POST A COMMENT

136 Comments

Back to Article

  • solarisking - Wednesday, September 10, 2014 - link

    I am bitterly disappointed with the A8 having only 1GB of RAM, assuming it's true. How does a CPU double from 1 Billion to 2 Billion transistors without adding one of the most customer facing features of all - the RAM!!!! With 2GB of RAM, lots of stuff won't have to be reloaded over and over and over again, especially given the 64bit architecture. It seems like the most obvious feature improvement of all has not been implemented. Damn, I hope you guys are wrong. Reply
  • zaza - Wednesday, September 10, 2014 - link

    I think the RAMs resides outside the SoC if I am not mistaken. I also think that apple left some features for the future iPhones. better CPU better RAMs and better main camera and bigger battery comes to mind. Since they already consider this a big upgrade. screen size and all. Reply
  • solarisking - Wednesday, September 10, 2014 - link

    The RAM is on chip, just like most everything else. You see THIS IS the second generation. The FIRST generation was for get-it-done and we'll improve it later. This is the step whee they're supposed to fix things. This is later!!! It's fucking twice the transistors!!!!!! Dammit, it should be fixed in this product!!!!! Reply
  • kron123456789 - Wednesday, September 10, 2014 - link

    Why increase performance and not increase amount of RAM? That doesn't make any sense. Reply
  • jameskatt - Wednesday, September 10, 2014 - link

    1. Apple would increase the RAM only if it made sense to increasing performance. Don't you think that Apple would want MORE than a 25% increase in performance particularly since RAM is cheap to add??

    2. Since Apple custom designs the A8, obviously their testing showed that increasing RAM from 1 GB to 2 GB DID NOT significantly increase performance on the iPhone 6.

    3. AND increasing RAM from 1 GB to 2 GB DECREASED battery time by increasing power usage.

    Therefore, from an engineering perspective, it makes sense for Apple to not increase RAM.
    Reply
  • GC2:CS - Wednesday, September 10, 2014 - link

    I think is extremely simple. Apple waits for DDR4 power savings to implement 2GB of RAM. No DDR4 no more RAM, simple as that.

    Again A8 is extremely focused on power comsumption so much that even few additional milliwatts were out of the picture.
    Reply
  • bigstrudel - Wednesday, September 10, 2014 - link

    This. If the RAM is not needed for daily use, power savings is more important. Reply
  • ninjaquick - Wednesday, September 10, 2014 - link

    Ram only really draws large amounts of juice when reading and writing. Static address space draws a negligible current, so low that if it is enough to break the power budget, then the design flaws are elsewhere. Furthermore, even at 1GB, the package has a peak R/W, which means that even at 4GB ram, the peak consumption would remain largely the same. Even at 100% usage, as long as the R/W is within the spec limit, consumption simply does not increase.

    Furthermore, NAND access requires more power than RAM activity, so having more memory, meaning less NAND access because apps are kept dormant on ram, means less power is used overall.

    Simply put, the only reason Apple would not increase RAM footprint is to drive costs down in manufacturing. 8Gb lpddr3 are not the cheapest components. They are by no means expensive, but if you ship 15 million units, that totals 112.5 million dollars less profits to double memory capacity. lpddr3 costs about 7.5 dollars per 8Gb unit, and 8Gb = 1GB.
    Reply
  • ninjaquick - Wednesday, September 10, 2014 - link

    http://www.dramexchange.com/WeeklyResearch/Post/2/... source to lpddr3 pricing. Reply
  • vFunct - Wednesday, September 10, 2014 - link

    DRAM needs constant power to refresh every cell. Reply
  • gngl - Saturday, September 13, 2014 - link

    "DRAM needs constant power to refresh every cell."

    Constant (well, more like periodic) but <10% of the total energy budget, isn't it?
    Reply
  • Flunk - Friday, September 12, 2014 - link

    No, I don't think that's it. I think Apple believes they can get away with it and as long as they can get away with it they will. Every cent of money saved increases their bottom line and as a corporation that's their job, maximize shareholder value. Reply
  • McD - Saturday, September 13, 2014 - link

    Kind of. They will get away with it for as long as they can - to benefit the consumer. When they increase the memory footprint the apps are optimised to take advantage of the extra memory which is the death knell for the older devices with smaller footprints. Maintaining the smaller footprint protracts obsolescence and extend the product's useful life.

