in standard x86 systems. (My new portable is being configured with a 256Gb SSD and the cost to do so isn’t all that bad compared to the performance boost.

    Tech specs by themselves aren’t going to bring ARM into the mainstream IMHO. What will is the perfect storm of desirable applications running on top of a widely used OS which happens to run on top of ARM bease systems. Android may be that vehicle, but Android seems only assured a place in the smart phone and possible the iPad type tablet market. The few laptop/netbook implementations that exist seem to be going nowhere.

    I guess we shall just have to see how it works out.

  3. Most of the slowness of systems these days is not the CPU. It is the hard drive. The typical PC has a 500gB hard drive that can deliver 120 MB/s sustained reading but still has 8 milliseconds seek-time. The CPU can move a bunch of gB per second as can the RAM and the motherboard. The bottleneck is the storage. I use old clunkers and it is the same. The bottleneck is the hard drive. If I use RAID or a network connection to a server with cached files, the performance increases immensely. ARM CPUs would not be the bottleneck especially with SSD these days. Early generations had small slow RAM but they are using more modern RAM with gB capacities now. These CPUs are not slow. See some of the demos on Youtube.

  4. NotZed says:

    The year of the ARM was some time ago. They sell billions of unit licenses every year. Billions every quarter.

    But for desktop computers, they just aren’t all that fast. People are used to speed – they wont give it up easily even if it uses less power and gets the job done. Look how much of the power budget is spent on graphics these days despite the little need for it.

    As for drivers – most of the peripheral hardware is the same, and so the drivers are the same (abstracted by kernel interfaces where necessary). Many peripherals are just USB anyway – same drivers. The CPU’s are modern cpu’s with floating point, vector units, and virtual memory – there’s no real difference compiling user software for ARM or any other 32 bit processor once the kernel is available and the compiler is written – which are all done. And like any hardware, it’s up to the hardware vendors to work on drivers – which they do (otherwise nobody would build stuff with their chips).

    So, software is a complete non-issue. You’re going to have more problems `porting’ most software to 64 bit architectures (powerpc, amd64) than you will recompiling on ARM. Most applications will not require porting as such.

    The big drawback is performance – the compilers are behind, and the hardware is as well. x86 is particularly good at running shitty code fast so the compiler doesn’t have to work so hard to get decent results. And most arm designs have been for very low power devices – which means a slower system architecture as well. i.e. mobile chipsets such as in the beagleboard/ai touchbook/open pandora.

    The other problem is software in general has bloated out a huge amount. Whereas you could easily get work done on a machine made 10 years ago at the time you probably couldn’t even install a modern gnu/linux system on it now through lack of memory or patience.

    Having said all that, x86 is a hacked up load of shit, i look forward to the day it isn’t around.

  5. oldman says:

    @Dann:

    “Even if using ARM at first would mean I wouldn’t have full functionality, it’s still something I would go for. ”

    Fair enough, but for those of use who have gotten used to certain standards from running various wintel configurations. Those limitations will be a turnoff.

    I think the thing to keep in mind is that for most people, computer’s are a means to accomplish a task, but an object to be tinkered with. What may appeal to you as an aficionado of cool technology might be a turnoff for others.

  6. oldman says:

    “Things that call for ARM:
    1)Working in a quiet place, hearing your thoughts, a fanless CPU is helpful.

    2)Working any place, a smaller box gives you more room to work, lay papers, books, write with pen and paper.”

    But most modern desktops from main stream vendors are already quiet. For instance,

    I am working in a quiet place as I type this on My Lenovo R61 portable, the only way that I notice that its fan is running is when I move my hand to the vent and feel the air exhausting out. My Dell Studio xps desktop is equally quiet.

    “ARM makes great thin clients that can do this ”

    Actually the OEM’s like Wyse and HP who license ARM designs make great thin clients. (Most of which are running the Citrix Reflector BTW). But his is not going to be something that home users are going to use, unless you include in the thin client definition something like the code the underpins Google TV.

    “Did you see the specs on this thing? ECC? Quad core? It’s quite capable of doing anything 90% of users do. ”

    Frankly I’m more impressed by the specs on the IBM Power7 CPU. But at the same time I dont expect that beast to be on an affordable desktop running the applications that I use any more than I expect an ARM processor to do so.

