ARM for the future?
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 12079 from 2003/5/22
    From: Germany
    > Cortex-A15 is due this year...

    I really didn't expect it so soon:

    http://en.wikipedia.org/wiki/Apple_A6

    Apple being first is especially funny as two months ago Samsung claimed their Exynos 5 chip
    to be "the first chip in the market to integrate Cortex-A15 dual core" (page 4).

    Apple A6 having Cortex-A15 cores has not been officially confirmed yet, but there are strong pro arguments:
    http://munsie.tumblr.com/post/31483017459/iphone-5-processor
  • »12.09.12 - 23:18
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 12079 from 2003/5/22
    From: Germany
    Correction:

    > Apple A6 having Cortex-A15 cores has not been officially confirmed yet

    As it has been revealed now, it's not Cortex-A15. It's not Cortex at all but rather Apple's very own implementation of the ARMv7-A ISA. With this, Apple joins the ranks of Marvell (Sheeva PJ4/PJ4B) and Qualcomm (Scorpion, Krait) in developing their own ARMv7 cores rather than just licensing ready-made Cortex-A cores from ARM Ltd.

    http://www.anandtech.com/show/6292/iphone-5-a6-not-a15-custom-core

    I'm anxious to know how Apple's A6 core compares to Cortex-A15 and Krait performance-wise. And I guess that Samsung still can manage to have "the first chip in the market to integrate Cortex-A15 dual core" with their Exynos 5 :-)
  • »15.09.12 - 21:52
    Profile
  • Order of the Butterfly
    Order of the Butterfly
    minator
    Posts: 365 from 2003/3/28
    Andreas_Wolf,
    Quote:

    As it has been revealed now, it's not Cortex-A15. It's not Cortex at all but rather Apple's very own implementation of the ARMv7-A ISA.


    When they bought they bought PA-Semi I suspected they'd do their own CPU.
    I've been wondering if it was something of their own in the A6.
  • »15.09.12 - 23:37
    Profile Visit Website
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 12079 from 2003/5/22
    From: Germany
    > When they bought they bought PA-Semi I suspected they'd do their own CPU.

    Let's not forget Intrinsity here :-)
  • »16.09.12 - 00:04
    Profile
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4977 from 2009/1/28
    From: Delaware, USA
    Andreas_Wolf,
    Quote:

    Let's not forget Intrinsity here :-)


    Think that they will use any of Intrinsity's gating technology?
    "Never attribute to malice what can more readily explained by incompetence"
  • »16.09.12 - 04:57
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 12079 from 2003/5/22
    From: Germany
    > Think that they will use any of Intrinsity's gating technology?

    It's well known that the Cortex-A8 core in Apple's A4 chip uses Fast14, and in the linked posting minator said:

    "Before they were bought, there were rumours of Intrinsity working on a 2Ghz Cortex-A9."

    From this I presume that also the Cortex-A9 core in Apple's A5 and A5X chips uses the technology. Thus I find it highly plausible that Fast14 technology has been applied to the custom core in Apple's A6 chip as well.
  • »16.09.12 - 11:19
    Profile
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4977 from 2009/1/28
    From: Delaware, USA
    That could lead to a really frightening prcessor.
    "Never attribute to malice what can more readily explained by incompetence"
  • »16.09.12 - 16:03
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 12079 from 2003/5/22
    From: Germany
    > That could lead to a really frightening prcessor.

    In the case of the A6 to mediocre performance at a frighteningly low power consumption, yes.
  • »16.09.12 - 16:33
    Profile
  • Order of the Butterfly
    Order of the Butterfly
    minator
    Posts: 365 from 2003/3/28
    Quote:

    In the case of the A6 to mediocre performance at a frighteningly low power consumption, yes.


    mediocre performance?
    Why do you think this?
  • »17.09.12 - 00:41
    Profile Visit Website
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 12079 from 2003/5/22
    From: Germany
    > mediocre performance? Why do you think this?

    First off, I'm talking about CPU performance alone, i.e. not taking the GPU into account. The A6 CPU performance should be mediocre in comparison because the A6 in the iPhone 5 is a dual-core chip running at a fairly low 1.0 GHz. If we assume a per-clock performance similar to that of the Krait core found in Qualcomm's Snapdragon S4 chips, then the result of a performance comparison between the dual-core 1.0 GHz A6 and the dual-core 1.5 GHz Snapdragon S4 (as can be found in current phones from various manufacturers) should be pretty predictable I'd say.
  • »17.09.12 - 08:30
    Profile
  • Order of the Butterfly
    Order of the Butterfly
    minator
    Posts: 365 from 2003/3/28
    Quote:

    The A6 CPU performance should be mediocre in comparison because the A6 in the iPhone 5 is a dual-core chip running at a fairly low 1.0 GHz.


