Waiting for a MBP which is Meltdown and Spectre proof, has 32GB of RAM and HBM GPU-2019 or 2020?

Discussion in 'MacBook Pro' started by KensaiMage, May 4, 2018.

  1. KensaiMage macrumors regular

    KensaiMage

    Joined:
    May 25, 2017
    #1
    As the topic says.

    Who is waiting for such MBP? There are many people who say they are fine even with their 2012 MBPs cause they see only slight improvement in specs nowadays.

    When will it come out?

    Are you going to go with it in 2019 or in 2020 when the MBP gets redesigned again (fixed keyboard)?

    I'm probably going to buy a MBP now after WWDC + apple care and sell it in 3 years buying 2-nd gen redesigned one. (first gen normally sucks). So, first purchase in 2018 and second in 2021. Does it sound to you as a good plan? I'm going to do so cause I need a mac now, and ofc I want it too ^^.

    Kind regards
     
  2. Falhófnir macrumors 68030

    Falhófnir

    Joined:
    Aug 19, 2017
    #2
    Almost certainly going to take until the redesign in 2020, possibly even beyond that. Though there is a rumour Apple will switch to A series chips in 2020 so they will have much greater control if that is the case, and true of the MacBook Pro.
     
  3. flyinmac macrumors 68040

    flyinmac

    Joined:
    Sep 2, 2006
    Location:
    United States
    #3
    You just described yourself as Apple’s favorite type of customer.

    I’m not particularly concerned about the patched vulnerability. There will be more. No matter what we do, or which chip / OS we use, we will be vulnerable to something. Unless we go with really old technology. For example, I wouldn’t be overly worried about an Apple IIgs or Commodore 64 being vulnerable on the Internet (and yes, I have previously browsed the Internet using a graphical GUI on both using a broadband connection).

    Aside from going old school, or ceasing to buy anything until it’s perfect, you’re only other option is to stop using computers if you don’t want to be vulnerable (though whoever else has data on you may still leak it).
     
  4. wxkkeup macrumors member

    wxkkeup

    Joined:
    Apr 2, 2018
    #4
    There’s just so much that we don’t know at this point. Like.. the fact that intel is likely to face issues with that processor which will delay the release to many years after 2020...

    Like, I agree in that I really want that MBP (ice lake..?) but it just doesn’t seem reasonable to keep my hopes up for it.

    I will buy this year’s MBP if it is released, but only because I speculate that this could be the last intel MBP. Well, that.. and I actually need a new laptop, but that’s beside the point lol. My point is that people have been waiting more than 5 years for an updated Mac Pro because of .. production problems or whatever.

    Who’s not to say that the same thing wont happen to MPB after 2018? Ugh.. sorry I’m being so bleak, but I am just irritated at the fact that I can’t tell the future.
     
  5. Queen6 macrumors 604

    Queen6

    Joined:
    Dec 11, 2008
    Location:
    Cocked, Locked, Ready to Rock
    #5
    Here's now:
    Screen-Shot-2018-05-01-at-4.34.53-PM.jpg
    Future's very likely going to get bleaker. Not that this brings any joy, and why the Mac line up is so poor...

    Q-6
     
  6. wxkkeup macrumors member

    wxkkeup

    Joined:
    Apr 2, 2018
  7. cube macrumors Pentium

    Joined:
    May 10, 2004
  8. jerryk macrumors 601

    Joined:
    Nov 3, 2011
    Location:
    SF Bay Area
    #8
    I want one without a AMD GPU. Nvidia with CUDA support please.

    I don't think it will happen, but a guy can dream.
     
  9. cube, May 6, 2018
    Last edited: May 6, 2018

    cube macrumors Pentium

    Joined:
    May 10, 2004
    #9
    HSA > CUDA

    With the convenient side effect that AMD CPUs have no Meltdown.
     
  10. jerryk macrumors 601

    Joined:
    Nov 3, 2011
    Location:
    SF Bay Area
    #10
    Doesn't matter if performance is better if I cannot access it with the Machine Learning software. CUDA is it in ML/AI
     
  11. cube macrumors Pentium

    Joined:
    May 10, 2004
    #11
    Maybe the software needs to support OpenCL instead of Apple implementing a proprietary technology.
     
  12. jerryk, May 6, 2018
    Last edited: May 6, 2018

    jerryk macrumors 601

    Joined:
    Nov 3, 2011
    Location:
    SF Bay Area
    #12
    It should, but none of the platforms for ML do. The Open CL group has been very slow to react to the needs of the ML community to the point were no one wants to work with them. In contrast, Nvidia has dedicated large amounts of resources and created an easy to use "walled garden" of tools and support. And we all know how well an easy to use "walled garden" that "just works" sells.
     
