ARM for the future?
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 11572 from 2003/5/22
    From: Germany
    > More tech details, performance and such would interesting though.

    http://en.wikipedia.org/wiki/Raspberry_Pi
    -> "Broadcom processor 2763" ->
    http://en.wikipedia.org/wiki/VideoCore

    More:
    http://www.google.com/search?q=bcm2763+arm11
  • »25.05.11 - 18:17
    Profile
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4957 from 2009/1/28
    From: Delaware, USA
    http://en.wikipedia.org/wiki/Raspberry_Pi

    Whoa! Neat device. I wouldn't mind trying to figure out how to connect multiple Raspberries to a central hub or another motherboard.
    "Never attribute to malice what can more readily explained by incompetence"
  • »26.05.11 - 02:11
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    amigadave
    Posts: 2792 from 2006/3/21
    From: Northern Calif...
    The heck with multiple Raspberry Pi devices being connected together, I wonder how well it would run MorphOS2.x with just one device?

    Wouldn't it be faster than an Efika5200b? Of course it is only theoretical, as MorphOS2.x does not run on ARM architecture. Perhaps it could run AROS though?
    MorphOS - The best Next Gen Amiga choice.
  • »26.05.11 - 07:25
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Zylesea
    Posts: 2028 from 2003/6/4
    Well, probabbly. an ARM11 700MHz based device would be faster than an e300/400MHZ. From raw computing power I think it should be more than two times as fast as the e300/400 (about 1800 MIPS vs. 800 MIPS AFAIK). Plus the Efika 5200 has some additional bottlenecks.
    For the ARM thingie there's the question how the video engine could be used, but it can deliver quite some multimedia content. Seems pretty attractive. Anyway, I guess a 5125 device could deliver quite some fun as well. But it is all rather void: there's no MorphOS for ARM (yet) and no cheap 5125 device.

    What I think is remarkable: A never heard of before prototype project gains pretty much PR (I read about that in my local newspaper a few days ago). If, years ago, Genesi would have had that PR with their 99 US$ Efika, maybe more ppl would have bought such a thing. Years ago the 99 US$ Efika was quite attractive. But it had its design flaws (my major critics back then: ATX instead of 5VDC, usb1.1 only), too.
    --
    http://www.via-altera.de

    Whenever you're sad just remember the world is 4.543 billion years old and you somehow managed to exist at the same time as David Bowie.
    ...and Matthias , my friend - RIP
  • »26.05.11 - 09:42
    Profile Visit Website
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 11572 from 2003/5/22
    From: Germany
    > an ARM11 700MHz based device would be faster than an e300/400MHZ.

    When takemehomegrandma claimed two weeks ago in this very thread that an 800 MHz ARM11 core outperforms a 400 MHz e300c4 core I called him out on that. Needless to say that he has failed to provide evidence.

    > From raw computing power I think it should be more than two times
    > as fast as the e300/400 (about 1800 MIPS vs. 800 MIPS AFAIK).

    With "MIPS" you mean "DMIPS", right? DMIPS performance of MPC5200B's e300c0 is only 4% less than that of e300c4, so they're about on par. 1800 DMIPS for 700 MHz would mean about 2.6 DMIPS per MHz for ARM11. That would be more than Cortex-A9 (2.5 DMIPS per MHz), so very unlikely. Where do you have that figure from? According to ARM Ltd. the ARM11 core delivers about half that performance, namely between 1.18 and 1.54 DMIPS per MHz (depending on the specific type of ARM11 core) and thus between 826 and 1078 DMIPS at 700 MHz. Add to this that the Dhrystone benchmark is probably not suited to benefit from a processor's out-of-order execution capability which the e300 provides but the ARM11 does not.

    DMIPS per MHz figures for different ARM11 types:
    1.18: http://www.arm.com/products/processors/classic/arm11/arm1136.php?tab=Performance
    1.25: http://www.arm.com/products/processors/classic/arm11/arm1136.php?tab=Specifications
    1.25: http://www.arm.com/products/processors/classic/arm11/arm1176.php?tab=Specifications
    1.41: http://www.arm.com/products/processors/classic/arm11/arm1156.php?tab=Performance
    1.54: http://www.arm.com/products/processors/classic/arm11/arm1156.php?tab=Specifications

    I don't know which one of these cores the BCM2763 incorporates though.
  • »26.05.11 - 11:56
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Zylesea
    Posts: 2028 from 2003/6/4
    Andreas_Wolf,
    Quote:


    With "MIPS" you mean "DMIPS", right? DMIPS performance of MPC5200B's e300c0 is only 4% less than that of e300c4, so they're about on par. 1800 DMIPS for 700 MHz would mean about 2.6 DMIPS per MHz for ARM11. That would be more than Cortex-A9 (2.5 DMIPS per MHz), so very unlikely. Where do you have that figure from? According to ARM Ltd. the ARM11 core delivers about half that performance, namely between 1.18 and 1.54 DMIPS per MHz (depending on the specific type of ARM11 core) and thus between 826 and 1078 DMIPS at 700 MHz.