    Unlike its competitors, Apple used the transition to mobile to remove many of the poor habits adopted by PC products including poor resource management symptomatic of unfettered multi-tasking; on iOS extra memory does present an automatic benefit to the consumer. The previous wisdom of more RAM = faster no longer applies.

    Apples gated approach to RAM increases benefits the consumer but it is time to lose that iPhone4.
    Reply
  • techchief - Sunday, September 14, 2014 - link

    Nothing Apple ever does is for the benefit of the consumer. They are the anti-consumer company. Apple was found guilty of collusion with e-book publishers and was the chief leader for collusion against engineer poaching which kept salaries down for their engineering stuff. Same with NFC. It would have cost them just 10c to have included NFC into the iPhone 5S which also had the Touch ID. As it stands, no iphone out there can use Apple pay but only the brand new one... all for savings of 10c. Apple is the anti-consumer company. They are the greedy company. And if you are a consumer, you wouldn't want any of your dollars going to that kind of a company . Reply
  • mesahusa - Monday, September 15, 2014 - link

    apple isnt just going to shove random shit into their phones like other companies do just to see if it works. as you can tell, google wallet has failed miserably despite a ton of flagship phones having nfc. watch the keynote. apple got dozens of AAA companies to adopt it including macys, mcdonalds, whole foods, etc. its not easy to convince huge companies like those to buy millions of nfc detectors for their stores and its even more work to get the banks to do it. the way apple's nfc works is way different than android's. it doesnt just imprint your card # and security code onto the machine. it generates a random code that goes through many security steps so that you dont have to worry about cancelling all of your cards when you leave your phone at a local diner. nfc, along with just about any other feature requires a backbone to support it. companies had no incentive to support it, because google didnt give them one. now android phones are just gonna dickride on the support apple has set up. so enjoy it, not bitch about it. Reply
  • darwiniandude - Tuesday, September 16, 2014 - link

    What if they're keeping 1GB as the ceiling as it is 'enough' to ensure developers don't simply code and test for the new devices only. Reply
  • HisDivineOrder - Wednesday, September 10, 2014 - link

    To have a bullet for another release next year. How are they going to convince you that upcoming apps require you to buy a whole new phone/tablet if they don't have a hard limit that basically obsoletes your current device?

    If they gave you 4GB ram right away, then why would you go and buy another iPhone next year after all? Already, too many people are sitting out entire cycles and Apple is starting to see declines in the tablet space. The phone space could follow if they don't keep you wanting to upgrade...
    Reply
  • Montrey - Tuesday, September 16, 2014 - link

    Will 2GB mean faster Facebook? Probably not for most users, no more than a 13MP camera necessarily makes better looking pictures than an 8MP camera. My Galaxy S4 has bigger numbers than my wife's iPhone 5 in almost every respect, but apart from screen size, it falls flat in user experience. Reply
  • nsiltane - Wednesday, September 10, 2014 - link

    Ram might be on package but it's definitely not on the SOC die itself. Reply
  • solarisking - Wednesday, September 10, 2014 - link

    After looking it up, I see the RAM for the A7/A8 is not on the SOC itself, but package-on-package (PoP), stacked vertically, for the iPhone. For the iPad it's not PoP but on-board. This would make it even easier to add RAM. I'm guessing they will for the iPhone 6S in a year. Reply
  • extide - Wednesday, September 10, 2014 - link

    The RAM is not on-chip. It is a PoP (Package on Package), which means they could easily offer up versions with 2GB, but yeah, wow, they are really that chintzy! I mean the difference is probably no more than like $15-20!! Reply
  • ninjaquick - Wednesday, September 10, 2014 - link

    http://www.dramexchange.com/WeeklyResearch/Post/2/... it is actually 8-9 dollars.

    That is over 100 million more profits by not going with more memory.
    Reply
  • Stochastic - Wednesday, September 10, 2014 - link

    That's chump change for Apple. Reply
  • NoMixedCoin - Saturday, September 13, 2014 - link

    That's it. They spent that money on U2, none left for the DRAM. Reply
  • lilmoe - Wednesday, September 10, 2014 - link

    You'd think they'll highlight it in their keynote if it came with more than 1GB of RAM just like they did when they upgraded from 512MB... But it surely is odd, not only is more RAM necessary for those who use their smartphones like smartphones are *normally* used (more than 2 apps/tabs open), but it's even more so for a 64bit architecture and higher resolution screens. They don't need 4GB to be relevant, but they surely need more than 1GB. I guess Apple is being Apple, and it leaves them space for future updates to be seem meaningful to consumers... Reply
  • MonkeyPaw - Wednesday, September 10, 2014 - link

