All iPads Future iPad CPU

Discussion in 'iPad' started by Silver78, Aug 24, 2014.

  1. Silver78, Aug 24, 2014
    Last edited: Aug 24, 2014

    Silver78 macrumors 6502

    Silver78

    Joined:
    Aug 24, 2013
    Location:
    Denmark
    #1
    The step towards haswell and now broadwell 14nm give great overall performance boost options.
    Will the step from 28nm down to 14 nm give the same big overall boost option for ipad?

    I think if the ipad cpu already use so little power that the boost will not be that great?



    If a5 was 45/32nm a6 32nm and a7 28nm. Gave 3x and 2x performance?

    What nm is the a8? And How much to expect from this a9 supposed 14nm?
     
  2. TommyA6 macrumors 65816

    Joined:
    May 15, 2013
    #2
    A8 will be 20 nm, so it will be more power efficient than the 28 nm A7, but it will not be huge difference. If A8 were to keep the identical design (clock speed, same number of transistors...) as the A7, it would be a bit more noticeable, but don't except Apple to do that. A8 will be significantly more powerful (up to 2x), with slight increase in power efficiency.

    The same probably applies to A9.
     
  3. Count Blah macrumors 68030

    Count Blah

    Joined:
    Jan 6, 2004
    Location:
    US of A
    #3
    The CPU is not the limiting factor for me on the iPAD, the lack of RAM and constant tab refreshes in Safari are. Couple that with the lack of RAM for true multitasking, and the CPU can be 5nm for all I care.
     
  4. Badrottie Suspended

    Badrottie

    Joined:
    May 8, 2011
    Location:
    Los Angeles
    #4
    2X is not that much. I would like to see 5X or more. Apple is slow to modern technology :apple:
     
  5. Piggie macrumors 604

    Piggie

    Joined:
    Feb 23, 2010
    #5
    My only problem with what Apple say in BIG NUMBERS on screen is like a typical sale in a closing down store that's trying to trick customers, with a UP TO 50% discount.

    You say UP TO 2X FASTER GRAPHICS PERFORMANCE With a GIANT X2 on the screen.

    When what you mean in reality is that a select few, could be a tiny amount of the output coming from the GPU, some calls, some very simple routines, perhaps a line, or a pixel plot is twice the speed.

    For an actual game, which uses all manner of graphics effects to product a screen full of nice looking visuals only runs perhaps 10% faster overall.

    I know why they do this, but it's misleading as all people remember is the BIG X2 words on the screen.
     
  6. Erasmus macrumors 68030

    Erasmus

    Joined:
    Jun 22, 2006
    Location:
    Hiding from Omnius in Australia
    #6
    I think "misleading" is much too strong a word here. Apple are simply stating the facts, that the new device is twice as fast as the old one in terms of raw processing power, and providing a few examples to show it. No, that doesn't mean that in all cases the new device will run twice as fast, simply that it has the capability to, if whoever created the program needs it to, and can use the extra power.

    And remember, if a program is limited for some reason to running at a similar speed to a much slower device, it is very likely that the newer device is performing those same operations using much less power.
     
  7. Piggie macrumors 604

    Piggie

    Joined:
    Feb 23, 2010
    #7
    No No No.....

    You are making the same mistake that Apple want you to make.

    If I say my GPU is UP TO 2X as fast.

    I am not saying it's twice as fast, I am saying some operations, the GPU performs in some circumstances are twice as fast.

    It would like me selling 50 apples and they all rot after 1 week.
    I make a change to the my apple tree and now, 1 or 2 apples stay fresh for 2 weeks.
    Now, I sell you 50 apples and I know that 1 or 2 of them will now last for 2 weeks and not 1 as before.

    I can now say "NEW APPLES, NOW LAST UP TO TWICE AS LONG AS BEFORE" 2X Life.

    I'm telling the truth, but it's not what you think.

    With Apple and their GPU, some graphics routines may be the same speed, some could be slower, many could be a few % faster, and a couple could be twice the speed.

    This may speed games up say 10% on average as the new MUCH faster routines are not very significant to a typical game, just one small aspect of it.