    Somewhere I read 2600 DMIPS for 1 GHz and drew my conclusion for the 700MHz chip. Sorry, messed up values, since the 2600 was not the 1GHz 1core version, but for some 4 core version. Had only a quick look (and false) look regarding computing power for this chip yet. I now also read about ~1.2 DMIPS/MHz. Well, with DMIPS values like that the device is of course still highly interesting and impressive for that little money, but not that otally blowing as I first thought - i.e. it then seems rather on par with e300/400. The inbuild powerful video codec is a very nice feature though of course.
    --
    http://www.via-altera.de

    Whenever you're sad just remember the world is 4.543 billion years old and you somehow managed to exist at the same time as David Bowie.
    ...and Matthias , my friend - RIP
  • »26.05.11 - 12:42
    Profile Visit Website
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 11572 from 2003/5/22
    From: Germany
    > Somewhere I read 2600 DMIPS for 1 GHz and drew my conclusion for
    > the 700MHz chip. Sorry, messed up values, since the 2600 was not the
    > 1GHz 1core version, but for some 4 core version.

    I think that must have been a dual-core ARM11MPCore chip, not quad-core. Your figure was about double the true figure, not quadruple.

    http://www.arm.com/products/processors/classic/arm11/arm11-mpcore.php?tab=Performance
  • »26.05.11 - 12:58
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 11572 from 2003/5/22
    From: Germany
    > Perhaps it could run AROS though?

    There's a short thread on this device on aros-exec.org:

    http://aros-exec.org/modules/newbb/viewtopic.php?topic_id=5956&forum=4
  • »26.05.11 - 13:25
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    takemehomegrandma
    Posts: 2708 from 2003/2/24
    Quote:

    When takemehomegrandma claimed two weeks ago in this very thread that an 800 MHz ARM11 core outperforms a 400 MHz e300c4 core I called him out on that. Needless to say that he has failed to provide evidence.


    I never said I was going to serve you with anything Andreas, so I haven't failed. It's all in your head, like so many other things, and I really wish you would keep it there. Your brains incapability to follow a discussion at large, and comprehend the essence of it, is notorious; all the time you seem to stumble and fall on the words a discussion is made of, with the result of you completely missing out on the discussion itself. You have showed this behavior in practically all threads you have taken part in, and this one as well. If your Asperger brain can't identify a pattern it can handle and process (preferably in the shape of some list of indexable details you can save as links) it tries to recompute it into a pattern it *can* handle and process (which probably will be some list of indexable details you can save as links), and then you will start discussing *this* instead of what was *really* being discussed... and yet another thread derails! Yet again! The ironic thing is that when people simply stops replying to your posts, you probably feel like you have "won" the discussion, and that you are right in your claims. You are actually showing signs of it here, in this very thread. But don't worry, I actually think all this (and you as a character as well) is highly amusing, and I honestly think that every community should have room for some originals, as it sure spices up the atmosphere...
    MorphOS is Amiga done right! :-)
    MorphOS NG will be AROS done right! :-)
  • »26.05.11 - 16:18
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 11572 from 2003/5/22
    From: Germany
    > I never said I was going to serve you with anything Andreas, so I haven't failed.

    You made a claim and didn't back it up when called out on it. That's what I call a failure to provide evidence. Whether you failed deliberately or not is another matter.

    > It's all in your head

    No, your claim about ARM11 outperforming e300c4 is right here in this thread for everybody to read:

    "the PPC camp would probably put forward the 5121e, and the ARM camp could choose from, say, the i.MX51, i.MX53, Tegra 1 or Tegra 2. All of those ARM chips performs much better than the PPC"
    https://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=7675&start=89

    As you know, it's the "Tegra 1" part of your claim I've been objecting to. Tegra 1 has an ARM11 core and clocks up to 800 MHz, MPC5121e has an e300c4 core and clocks up to 400 MHz. So it's your claim that 800 MHz ARM11 "performs much better than" 400 MHz e300c4. I can't see why that claim should be in my head when in fact it's in what you wrote.