    Nobody knows the clock rate - Geekbench tends to report incorrect results.

    Quote:

    If we assume a per-clock performance similar to that of the Krait core found in Qualcomm's Snapdragon S4 chips,


    That's a big assumption.

    Quote:

    then the result of a performance comparison between the dual-core 1.0 GHz A6 and the dual-core 1.5 GHz Snapdragon S4 (as can be found in current phones from various manufacturers) should be pretty predictable I'd say.


    The Geekbench results (assuming they are accurate) show it to be within 20% of a 1.5GHz S4 based phone. If it really is running at 1GHz (probably not, I'd guess around 1.2GHz) it sounds like it has higher IPC.
  • »17.09.12 - 21:36
    Profile Visit Website
  • Order of the Butterfly
    Order of the Butterfly
    minator
    Posts: 365 from 2003/3/28
    Quote:

    The Geekbench results (assuming they are accurate) show it to be within 20% of a 1.5GHz S4 based phone. If it really is running at 1GHz (probably not, I'd guess around 1.2GHz) it sounds like it has higher IPC.


    Scrub that.
    Looks like the result I look at might have been overclocked and the A6 might actually be faster.

    I was comparing it to a HTC One XL and there are a few results below the 1600 the iPhone 5 posted.
  • »17.09.12 - 21:45
    Profile Visit Website
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 12079 from 2003/5/22
    From: Germany
    > Nobody knows the clock rate - Geekbench tends to report incorrect results.

    Ah okay, I didn't know that. So let's wait and see what the clock rate really is.

    >> If we assume a per-clock performance similar to that of the Krait core
    >> found in Qualcomm's Snapdragon S4 chips,

    > That's a big assumption.

    What's your estimation then on how the A6 core compares to Krait (or Cortex-A15) per clock?

    Btw, I'm in good company regarding this point:
    "We believe the custom A6 CPU is similar in complexity and performance to Cortex-A15 as well as to the Krait CPU that appears in Qualcomm’s newest processors. To reach Apple’s claim of a 2x performance gain over the iPhone 4S (which uses the Apple A5 processor), we expect the A6 contains two CPU cores clocking at roughly 1.2GHz. This speed is lower than what competing A15-class CPUs achieve, presumably to save power. As a result, the iPhone 5 will probably fall short of other high-end smartphones in raw CPU performance, although the same has been true of previous iPhone generations."
    http://www.linleygroup.com/newsletters/newsletter_detail.php?num=4881

    > I'd guess around 1.2GHz

    You're in good company then (see above) ;-)

    > I was comparing it to a HTC One XL and there are a few results below
    > the 1600 the iPhone 5 posted.

    Yeah, it seems that Geekbench favours the A6 very much, or that the A6 favours the benchmarks Geekbench consists of, depending on how you look at it. I'm surprised.
  • »17.09.12 - 23:57
    Profile
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4977 from 2009/1/28
    From: Delaware, USA
    >mediocre performance?

    Yeah Andreas,
    I think you're being a little harsh here, too.
    Even at 1.2 this looks pretty powerful.
    And with Intrinsity's gating it might clock higher.
    "Never attribute to malice what can more readily explained by incompetence"
  • »18.09.12 - 00:50
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 12079 from 2003/5/22
    From: Germany
    > Yeah Andreas, I think you're being a little harsh here, too.

    This may well be. I'm really surprised about how the dual-core A6 seems to beat the 1.5 GHz dual-Krait CPUs in Geekbench at a significantly lower clock rate (which also includes 1.2 GHz). I didn't expect that. This would mean the per-clock performance of the A6 core to be significantly higher than Krait's, and thus maybe even Cortex-A15's.

    > Even at 1.2 this looks pretty powerful. And with
    > Intrinsity's gating it might clock higher.

    I doubt the A6 will clock higher than 1.2 GHz in the iPhone. Maybe later in the next iPad, as the tablet form factor allows for a bit more power consumption.
  • »18.09.12 - 09:14
    Profile
  • Order of the Butterfly
    Order of the Butterfly
    minator
    Posts: 365 from 2003/3/28
    Jim
    Quote:

    Even at 1.2 this looks pretty powerful.
    And with Intrinsity's gating it might clock higher.