  13. Naimfan Suspended

    Naimfan

    Joined:
    Jan 15, 2003
    #13
    Never.
     
  14. cube macrumors Pentium

    Joined:
    May 10, 2004
    #14
    Is it Khronos that is slow or NVIDIA and Apple?

    AMD and Intel are at OpenCL 2.0, while NVIDIA and Apple are still at 1.2 .
     
  15. jerryk macrumors 601

    Joined:
    Nov 3, 2011
    Location:
    SF Bay Area
    #15
    I think you are talking gaming usage. I am talking Machine Learning.

    Machine Learning toolkits utilize the Nvidia cuDDN (Cuda Deep Neural Network Library) to execute and manage mathematical operations on the GPU instead of the CPU. By doing this they get a large increases in performance by running their simple math operation in parallel on many Cuda cores.

    Unfortunately Nvidia early lead in this area has made them about as embedded as Microsoft Office is in the business world.
     
  16. cube macrumors Pentium

    Joined:
    May 10, 2004
    #16
    That is like saying that many games use GameWorks, which is also a wrong approach.
     
  17. jerryk macrumors 601

    Joined:
    Nov 3, 2011
    Location:
    SF Bay Area
    #17
    I don't do games so I have no idea what you are talking about.
     
  18. cube macrumors Pentium

    Joined:
    May 10, 2004
  19. duervo macrumors 68020

    duervo

    Joined:
    Feb 5, 2011
    #19
  20. cube macrumors Pentium

    Joined:
    May 10, 2004
    #20
    I don't care. There's an open framework and you just don't have to be lazy.
     
  21. jerryk macrumors 601

    Joined:
    Nov 3, 2011
    Location:
    SF Bay Area
    #21
    Ok. I am very familiar with this debate on the TensorFlow community, the largest deep learning framework by far. I am not familiar where Facebook is OpenCL support in Caffe. However their does not seem to be a lot of activity in the Github repo for adding Open CL support to Caffe.

    Regarding TensorFlow there seems to be some people pushing support for OpenCL, but they do not work for Google Brain the real muscle behind TensorFlow. Tensorflow has a user Contrib branch that may include such support. However, one needs to remember that Google also makes their own GPU like hardware called a TPU (Tensor Processing Unit). And has invested heavily in it since it is proprietary to Google Cloud.

    It will be interesting to see where the OpenCL goes. But as I see it, if the large cloud providers, Google, Amazon, Microsoft, etc keep driving their prices for GPU cycles lower and lower, AMD may have missed the boat, since the need to run on a local machine decreases every day. Still it would be nice to have the option, and right now that option means Cuda.

    I believe Nvidia sees this change in the market and is positioning themselves as a vendor of GPUs for commercial purposes like autonomous vehicles and in data centers like those that power the Amazon and Microsoft cloud offerings.

    This difference in this effectiveness of this strategy is reflected in the relative finances of the companies.

    AMD Nvidia
    Revenue $5.33 billion $9.71 billion
    Operating income $0.20 billion $3.21 billion
    Net income $0.04 billion $3.05 billion
    Operating cash flow $0.07 billion $3.50 billion
    Free cash flow ($0.05 billion) $2.91 billion
    Cash & investments $1.18 billion $7.11 billion
    Debt $1.40 billion $2.00 billion

    stock prices $11.28 $239.20
    stock price in 2014 $5.20 $14.75

    Let me also say that owned stock in both companies in 2014. I liquidated my holdings of AMD long ago, but still own Nvidia.
     
  22. cube macrumors Pentium

    Joined:
    May 10, 2004
    #22
    If you look at the HIP and MIOpen projects, you will see that they are basically working around the sandbagging of OpenCL.
     
  23. jerryk macrumors 601

    Joined:
    Nov 3, 2011
    Location:
    SF Bay Area
    #23
    I will take a closer look. However, on first view it appears to be an AMD effort.
     
  24. leman macrumors G3

    Joined:
    Oct 14, 2008
    #24
    To the original question: probably 2018.

    However, I very much suspect that the programming paradigm will change within not too distant future. The CPUs will probably move towards high-bandwidth memory and will adopt data-parallel programming techniques even more then now. We could even get high-bandwidth memory that is shared by both CPU and GPU. This will create new challenges, but the payoff will be very high as the processing throughput will increase dramatically.
     
  25. robvas macrumors 68030

    Joined:
    Mar 29, 2009
    Location:
    USA

Share This Page