    “Many consumers will buy whatever is on the shelves regardless of CPU. We saw that with Vista when consumers bought machines with only 512MB cache and were slow as molasses.”

    But as of now the only things that those consumers have in front of them are Wintel systems. There are no ARM desktops in sight, nor are there significant android based tablets on the market beyond the iPad.

  7. Dann says:

    @oldman

    OK. I can see where you are coming from. That school of thought does put a damper on the mood, but about drivers…
    Shouldn’t it be fairly simple to cross compile open-source drivers onto an arm arch? Obviously proprietary x86/amd64-based nvidia/ati/etc won’t port.

    Then again, now I can see more clearly about how efficient ARM would be on servers, with few peripherals and required throughput/efficiency.
    Or even using ARM in a cluster would be impressive.

    Even if using ARM at first would mean I wouldn’t have full functionality, it’s still something I would go for.

    One great use I could see for ARM boxes is in the media center business. For instance, Dell is releasing (or has already) the Boxee Box, an opensource media center machine for about $100-200. I’ve used boxee and it’s impressive. Having it ported to ARM, or even built INTO tv’s, could be an excellent use case for a manufacturer. Get enough ARM cores in there and you could potentially playback 3D-tv without issue.

    An ARM netbook that is in BETA and has had preorders sell out is the Always Innovating netbook.

    http://www.alwaysinnovating.com/products/

    All ARM (Cortex A8 I think). And I want one.
    For their first models, I believe battery life was 8-10 hours. Not sure if they’ve updated it.

  8. Things that call for ARM:
    1)Working in a quiet place, hearing your thoughts, a fanless CPU is helpful.

    2)Working any place, a smaller box gives you more room to work, lay papers, books, write with pen and paper.

    ARM makes great thin clients that can do this very well as can some Atoms and Via CPUs but now ARM seems much more capable of running a decent thick client for those who need it. Did you see the specs on this thing? ECC? Quad core? It’s quite capable of doing anything 90% of users do. It also will cost less because there’s just less silicon in it and ARM charges very modest per-chip licensing fees leaving the OEM with lots of room to undercut Intel/AMD.

    I think it has only been the last year that good ARM chips for desktop thick clients have become available. They make sense in lots of places. A great one is the all-in-one PC where the guts are in the monitor or keyboard.

    As for software compatibility, the Android OS makes sense since many useful apps are available in Java already so they should run in Dalvik after translation. Debian has a distro-port that runs on smart-phones natively (not relying on Dalvik). OpenOffice will run on these things if they have enough RAM. Some still have only 256 MB/512MB but that is not likely these days because RAM is so cheap. Time to market is incredibly short with ARM designs. The current delays are more related to responding to the iPad and waiting for the next release of Android as much or more than there not being a market. Many consumers will buy whatever is on the shelves regardless of CPU. We saw that with Vista when consumers bought machines with only 512MB cache and were slow as molasses.

    There are already notebooks on the market with dual processors (ARM+Intel) for mobile/stationary work. It is just a matter of time before ARM will be seen capable of doing the whole job.

    see http://simonbramfitt.com/2010/10/arm-powers-innovation-in-the-enterprise.html

    see http://www.digitimes.com/news/a20101101PD223.html

  9. oldman says:

    “Well, you don’t have to do all that work. Folks can run OpenOffice.org on these things if it will fit thanks to Debian and others. You do not have to maintain the code at all. They do it freely.”

    So are you telling me that all of the Major FOSS apps have been converted to ARM? The fact that you use the qualifier “if it will fit” makes me suspicious.

    Being able to cross compile to an ARM CPU is a neat trick, but IMHO ultimately meaningless as there is not only no ARM workstation to run it on , but there is no Standardized system design for ARM based workstations the way that there is for x86 desktops.

    I don’t to be contrary Pog, but I just don’t see where your ARM based desktop is coming from – there simply seems to be no demand for it!.

  10. oldman says:

    @Dann

    “Why so indifferent? Is it because you realize the potential of ARM to displace a certain OS?”

    Lets put it this way:

    Right now I have a workstation quad core i7-970 on my desk. within a year or two I expect have the workstation version of the 8 core Nehelem EX at the same price. That system will run all of the applications that I use now both FOSS and commercial, and even runs an instance of the Linux that I worl with regularly (Red Hat Enterprise Linux 5 x64)in a virtual machine. These are applications that I use professionally, and many of them have no available equivalents in FOSS or have equivalents that are less capable than I have now.