    > I really wish you would keep it there.

    I really wish you'd keep your nonsense claims wherever they are prior to you spouting them out in public.

    > Your brains incapability to follow a discussion at large, and
    > comprehend the essence of it, is notorious

    The essence of a discussion (if there is such thing by default at all) is determined by the essences of the statements that discussion is compound of. You made several dubious statements on the topic of ARM in this thread, which I listed there:

    https://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=7675&start=92 (last part)

    So far, you chose to not address my objections and questions regarding those statements.

    > all the time you seem to stumble and fall on the words a discussion is made of

    Words are what constitute the meaning of a proposition for its recipient. You wrote "Tegra 1" but didn't mean to write "Tegra 1"? Your problem, not mine. It could have been easy for you to retract the "Tegra 1" part of your claim and I'd have been fine by it, but you chose not to retract it. You're the only one responsible for your choices.

    > with the result of you completely missing out on the discussion itself.

    No, I understand the "discussion itself" very well. But that doesn't mean I won't address what I believe are lies or false statements or ask questions regarding what I believe are dubious statements, be they part or not of what you think the "essence" is.

    > You have showed this behavior in practically all threads you
    > have taken part in, and this one as well.

    That's true. Whenever I see something on the message boards I'm active on which I think is a lie, a false statement or a dubious statement I feel free to address it either by outright correcting it (often with links) or by asking questions.

    > If your Asperger brain can't identify a pattern it can handle and process

    Huh? I *can* identify, handle and process the "Tegra 1" part of your claim. That's the reason I objected to it. Or what alleged "pattern" do you refer to regarding your "performs much better than" claim?

    > it tries to recompute it into a pattern it *can* handle and process

    Huh? I didn't "recompute" anything. Your claim that Tegra 1 "performs much better than" MPC5121e is written by you as clear as it could be. Substituting "Tegra 1" by "ARM11" and "5121e" by "e300c4" was done by me in order to abstract from the whole SoC to the part delivering the core performance. The clock frequencies were mentioned by me to point at the maximum core frequencies the Tegra 1 and the MPC5121e can operate at (to make a fair assessment and not compare to a lowly clocked 400 MHz Tegra 1 for instance). So where did I "recompute" anything?

    > then you will start discussing *this* instead of what was *really* being discussed...

    If you don't want things being discussed then don't mention them. In particular: If you don't want your performance claims regarding Tegra 1 (in comparison to MPC5121e) being discussed then don't make them in the first place. Simple.

    > yet another thread derails! Yet again!

    You're the one derailing it. Performance comparison between Tegra 1 and MPC5121e is well within the scope of this thread's topic I'd say (else your original attempt at comparing them would have been off-topic already). Your discussion of my "Asperger brain" and my posting habits is definitely not. (Btw, what do "yet another" and "yet again" refer to specifically?)

    > The ironic thing is that when people simply stops replying to your posts,
    > you probably feel like you have "won" the discussion

    If someone makes a certain claim and I ask for evidence regarding this claim but this someone won't address this request I'm inclined to believe he has no such evidence but pulled the claim from his behind. There's nothing ironic about that.

    > and that you are right in your claims.

    Huh? It's *your* claim that Tegra 1 "performs much better than" MPC5121e, not mine. I didn't even claim the opposite but just asked you for evidence to back up your claim.

    > I actually think all this (and you as a character as well) is highly amusing

    I'm glad to be able to bring some joy into your life :-)
  • »26.05.11 - 17:46
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    takemehomegrandma
    Posts: 2708 from 2003/2/24
    You *still* don't get it, do you? Well, I rest my case...
    MorphOS is Amiga done right! :-)
    MorphOS NG will be AROS done right! :-)
  • »27.05.11 - 07:53
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 11572 from 2003/5/22
    From: Germany
    > You *still* don't get it, do you?

    Right, I still don't get your claim that Tegra 1 "performs much better than" MPC5121e.

    > I rest my case...