    It might be the other way around - they might have designed the frequency significantly lower and used the clever gating to bring it up. This would mean it's a wide but has a short pipeline - a good power saving trick.
  • »23.09.12 - 13:50
    Profile Visit Website
  • Order of the Butterfly
    Order of the Butterfly
    minator
    Posts: 365 from 2003/3/28
    Here's one Andreas missed:

    Project boulder. If it's true, it sounds like a much more high end device and possibly the first time we'll see anyone building an ARM core for the high end.

    In other news Global foundries have pulled in their 14nm process to 2014, this could get very interesting indeed as that's when they'll add FinFets.
  • »23.09.12 - 14:17
    Profile Visit Website
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4977 from 2009/1/28
    From: Delaware, USA
    Well, so much for the PandaBoard.
    Here's my new favorite.
    1.4 GHz Quad Core Cortex A9 for $129 shipped worldwide.

    http://www.hardkernel.com/renewal_2011/products/prdt_info.php?g_code=G133999328931
    "Never attribute to malice what can more readily explained by incompetence"
  • »30.09.12 - 01:01
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 12079 from 2003/5/22
    From: Germany
    > Project boulder [...] sounds like a much more high end device

    Much more high-end than what? If I have understood right, Project Boulder is supposed to be a server chip that incorporates Project Denver's CPU core.

    > and possibly the first time we'll see anyone building an ARM core for the high end.

    If my assumption above is correct, the ARM *core* would still be Project Denver, not Project Boulder. And of course, Project Denver would have to be completed before Project Boulder.
    Furthermore, I think Applied Micro (X-Gene) or even Cavium (Project Thunder) might arrive in high-end ARM land before nVidia (Project Denver/Boulder).
  • »30.09.12 - 09:34
    Profile
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 12079 from 2003/5/22
    From: Germany
    Update:

    >>> the A6 in the iPhone 5 is a dual-core chip running at a fairly low 1.0 GHz.

    >> Nobody knows the clock rate - Geekbench tends to report incorrect results.

    > Ah okay, I didn't know that. So let's wait and see what the clock rate really is.

    It's said to be actually 1.3 GHz:

    http://www.engadget.com/2012/09/26/apple-a6-cpu-13ghz-geekbench-confirmed-overclocking/
  • »08.10.12 - 22:01
    Profile
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4977 from 2009/1/28
    From: Delaware, USA
    Received TI Stellaris LM4F120 LaunchPad evaluation board today.
    Pretty neat for $5.
    "Never attribute to malice what can more readily explained by incompetence"
  • »10.10.12 - 00:00
    Profile
  • Order of the Butterfly
    Order of the Butterfly
    minator
    Posts: 365 from 2003/3/28
    If anyone is actually looking at ARM this might be of some use:

    http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dai0245b/index.html
  • »12.10.12 - 22:36
    Profile Visit Website
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Andreas_Wolf
    Posts: 12079 from 2003/5/22
    From: Germany
    > http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dai0245b/index.html

    I had expected to see this link here sooner ;-)

    http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=35953&forum=33#670028
    http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=35953&forum=33&start=20#670258
  • »12.10.12 - 22:54
    Profile
  • ASiegel
    Posts: 1370 from 2003/2/15
    From: Central Europe
    Quote:

    Well, so much for the PandaBoard. Here's my new favorite.
    1.4 GHz Quad Core Cortex A9 for $129 shipped worldwide.

    http://www.hardkernel.com/renewal_2011/products/prdt_info.php?g_code=G133999328931


    I would like to point out that there is a big difference between a consumer product and a development platform which is being sold at a subsidized price (PandaBoard, etc.).

    Consumer products using the same Samsung processor are substantially more expensive.
  • »12.10.12 - 23:48
    Profile
  • Jim
  • Yokemate of Keyboards
    Yokemate of Keyboards
    Jim
    Posts: 4977 from 2009/1/28
    From: Delaware, USA
    ASiegel,Quote:


    I would like to point out that there is a big difference between a consumer product and a development platform which is being sold at a subsidized price (PandaBoard, etc.).

    Consumer products using the same Samsung processor are substantially more expensive.


    As long as it is sold to everyone without qualification, its still worthy of consideration.
    "Never attribute to malice what can more readily explained by incompetence"
  • »13.10.12 - 00:27
    Profile