    Yeah, if it had a RAM increase, it would have been a slide all to itself. There's little doubt in my mind that it's still 1GB. They have to add something for the 6S, and my guess is it will be more RAM. Reply
  • Brandon Chester - Wednesday, September 10, 2014 - link

    Apple doesn't state the amount of RAM in their presentations for iPhone/iPad. Reply
  • tipoo - Wednesday, September 10, 2014 - link

    When they moved to 1GB, they mentioned having twice as much as the then-current generation consoles. Reply
  • lilo777 - Thursday, September 11, 2014 - link

    "Apple doesn't state the amount of RAM in their presentations for iPhone/iPad."
    Indeed they don't. This is because iDevices always have embarrassingly small RAMs. That's why they prefer to talk about things like "retina HD" displays.
    Reply
  • Laxaa - Wednesday, September 10, 2014 - link

    Some rumors point to the 5.5 inch version getting 2GBs of RAM. If that's the case, they will probably run into some fragmentation issues down the line. Reply
  • iwod - Wednesday, September 10, 2014 - link

    My only guess is that it hurt battery, and LPDDR4 still isn't ready.
    But honestly this sucks, unless they have other tricks to help minimize out of memory crash and reload.
    Reply
  • ninjaquick - Wednesday, September 10, 2014 - link

    Memory capacity generally increases battery life by reducing CPU active time due to file system access in multi-tasking scenarios.

    There is a negligible baseline increase in consumption from power draw to keep data addresses alive, but the biggest consumers of power relevant to memory are the IMCs on CPU, which do not draw more power from increased memory capacity. The choice of limiting the memory footprint is entirely a greed one: Doubling the DRAM capacity would cut into their profits massively. 15 million unit sales would see a decrease in profits of over 120 million dollars.
    Reply
  • BillBear - Wednesday, September 10, 2014 - link

    >How does a CPU double from 1 Billion to 2 Billion transistors without adding one of the most customer facing features of all - the RAM!!!!

    Memory is not on the same die as the CPU. The iPhone has always stacked the memory chip on top of the SOC however in the package on package format.

    http://en.wikipedia.org/wiki/Apple_A7#Apple_A7_.28...
    Reply
  • damianrobertjones - Wednesday, September 10, 2014 - link

    Why add the ram now when, at a later date, you can fleece people AGAIN with minor changes/updates. It makes business sense to keep the ram at 1GB... for now. Sod the customer. Plus saving 1p on every single phone, more or less, will save the company mega $$$$$$$ Reply
  • ninjaquick - Wednesday, September 10, 2014 - link

    It is actually 9 dollars, give or take, to double from 1GB to 2GB (8Gb to 16Gb): http://www.dramexchange.com/WeeklyResearch/Post/2/... Reply
  • jameskatt - Wednesday, September 10, 2014 - link

    Apple focuses on the customer experience. Obviously, increasing RAM from 1 GB to 2 GB did not improve this. Reply
  • Spunjji - Thursday, September 11, 2014 - link

    Except that's patently untrue. Reply
  • lilo777 - Thursday, September 11, 2014 - link

    I remember how Apple was releasing laptops with USB2 ports a year after Windows laptops got USB3 because Intel chipset did not support USB3 yet and getting USB3 required adding a separate controller/chip that cost about $5. Yes Apple is that greedy. No need trying that hard to justify their behavior. Reply
  • vFunct - Wednesday, September 10, 2014 - link

    Adding RAM adds no benefit. It only takes away power.

    Phones aren't meant to be workstations. Phones are defined by how LITTLE they do, not how much more. You save the heavy-duty compute work for your desktop.

    So, given a choice between 1GB and 2GB for a phone, always choose 1GB, and limit what you do with it. If you need more than 1GB for a phone, you're doing something wrong.

    Right now there are no use cases for more memory in a phone. No developers are asking for more memory. We're at a point where phone memory requirements have plateaued, like how CPU performance has largely stayed still over the last 10 years. You just don't need anything more, as no developers are doing anything with it.

    Because phone memory requirements have plateaued, you can expect to see 1GB memory in phones for the next 10 years.

    The only devices where you see more memory are badly designed systems, like Android, where they mistakenly think people want to multitask on a phone. Of course, those systems are terribly engineered and are always big/heavy because they have such huge batteries because of their bad systems design.

