Intel splits on Atom after the mobile relevance of x86 whacked by Apple's Ax

2456

Comments

  • Reply 21 of 102
    SnRaSnRa Posts: 65member
    SnRa said:
    Apollo Lake with Goldmont cores is still coming (replacing Braswell), so that shouldn't be a problem. Only SoFIA and Broxton (smartphones/tablets) have been cancelled.
    Thank you for the tips. Now can you comment please on a much discussed issue such as the lack of Thunderbolt 3 in the new Retina Macbook? Does Core M support TB 3, or if not what is Intel's solution?
    Yes, Skylake Core m devices can support Thunderbolt 3. A number of Windows based Core m tablets already have Thunderbolt 3 (example: Dell XPS 12, Acer Switch Alpha 12).
    cnocbui
  • Reply 22 of 102
    irelandireland Posts: 17,785member
    Please ditch that Trump avatar. Kills the vibe every time I visit.
    SnRajustadcomicsmagman1979revenantmobiusquadra 610pscooter63
  • Reply 23 of 102
    Herbivore2Herbivore2 Posts: 367member
    SnRa said:
    Proof? 

    Core M also costs $281-$393 per chip. That's not smartphone material, even if you assume a volume discount. 
    http://www.engadget.com/2015/09/02/intel-s-core-m-processors-are-being-tested-in-phones/

    I wouldn't see core m being a mainstream product for phones, but high end devices like that "rumored" Surface phone could still be possible.

    Also, don't be fooled by the list price of that chip. All said and done, no one is paying that much.
    Are you saying that someone like Apple would be able to purchase A core M for approximately the same price they are paying for the A9X?!? There is no way that Intel would ever sell a core M CPU for the price of the A9X. The core M is still far too expensive even at a discount. They lost money on the inexpensive Atom chips as ARM chips are still even cheaper. 

    I do agree that the core M would never go into a smartphone. Intel and Microsoft are both in trouble. 

    Smart guy that Otellini. Krzanich isn't any better either. And if I were Nadella of MSFT, I would be on the phone to Cook attempting to acquire A10X SOCs to build out a Windows phone anyhow. 
    tmaycornchip
  • Reply 24 of 102
    SnRaSnRa Posts: 65member
    SnRa said:
    http://www.engadget.com/2015/09/02/intel-s-core-m-processors-are-being-tested-in-phones/

    I wouldn't see core m being a mainstream product for phones, but high end devices like that "rumored" Surface phone could still be possible.

    Also, don't be fooled by the list price of that chip. All said and done, no one is paying that much.
    Are you saying that someone like Apple would be able to purchase A core M for approximately the same price they are paying for the A9X?!? There is no way that Intel would ever sell a core M CPU for the price of the A9X. The core M is still far too expensive even at a discount. They lost money on the inexpensive Atom chips as ARM chips are still even cheaper. 

    I do agree that the core M would never go into a smartphone. Intel and Microsoft are both in trouble. 

    Smart guy that Otellini. Krzanich isn't any better either. And if I were Nadella of MSFT, I would be on the phone to Cook attempting to acquire A10X SOCs to build out a Windows phone anyhow. 
    I never made such a claim, but the prices on ARK are not what OEM's are paying. For example, there are Core m devices selling for as low as $400.

    I don't understand your last comment. I don't know what Microsoft would do with an A10X for Windows phone. For tablets Microsoft currently has Skylake Core m and Core i options and for phones they have a number of SoC's at their choosing, such as the Snapdragon 820.
  • Reply 25 of 102
    macplusplusmacplusplus Posts: 2,096member
    SnRa said:
    Thank you for the tips. Now can you comment please on a much discussed issue such as the lack of Thunderbolt 3 in the new Retina Macbook? Does Core M support TB 3, or if not what is Intel's solution?
    Yes, Skylake Core m devices can support Thunderbolt 3. A number of Windows based Core m tablets already have Thunderbolt 3 (example: Dell XPS 12, Acer Switch Alpha 12).
    6th generation Core M features somewhat and partially overlap with some TB 3 features but to say that these ports fully provide TB3 benefits, serious testing must be done. TB 3 is also backwards compatible and you may well get TB 2 speeds with these TB 3 ports.
    edited April 2016
  • Reply 26 of 102
    ration alration al Posts: 81member
    SnRa said:
    Apollo Lake with Goldmont cores is still coming (replacing Braswell), so that shouldn't be a problem. Only SoFIA and Broxton (smartphones/tablets) have been cancelled.
    Thank you for the tips. Now can you comment please on a much discussed issue such as the lack of Thunderbolt 3 in the new Retina Macbook? Does Core M support TB 3, or if not what is Intel's solution?
    Thunderbolt 3 is implemented using a separate controller chip (DSL6540) which requires processor/system support for PCI express Generation 3.