    Apple can still say, and DO day..... UP TO TWICE AS FAST. And they are right, that is true. But it's not what you think it means.
     
  8. AxoNeuron, Aug 27, 2014
    Last edited: Aug 27, 2014

    AxoNeuron macrumors 65816

    AxoNeuron

    Joined:
    Apr 22, 2012
    Location:
    The Left Coast
    #8
    Moore's law is dead. Traditional CPU scaling just isn't what it used to be. Even when they get down to the 14nm node, it's not going to offer anywhere near 2x the performance of the 28nm node because of transistor density increase alone. It's the same reason Intel's Broadwell release has been pushed back for a full year. These transistors are getting so incredulously small that they can't really just make them smaller and call it progress. Quantum effects such as electron tunneling are preventing them from doing so, because on such small scales an electron can literally tunnel through the silicon (which is very bad). So you can cut the transistor size in half, but you're not actually decreasing the power usage at all.

    FinFET and a few other techniques will help but only for another two nodes, at BEST three. Then we're at the 7nm level, very nearly the limit of how small you can make a transistor.

    Hopefully, Apple has some architectural sorcery up their sleeves, design stuff that could make up for it. But it won't give us anywhere near the annual performance increases we've been used to for the last 50 years. Hopefully something like carbon nanotubes comes along and gives us super fast chip performance increases again.

    Truth is, if developers actually made use of all of the new programming technologies immediately upon launch (things like the Metal libraries) with every new release of iOS, you probably would be seeing 2x performance every generation. But they don't have enough time to reengineer their entire project to use the new tech.
     
  9. Erasmus macrumors 68030

    Erasmus

    Joined:
    Jun 22, 2006
    Location:
    Hiding from Omnius in Australia
    #9
    I'm not making any mistake. If Apple say "Now up to 2x as fast" They mean that the processor itself IS twice as fast as the old version. I am perfectly aware that some tasks may not execute in half the time, but I don't care. The capability is there for many critical tasks to execute in twice the speed, whether they are coded in such a way for this to happen (or not) is not Apple's problem.

    What WOULD be "misleading" is if Apple went to a quad core CPU of similar per-core performance, touted it as "twice as fast" and then limited threading to prevent any app from running on more than 2 cores at once.

    In summary, you cannot blame Apple for the laziness of third party programmers. Or blame third party programmers for failing to optimise code that already runs fast enough.

    ----------

    Quoted for truth.

    Silicon won't last much longer. There are fundamental quantum limits to the size of transistors, and we're pushing up towards them already. And we've already hit the limit of how fast you can switch a silicon transistor, we did that years ago. The only practical way of improving speed is to move to more exotic materials that can allow the construction of new transistors that can switch faster, using less power. Probably either optical or CNTs.
     
  10. Piggie macrumors 604

    Piggie

    Joined:
    Feb 23, 2010
    #10
    No. This is totally the wrong way of looking at it. I can't see why you don't understand how this phrase is or can be used.

    Let me TRY again to make you understand.

    I have a store selling many items. Let's say I have 500 items in my store.
    I have discounted say 50% of the items.

    Of the 250 items in my sale that I have discounted the prices on, say:

    5% on 150 of them
    10% on 50 of them
    20% on 30 of them
    30% on 10 of them
    50% on 5 of them
    60% on 3 of them
    80% on 2 of them

    I can quite honestly and legally, place a GIANT sign in my shop window saying MASSIVE Sale, up to 80% Discount in store.

    (Like Apple can and does say UP TO 2X Faster)

    In reality, out of my 500 items I sell, and perhaps out of 500 different operations Apple's CPU/GPU can do.

    Only 2 out of 500 are actually at the figure my BIG SIGN says.

    The bulk of my items, or Apples speed increase are either no faster, or only 5% faster.

    But there are one or a few instructions that are actually up to my claim.

    It's not a lie, but unless you give other figures, saying UP TO 80% sale, or UP TO 2X the speed, means ZERO in reality.

    Apple's CPU/GPU could be exactly the same speed for 499 functions, and 1 function runs at twice the speed, and they can say up to 2X the speed.