    Thought so. You're a loudmouth who's not able or willing to back up his claims with evidence when challenged on them.
  • »27.05.11 - 10:08
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    takemehomegrandma
    Posts: 2708 from 2003/2/24
    :lol:
    MorphOS is Amiga done right! :-)
    MorphOS NG will be AROS done right! :-)
  • »27.05.11 - 11:56
    Profile
  • Caterpillar
    Caterpillar
    Aramon
    Posts: 35 from 2011/4/21
    From: Germany - Hamburg
    I think the ARM platform is a very nice solution :-)
    Because the OS needs new computer. A SAm is to expensive and to slow.

    [ Editiert durch Aramon 27.05.2011 - 13:20 ]
    Mythana das kostenlose Browser-Rollenspiel: http://www.mythana.de
  • »27.05.11 - 12:18
    Profile Visit Website
  • Yokemate of Keyboards
    Yokemate of Keyboards
    amigadave
    Posts: 2792 from 2006/3/21
    From: Northern Calif...
    If the MorphOS Dev. Team was already working on a port to any Arm based hardware, the Raspberry Pi sure would be a great entry point, low end system, if it really does go into full scale production and can be sold for $25, like it is advertised as a target price.

    Porting to Arm would mean that soon MorphOS2.x could be running on all kinds of cell phones and tablets and netbooks (not that I really want to run MorphOS2.x on a phone).
    MorphOS - The best Next Gen Amiga choice.
  • »27.05.11 - 13:10
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Zylesea
    Posts: 2028 from 2003/6/4
    But keep in mind that MorphOS is no an OS very useful for phones. It is not secure per se and the UI is rather designed for the desktop.
    The arm-netbooks are another story. While MorphOS could shine on them in theory, their big success is long promised, but eventually at least very delayed.
    But okay, we discussed theoretical ISA switche backa and forth...

    Lets see Powerbook and G5 support first.

    But when this guy gets so much PR with his 25$ ARM stick, maybe it would be worthwhile to do something similar (probably a little bit more expensive though) with the 5125? Anyone some serious spare money to spend? Some expertise to bring in? I would join such an effort immediately. The 5125 is cheap, rather complete (sound is missing), probably similar powerful and energy efficient. C'mon - what we're waiting for :-)
    --
    http://www.via-altera.de

    Whenever you're sad just remember the world is 4.543 billion years old and you somehow managed to exist at the same time as David Bowie.
    ...and Matthias , my friend - RIP
  • »27.05.11 - 15:29
    Profile Visit Website
  • Caterpillar
    Caterpillar
    Aramon
    Posts: 35 from 2011/4/21
    From: Germany - Hamburg
    But the Question is....would like the MorphOS Team develop MorphOS for ARM?
    But I think, ARM or Intel are the only way to generate new customers.
    Mythana das kostenlose Browser-Rollenspiel: http://www.mythana.de
  • »27.05.11 - 16:03
    Profile Visit Website
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4957 from 2009/1/28
    From: Delaware, USA
    >Lets see Powerbook and G5 support first.

    I'd definitely second Zylesea's opinion. These solutions would be faster and still be compatible with our current base of software. They are also fairly inexpensive (although obviously more than $25).

    Frankly 700-800 Mhz isn't that impressive. And we still don't have SMP support, so multiple cores aren't useful.
    As ARM moves toward 2Ghz and higher, it then becomes a more attractive solution.
    "Never attribute to malice what can more readily explained by incompetence"
  • »27.05.11 - 16:55
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 11572 from 2003/5/22
    From: Germany
    > Frankly 700-800 Mhz isn't that impressive.

    Yes, especially with ARM11. Cortex-A9 offers about twice the performance per clock and core compared to ARM11, thus quadruple the single-core performance when clocked at around 1.5 GHz.
  • »27.05.11 - 17:05
    Profile
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4957 from 2009/1/28
    From: Delaware, USA
    >...especially with ARM11. Cortex-A9 offers about twice the performance per clock and core compared to ARM11, thus quadruple the single-core performance when clocked at around 1.5 GHz.

    Yes, and faster versions on the horizon.
    While I am impressed with ARM in general (and like the EfikaMX's design), the Cortex-A9 is the most impressive variant so far.
    If it wasn't a little buggy, I would already be using a Pandaboard.
    And TI next revision look even more promising.

    If we're going to replace our PPCs, I'd prefer it be with a MORE powerful solution (not a less powerful one).
    "Never attribute to malice what can more readily explained by incompetence"
  • »27.05.11 - 21:08
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 11572 from 2003/5/22
    From: Germany
    > the Cortex-A9 is the most impressive variant so far.