    http://ark.intel.com/products/87402/Intel-DSL6540-Thunderbolt-3-Controller

    This chip wasn't available before early 2015 when Apple launched the MacBook with a diminutive PCB to maximize space for battery. Its lack of support for it is probably a design trade-off between battery life and the power-user perception of a 40 Gbps transfer speed.
    tmaysteveh
  • Reply 27 of 102
    starwarsstarwars Posts: 72member
    This is the results when hardware and software maker don't go in tune, either hardware is not optimised for the software or vice versa. In the name of atom, I can't remember I ever had pleasant experience. The entire thing is slow and useless. Atom spells for ultra low performance. Should have ditched the atom long long ago.
    magman1979
  • Reply 28 of 102
    SnRaSnRa Posts: 65member
    SnRa said:
    Yes, Skylake Core m devices can support Thunderbolt 3. A number of Windows based Core m tablets already have Thunderbolt 3 (example: Dell XPS 12, Acer Switch Alpha 12).
    I asked for the direct support for Thunderbolt 3 built into the Core M, not the existence of TB 3 ports in Core M machines.

    To what extent these TB 3 ports support the TB 3 standard with that Core M? If the processor only supports TB 2 speeds (such as PCIe 2.0 support embedded in Core M), these ports can only run in TB 2 speeds regardless of the existence of the TB 3 controller, remember TB 3 is also backwards compatible !!!... 
    These devices have full TB3 support, they are not running at TB2 speeds. Skylake Core m support PCIe 3.0 x4, so 40 Gb/s through TB3 is possible.

    One company in particular, Acer, launched a Thunderbolt 3 graphics dock (GTX 960M 4GB) with their Aspire Switch 12 S tablet. It should also be noted that the graphics dock should work with any Thunderbolt 3 device and not just the Switch 12 S.
  • Reply 29 of 102
    misamisa Posts: 827member
    If I were Intel I'd scrap all the non-socket 2011 models and figure out how to produce parts that can be used in all verticals rather than making a hundred different parts that only differ in power consumption. If you could clock down a "server" class cpu to 1.9ghz and have it take 15 watts, that would be amazing, but no, servers are all 135 watt parts at 2Ghz instead of being able to scale to 4Ghz. Imagine being able to software-define which CPU configuration you want "higher single thread speed" eg 2 or 4 cores at 4Ghz, or "higher threading capacity" 16 cores at 2Ghz, or "Longest battery life" of running only 2 cores at 2Ghz and turning off the other 14 unless plugged in.

    The Atom chip ends up in products that are too underpowered (eg NAS devices and "cheap" x86 laptops that can barely run windows), combined with the extremely weak GPU performance, these were never going to end up in mobile phones and tablets.

    Apple's Ax series of CPU+GPU parts actually is close to Intel's laptop CPU (not Atom) parts in performance at much lower power usage. eg the A9 (1.85Ghz) = i5-4300U (1.9Ghz)

    But that ignores two things, the CPU performance is not at parity with a desktop to begin with:
    The i5-4300U is dual-core part running at 1.9Ghz
    Passmark Rating: 3751
    Passmark Single Thread Rating: 1606

    Intel's strongest CPU is the Intel Core i7-4790K @ 4.00GHz
    Passmark Rating: 11205
    Passmark Single Thread Rating: 2,529

    And the GPU performance is garbage.
    Both of those parts have a HD 4400 part, passmark GPU performance: 547
    The minimum GPU you need to get "game console performance" is the nVidia Geforce GTX 960: 5941

    Based on an image found here:
    http://appleinsider.com/articles/15/02/26/apple-partner-imagination-technologies-unveils-new-powervr-super-gpu-with-512-alu-cores
    The A9 is equal to an Xbox 360 GPU (vs PowerVR GT7600.) The next part up is compared to the Geforce 730m (vs PowerVR GT 7900) which has a GPU passmark of 800

    So we're still a far cry from desktop performance, both on the CPU and GPU side for the A9, but Intel's own (Atom) parts are weaker than what Apple is using in their own A9 parts, and consumes less power in the process.