    It won't mean anything speed wise to the whole device.

    Note: I'm not saying Apple stretch the truth this far 499-1, but they could, and always just saying UP TO anything is meaningless on it's own.

    It's used as a marketing trick to fool buyers.

    Worthwhile examples, would be to show how much faster the device runs well known things.
     
  11. darngooddesign macrumors G3

    Joined:
    Jul 4, 2007
    Location:
    Atlanta, GA
    #11
    Do you really want a game to literally run twice as fast. Are you a good enough gamer to handle all on screen elements moving twice as fast like someone doubled the speed of a video. Of course not. So its absurd to say that the graphics performance isn't twice as fast because the game only seems to run 10% faster.
     
  12. Piggie macrumors 604

    Piggie

    Joined:
    Feb 23, 2010
    #12
    No, twice as fast graphics does not mean the game is twice as quick.
    It means very basically it could run twice as smooth, same speed but more frames/second.
    Or you could get more detail on screen and keep the current frame rate.

    I have all along just been pointing out that "up to 2X the performance" means pretty much zero without any addition information. and could mean game in reality only have 5% or 10% more usable power.
     
  13. Erasmus macrumors 68030

    Erasmus

    Joined:
    Jun 22, 2006
    Location:
    Hiding from Omnius in Australia
    #13
    TLDR...

    Treating me like an idiot does you no favours. I can guarantee you I'm not. What YOU don't seem to understand is that I don't care that every process won't run at twice the speed on new hardware that has twice the processing capability.

    The fact is that the most processor intensive processes, the ones that take a long time to complete, are generally the ones that benefit most from more power, such as large matrix mathematics, and other complex parallel operations on large amounts of data. It's generally tasks that already complete very fast that are unlikely to benefit, and that is because they are hamstrung by things like memory bandwidth. Or, a game that already runs at 200 FPS is unlikely to run much faster with new hardware, but it doesn't need to. Instead, the developer could add a few new graphics effects, and still keep the game running at 200 FPS, while on older hardware it would slow to 100.

    To use your analogy, it's like your 2 items at 80% off are both fully upgraded Mac Pros, but the majority of the 150 items at 5% off are iPhone cases.

    What you still don't seem to understand is that programmers generally should be able to optimise their code to use more efficient processes that run much faster on new CPUs, but they don't always bother. And that is not Apple's fault.

    And please stop using your stupid analogies. Analogies very rarely work.
     
  14. TommyA6 macrumors 65816

    Joined:
    May 15, 2013
    #14
    The reason you don't see hugely improved graphics every year is because developers tend to optimize their apps/games to run great on older hardware (Currently A5 powered devices).

    There are plenty of benchmarks that measure CPU/GPU performance and each and every year (when Apple claimed 2x performance increase), those benchmarks indeed show those kind of improvements.
     
  15. cfedu macrumors 65816

    cfedu

    Joined:
    Mar 8, 2009
    Location:
    Toronto
    #15
    2 times the performance is 2 times the performance, not 10% not 300%.

    Apple saying that a GPU or CPU has 2 times the performance based on a benchmark means exactly that. If Apple said twice the FPS, encoding time, etc then that would be misleading.

    You made 3 analogies that do not make sense and that is why you are confused.

    If I buy a new car and have 2 options for engine, lets say 200 Bhp and 400 Bhp. I would not expect the 400 Bhp to go twice as fast yet I have exactly 200 Bhp more.

    To truly take advantage of the power you need the car to be optimized otherwise you will only get a small improvement. Aerodynamics, transmission, tires and so on.

    Unless the software is optimized you will not see much improvement.
     
  16. Piggie macrumors 604

    Piggie

    Joined:
    Feb 23, 2010
    #16
    Apple don't say that. That is the whole point.
    They say UP TO 2X the performance, which for the 100th time means nothing.
    Sheesh I'd not idea getting something simple was so dam hard on these forums.

    You could slow any iPad to probably 1 frame a second instantly with most high end software that needs a high end Intel i7 to run it well, with say 8 or 16GB Ram, and 2 to 4GB of video RAM.