    To give me what I have above:

    First you need a major manufacturer to commit to building a personal desktop on ARM. Doing this will require that someone checks to make sure that all of the drivers as well as any firmware are also adapted for ARM as well. Your would be ARM customers wont buy those systems without peripherals. Note that this will all be without any guarantee of a market!

    At the same time some major commercial linux distributor like Canonical or Red Hat has to commit to supporting and working with that vendor. IN addition, They or someone else is then going to have to take on the task of recompiling at least the major FOS apps to the people will have something to run on their new ARM desktops.

    Since there are NO ARM based desktops to work with , this effort will most likely be in the form of cross compiles. Take a look at the process needed to cross compile OpenOffice on to ARM at

    http://wiki.services.openoffice.org/wiki/ARM

    As you will note, this cross compilation is not the easiest task to perform. Multiply this out by the dozen or more major applications.

    A the end of this massive effort all that will be available is FOSS – You still haven’t addressed the needs or the users of even commercial Linux based apps.

    And you certainly havent addressed my computing needs.

    Simply put, I am indifferent because I doubt that anyone is going to produce an ARM based desktop any time soon.

  11. Dann says:

    Gentoo also has an ARM installation process. You can check out their handbook at gentoo.org.

    oldman, I don’t really understand what you mean by maintaining code yourself.
    The Gentoo maintainers do all that themselves, and since it’s just source code you download then compile, there shouldn’t be much difference between installing for x86 and ARM.

    And what’s not to get excited about? Imagine having the power of 100 cell phones in your desktop using about the same power as a regular processor? THat’s exciting! You have no idea how awesome it would be to have a 16-core arm desktop/laptop with hours and hours of battery life, doing video, compiling, rendering, flash (goodness knows how cpu intensive that plugin is), and transcoding all at the same time. Sure, having ARM on the servers will be a cataylist to its wide-spread adoption, but enthusiasts will be the ones who blog about how it crushes x86 and amd64 arches. Benchmarks will come out. Major retailers will stock laptops and desktops with massive core machines. Moore’s law might get a few jabs in.

    Not to mention the price difference.
    oldman, you don’t give third world countries enough credit. Hopefully your opinion will change drastically in your lifetime.

    Why so indifferent? Is it because you realize the potential of ARM to displace a certain OS?

    Regardless, I wait with baited breath for what ARM will bring.

  12. Well, you don’t have to do all that work. Folks can run OpenOffice.org on these things if it will fit thanks to Debian and others. You do not have to maintain the code at all. They do it freely.

    For local apps that could be a problem but then you can use Java for the local apps and there is no problem again.

    On servers, I expect Apache will find a way to run on them.

    My fascination is that I remember the 6502 which is an ancestor of ARM. I still have my Ohio Scientific Superboard II… All the developments that have gone into ARM and x86 have gone their separate ways until this year when they are competing head-to-head everywhere. The whole world has been complaining about how bulky and hot machines are. ARM does it better if the job is just moving data through for a particular level of power consumption. The reason is that the x86 and x86-64 chips just have way more bits flipping than is necessary to do the job. That’s a waste of silicon and energy. Thus ARM is far more cost-effective. I love that. There will always be complex tasks that x86 does better, probably super-computing where it’s the number of bits flipped per second in the CPU that really matters, but serving web pages etc. is mostly I/O which ARM can do pretty well.

    It seems to me that ARM and GNU/Linux just go well together. They are both modular and tight. That other OS does not belong on ARM.

    Also, competition is good. I will bet that as ARM pushes into thick clients and servers, the price of x86 CPUs finds a way to drop faster. ARM is becoming more like x86 in some ways and I would bet that x86 finds ways to become more like ARM giving us the best of both worlds.

  13. oldman says:

    Pog:
    I have never understood your fixation with a processor that has thus far only proven itself in cell phones, tablets and embedded systems.

    It seems to me that the fact that one has a non commercial distribution of Linux (i.e. debian) that has been compiled to run on some forms of ARM is really a non event. One most then recompile ( and possibly revise) all of the code that one needs to even get to what is available now on x86 linux. THEN you have to maintain it yourself. This is something that the average non-technical computer user simply is not going to do, even if they are located in the third world emerging economies that you seem to pin your hopes on.

Leave a Reply