    It could be reasoned that in the next two years Apple's A10 and A11 parts will leapfrog Intel's x86 offerings for all mobile devices, even the high end laptop parts. The A9 is already powerful enough to "beat the Xbox360", so it's reasonable to think by the A11 it could beat the PS4. 
    tmay
  • Reply 30 of 102
    macplusplusmacplusplus Posts: 2,096member
    SnRa said:
    I asked for the direct support for Thunderbolt 3 built into the Core M, not the existence of TB 3 ports in Core M machines.

    To what extent these TB 3 ports support the TB 3 standard with that Core M? If the processor only supports TB 2 speeds (such as PCIe 2.0 support embedded in Core M), these ports can only run in TB 2 speeds regardless of the existence of the TB 3 controller, remember TB 3 is also backwards compatible !!!... 
    These devices have full TB3 support, they are not running at TB2 speeds. Skylake Core m support PCIe 3.0 x4, so 40 Gb/s through TB3 is possible.

    One company in particular, Acer, launched a Thunderbolt 3 graphics dock (GTX 960M 4GB) with their Aspire Switch 12 S tablet. It should also be noted that the graphics dock should work with any Thunderbolt 3 device and not just the Switch 12 S.
    The point is, Intel has no specific datasheet for a specific Core-M model (m3/m5/m7), actual datasheets refer to a higher generation such as 6th Generation H-platforms and alike... These specs are obscure in whether some specs (for example as seen in display resolutions for 5th Generation Core M) require additional cooling or not. This is why I claim the support of Skylake Core M for TB 3 on these machines must be put under rigorous testing to see to what extent these ports provide all of TB 3 benefits.
    edited April 2016
  • Reply 31 of 102
    robjnrobjn Posts: 264member
    dysamoria said:
    "Fan fiction" ha ha ha  :D
    My next door neighbor here in Cupertino is an engineer at Intel. His job is (or was) to run benchmark tests on Atom chips in Android devices.

    This guy heavily praises Apple's chips when in casual conversation with me. He says they have some kind of secret in the way they "schedule" operations through the processor. He laments that no one at Intel knows how Apple do it!

    Of course, like most engineers here in Silicon Valley, he chooses to use an iPhone. This isn't "fan boy" fiction. The Atom chip has utterly failed to compete in the tablet and smartphone market.


    justadcomicsmagman1979nolamacguytmaypscooter63stevehbb-15badmonk
  • Reply 32 of 102
    SnRaSnRa Posts: 65member
    misa said:
    If I were Intel I'd scrap all the non-socket 2011 models
    Umm...

    As for the rest, I have no idea how you come to those conclusions. Honestly. Your benchmark claims are all over the place. And why are you talking about Haswell?... For that matter, why are you trying to compare x86 and ARM using Passmark? It would be terribly misleading should anyone actually take your post seriously.
    cnocbuinetmage
  • Reply 33 of 102
    macplusplusmacplusplus Posts: 2,096member
    ration al said:
    Thank you for the tips. Now can you comment please on a much discussed issue such as the lack of Thunderbolt 3 in the new Retina Macbook? Does Core M support TB 3, or if not what is Intel's solution?
    Thunderbolt 3 is implemented using a separate controller chip (DSL6540) which requires processor/system support for PCI express Generation 3.

    http://ark.intel.com/products/87402/Intel-DSL6540-Thunderbolt-3-Controller

    This chip wasn't available before early 2015 when Apple launched the MacBook with a diminutive PCB to maximize space for battery. Its lack of support for it is probably a design trade-off between battery life and the power-user perception of a 40 Gbps transfer speed.
    Only 6th Gen. i.e. Skylake Core M (m3/m5/m7) supports PCIe 3.0 and can benefit from that TB 3 controller. Intel's documentation is obscure in whether some TB 3 features require additional cooling or not, as seen in display resolutions of 5th Gen. Core M for example. For reasons only an OEM would know, Apple may have omitted TB 3 support in the rMB. The low-power variant of that controller for example is shown as suitable for MBA in wikipedia, and the MBA is not a fanless laptop.
    edited April 2016
  • Reply 34 of 102
    sprockketssprockkets Posts: 796member
    Hmm...maybe the reason why Intel failed is because there is no compelling advantage to using atom chips vs any arm chips, not the Apple A series. No one considers them high end. Android didn't fail to help Intel, but Intel failed to take over the market by force. Kinda hard to profit from those chips when you priced them to dump on the market and they weren't great performers for battery life.
     Only DED could frame this as an apple vs everyone else. So predictable.