    I accept of course you need to take advantage in software of any new CPU/GPU upgrades to make your program run it's best, but UP TO 2X speed means zero, as 90% of your game could be unaffected by these few routines that are running 2X the speed of before.

    The SALE analogy is perfect, saying UP T0 80% off, means zero, without saying anything more. It's not until you get into the store and look around you realize the 80% is totally and utterly misleading without any more details.
     
  17. cfedu macrumors 65816

    cfedu

    Joined:
    Mar 8, 2009
    Location:
    Toronto
    #17



    160 GFLOPS is 2X 80GFLOPS not zero!!!
     
  18. Erasmus macrumors 68030

    Erasmus

    Joined:
    Jun 22, 2006
    Location:
    Hiding from Omnius in Australia
    #18
    We trust Apple not to mislead us, and trust product reviewers to alert us in the event that they do.
     
  19. mangomind macrumors 6502a

    mangomind

    Joined:
    Mar 15, 2012
    #19
    "Nm" (nanometer) is a length. In this instance, this length relates to the size of an electrical component. A smaller electrical component uses less power than a larger electrical component. And making an electrical component extremely small introduces more error and makes manufacturing it less efficient.

    To answer your question, smaller CPUs are harder and more expensive to produce, but use less electricity. Making something smaller makes it smaller, and does not magically change it to make it better. However, Apple can and will improve the CPU's architecture to improve the CPU. But that is separate from the actual size of the CPU.
     
  20. KingKen1986 macrumors regular

    KingKen1986

    Joined:
    Aug 16, 2012
    #20
    "Piggie" you have the brain of a car salesman. When you actually learn what everyone is trying to explain to you it will be mind blowing...until then do us all a favor and don't talk about what you do not know.
     
  21. Silver78 thread starter macrumors 6502

    Silver78

    Joined:
    Aug 24, 2013
    Location:
    Denmark
    #21
    I try again without mention of 2x 3x faster.

    No one missed that 22nm haswell gave great overall performance boosts compared to 32nm sandy and ivy at 22nm (tick).
    Size, cooling, noise, batterylife, performance etc.

    And we are not talking only numbers... You actually could notice jump from sandy/ivy to haswell using the same amount of power right?

    The same is promised with 14nm broadwell.

    Again a5 45/32nm, a6 32nm, a7 28nm.

    Like with jump from 32nm sandy to haswell 22nm .... A8 expected 22nm and even a9.

    In theory then would we see same overall performance boost on ipad going from 32/28 to 22/16/14nm? ... Like 50% or something more batterylife like with macs (i know new ios also helped) ?

    How much does dubble the transistors do IRL?
     
  22. Piggie macrumors 604

    Piggie

    Joined:
    Feb 23, 2010
    #22
    Please note: The term 2X was never used.

    UP TO 2X was the term which means something totally different and is a marketing trap used to confuse consumers into thinking one phrase means something else.

    ----------

    Car Salesman, Steve Jobs, they are one and the same.
    Doing their very best to sell you the product they have on offer and VERY carefully choosing words and phrases to confuse the customer into thinking one thing, whilst they say another.
    Only if you go back and say "This is not what you said" do they then carefully explain the meaning of the actual words used.
    then you realize you misunderstood what was being said.

    Apple say "up to 2X" the speed. with a BIG 2X number on a stand.

    What people hear is "This new model is twice the speed of the old one"

    And that's not the same at all.

    ----------

    To be honest, Intel's latest run of die shrinks and model changes over the past 2 to 3 generations have been nothing short of disappointing.
    People like me and STILL waiting for a reason to upgrade to a worthy successor to Sandybridge.

    A large problem I feel is that Intel have pretty much zero serious competition so they can milk things for as long as they need to at the moment :(
     
  23. cynics macrumors G4

    Joined:
    Jan 8, 2012
    #23
    2X as fast etc is pure marketing, I deal with it daily. A giant 2X is very appealing to consumers.

    Let me ask. What is the standard measurement of a processor that was used to establish it was 2X as fast? We certainly know its not EVERY process because we tested and benchmarked and retested and know that its not 2X at everything.

    Strange argument based around a fictitious measurement.
     

Share This Page