    Yes, so far Cortex-A9 leads the bunch. Just for fun I compiled a small list* of DMIPS per MHz and core figures for various recent and future cores implementing ARM ISA:

    ARM Cortex-A8: 2.0
    Qualcomm Scorpion: 2.1
    Marvell Sheeva PJ4: 2.4
    ARM Cortex-A9: 2.5
    Qualcomm Krait: 3.0...3.5 (estimated)
    ARM Cortex-A15: 3.5 (estimated)

    * Edit: updated list there

    > If it wasn't a little buggy, I would already be using a Pandaboard.

    There are also other Cortex-A9 boards, for instance the Snowball (available end of Q2/2011 starting at 225 USD):

    http://www.igloocommunity.org
    http://www.calao-systems.com/articles.php?lng=en&pg=6186
    http://www.youtube.com/watch?v=ZUpgpjJl-5g

    [ Edited by Andreas_Wolf 14.11.2012 - 00:49 ]
  • »28.05.11 - 00:37
    Profile
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4957 from 2009/1/28
    From: Delaware, USA
    >- 1 x GPS (Ex. Antenna),
    - 1x 3 Axis Accelerometer,
    - 1x 3 Axis Magnetometer,
    - 1x 3 Axis Gyrometer,
    - 1x Pressure sensor,


    Interesting added features on this one Andreas. Thanks Andreas.
    ARM Cortex-A15 really looks promising.

    I can understand your disagreement with takemehomegrandma. ARM11 is a much older, weaker performing platform.The ARM Cortex-A8 in the Efika MX would definitely be a better processor.
    I wonder why, if he was going to compare PPC and ARM processors, he did choose to compare the CPUs used in the Efikas?
    "Never attribute to malice what can more readily explained by incompetence"
  • »28.05.11 - 14:14
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 11572 from 2003/5/22
    From: Germany
    > I wonder why, if he was going to compare PPC and ARM processors,
    > he did choose to compare the CPUs used in the Efikas?

    He didn't. He compared the MPC5121e (which is not used in any Efika) to i.MX51, i.MX53, Tegra 1 and Tegra 2 because he was talking about netbooks and nettops at this point. In fact, it's quite reasonable to compare these chips for those particular types of applications. Nothing wrong with that as there's currently no better integrated PPC chip than the MPC5121e for such purposes. What I found (and still find) debatable though was his following claim that "all of those ARM chips performs much better than" the MPC5121e as I think that might not be true for the ARM11 based Tegra 1. That's why I asked him for evidence regarding the "Tegra 1" part of his claim specifically.
  • »28.05.11 - 17:16
    Profile
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4957 from 2009/1/28
    From: Delaware, USA
    >That's why I asked him for evidence regarding the "Tegra 1" part of his claim specifically.

    Its is questionable. The performance of an older ARM design may not perform as well as even a relatively low end PPC. Later ARM design, as your figure pointed out, clearly would be competitive.

    While I find Zylesea's argument for a low end PPC system somewhat unappealing, I don't think a Tegra1 would be preferable. I not even sure that focusing on the Cortex A9 is wise with the A15 coming soon.

    I know I keep beating on this same drum, but there is never too much CPU power. If we are to replace our PPCs (which perform more then adequately), we should aim higher. Cortex A9 would be a bare minimum. They're cheap.
    "Never attribute to malice what can more readily explained by incompetence"
  • »28.05.11 - 18:43
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 11572 from 2003/5/22
    From: Germany
    > Its is questionable.

    Yes, highly. But apparently takemehomegrandma doesn't think so. At least he actively chose not to retract or qualify his claim.

    > Later ARM design, as your figure pointed out, clearly would be competitive.

    Yes, that's why I didn't challenge the "i.MX51", "i.MX53" and "Tegra 2" parts of takemehomegrandma's claim. It's obvious that those chips are not only competitive but outperform the MPC5121e. And btw, I didn't say Tegra 1 was not competitive compared to the MPC5121e. What I said was that I doubt that Tegra 1 "performs much better than" MPC5121e. That's not the same statement, as something that isn't much better wouldn't have to be worse. It can very well be only slightly better or about on par.

    > with the A15 coming soon.

    "Soon" as in "in 1.5+ years" ;-) Btw, Qualcomm announced Krait sampling for Q2/2011, so while probably not quite as fast as Cortex-A15 it should be there way earlier and still be faster than Cortex-A9.
  • »28.05.11 - 23:40
    Profile