     "As noted yesterday by PC World, the cancelation of phone-scaled Atom chips by Intel also "casts a dark shadow over the rumored Surface Phone," a vaporware product that imagined packing a desktop PC into a smartphone, which could then be plugged into a monitor and keyboard using Microsoft's "Continuum" strategy to run legacy x86 software."

     Microsoft was NEVER going to release a phone compatible with the ancient win32 programs. That would be stupid and violate every security feature WM10 has.
    edited April 2016 jonlcnocbui
  • Reply 35 of 102
    Rayz2016Rayz2016 Posts: 6,957member
    So if I'm reading this right. 

    Intel failed because they failed to win the single customer who is swallowing all the profits in the mobile market?

    I really don't understand their attempt but their way into the market with subsidies. What is the point of buying into a market where no one seems to be making any money?

    edited May 2016 tmaycornchipbadmonk
  • Reply 36 of 102
    vision33rvision33r Posts: 213member
    robjn said:
    dysamoria said:
    "Fan fiction" ha ha ha  :D
    My next door neighbor here in Cupertino is an engineer at Intel. His job is (or was) to run benchmark tests on Atom chips in Android devices.

    This guy heavily praises Apple's chips when in casual conversation with me. He says they have some kind of secret in the way they "schedule" operations through the processor. He laments that no one at Intel knows how Apple do it!

    Of course, like most engineers here in Silicon Valley, he chooses to use an iPhone. This isn't "fan boy" fiction. The Atom chip has utterly failed to compete in the tablet and smartphone market.


    The secret is simple, design the CPU and have the guys next door in your office make the OS and APIs built for it.
    tmaycornchipbadmonk
  • Reply 37 of 102
    ration alration al Posts: 81member
    ration al said:
    Thunderbolt 3 is implemented using a separate controller chip (DSL6540) which requires processor/system support for PCI express Generation 3.

    http://ark.intel.com/products/87402/Intel-DSL6540-Thunderbolt-3-Controller

    This chip wasn't available before early 2015 when Apple launched the MacBook with a diminutive PCB to maximize space for battery. Its lack of support for it is probably a design trade-off between battery life and the power-user perception of a 40 Gbps transfer speed.
    Only 6th Gen. i.e. Skylake Core M (m3/m5/m7) supports PCIe 3.0 and can benefit from that TB 3 controller. Intel's documentation is obscure in whether some TB 3 features require additional cooling or not, as seen in display resolutions of 5th Gen. Core M for example. For reasons only an OEM would know, Apple may have omitted TB 3 support in the rMB. The low-power variant of that controller for example is shown as suitable for MBA in wikipedia, and the MBA is not a fanless laptop.
    The new Dell XPS 12 2-in-1 is the only other Skylake Core M based unit I could find that has 2x TB3 ports. All the reviews speak about 40 C spot temperatures on the case under processor full load and throttling down to 1 Ghz in short order.
    They also mentioned the mid-level m5 version doing light browsing for only 6.5 hours versus closer to 10 hours for the new MacBook on a similar sized battery.
    Too bad USB3.1 Gen 2 goes out the window with the baby because it's implemented using the same controller as TB3 (Alpine Ridge). It could have been a good compromise otherwise.
    tmay
  • Reply 38 of 102
    qwweraqwwera Posts: 281member
    And to top it all off, the newest Apple Ax chips are not just power efficient, the devices are not even able to utilize the complete power of these beasts!
    cornchip
  • Reply 39 of 102
    christopher126christopher126 Posts: 4,366member
    When the founder of a company dies, the following CEO's drive the business into the ground 95% of the time. Ford, Apple, McDonald's, Boeing are the exceptions not the rule. America's fascination with CEO's is misguided.

    Of the Fortune 500 companies in 1950, 88% (440) are out of business today!

    Thank you Otellini for "passing" on the mobile revolution-brilliant! :(

    Best Regards
  • Reply 40 of 102
    quadra 610quadra 610 Posts: 6,756member
    Intel's gone fission, now that the mobile business has reached critical mass without them. 

    I wonder how they'll handle the fallout.
    cornchip
Sign In or Register to comment.