    Meanwhile, Apple balanced their design perfectly. Nobody wants power-hungry phones. Ever.
    Reply
  • extide - Wednesday, September 10, 2014 - link

    Wow! What are you smoking!! Reply
  • ishould - Wednesday, September 10, 2014 - link

    "No one will need more than 637 kB of memory for a personal computer"
    "No one will need more than 1GB of memory for a smartphone"
    Give the developers more memory and they will find a use for more memory. Just look at the PS4/Xbox One
    http://wccftech.com/crytek-8gb-ps4xbox-ram-surely-...
    Reply
  • vFunct - Wednesday, September 10, 2014 - link

    Those are only for games.

    Nobody buys their phones to play games. They buy their phones to communicate, and play games when they have nothing else to do.

    You don't build a product for "nothing else to do". You build it for a need first.
    Reply
  • kron123456789 - Friday, September 12, 2014 - link

    "Nobody buys their phones to play games. They buy their phones to communicate" - Oh, yeah? To communicate you don't need THAT CPU and GPU performance. That performance is used ONLY in games. If nobody buys their phones for playing, why increase CPU and GPU performance in the first place? Reply
  • Alexey291 - Saturday, September 13, 2014 - link

    ORLY? I bought my phone to play games on it. I like emulators (with a moga pro controller) and I like the other moga supported games. I also love some of the lovely games on the play store.

    I think you've taken shilling just a touch too far mate.

    I mean c'mon lets remember... When was it that Apple has announced a game exclusive along with its phone? Oh yeah last year. And what about "Real Racing will run on higher quality settings on an Iphone than anywhere else" Oh yeah that was actually just 2 years ago...

    And considering that games are essentially 75% of all large app markets and earn WAY more than any other "productivity" app out there. (Current Play Store top 10 top grossing apps are all games at the time of writing) We can conclude that people don't buy their phones to play games and you are indeed 100% correct. Nobody ever does any of that.
    Reply
  • mkozakewich - Wednesday, September 10, 2014 - link

    Memory definitely hasn't plateaued in the last five years. We only needed 1 GB, and now we need 4 GB. I'd prefer if I had 8 GB in my current system. Reply
  • jameskatt - Thursday, September 11, 2014 - link

    With the logic in this thread I wonder why posters aren't clamoring for 32 GB of RAM on the iPhone 6. After all, there will always be a handfull of geeks that can use that much RAM. It would be double the RAM of the highest end MacBook Pro. Heck, someone would probably port Photoshop to it. Reply
  • Spunjji - Thursday, September 11, 2014 - link

    It's not about wanting more for the sake of more, it's simple mathematics. Apple have increased the memory space required by applications from the 64bit transition and increased the frame buffer space required by boosting the screen resolution. This results in a net loss in usable RAM from the previous 2 generations. This already had readily observable consequences in the 5S and these will worsen with the 6/6+. Reply
  • lilo777 - Thursday, September 11, 2014 - link

    I am genuinely surprised to read posts like this on AT. Macrumors? Sure. But on AT? Reply
  • calden - Tuesday, September 16, 2014 - link

    Ridiculous post, 10 years at 1GB. I work for a bank in Europe as a software engineer, our department is responsible for creating trading apps for our noninstitutional customers, general public. We have some very talented people working for us and we have defiantly hit the ceiling in terms of what we can do because of the limited amount of memory in both the iPad and iPhone. You have no idea what you are talking about, there isn't a single app designer out there who wouldn't want more memory to utilize and to suggest otherwise means your simply a fanboy who has been brain washed to except anything Apple turns out as Gospel. Simply ridiculous, this is about maintaining high margins and nothing more, especially about the silly notion of power consumption as the extra power draw is minimal. Reply
  • Ri73 - Wednesday, September 10, 2014 - link

    I think memory expansions will be comming to iPads and not iPhones for the time being. I actually think 64 bits makes no sense in a smartphone but will enable almost desktop class apps in tablets and that's where you're needing big memory Reply
  • D V - Monday, September 15, 2014 - link

    I wouldn't be surprised that any future product for which they want to turn on multitasking in two windows concurrently will have 2GB RAM. Best candidate: iPad Air and future large iPad... and just maybe the iPhone 6 Plus. Reply
  • Tangey - Wednesday, September 10, 2014 - link

    According to IMG, series 6 -> 6XT gains "up to 50%" improvement on a clock to clock basis, from architectural improvements. Apple also referred to "up to 50% in the presentation.

    So its possible that it is GX6450 with a clock bump. Alternatively it might indicated a G6650 clocking slower than the GPU in the A7.
    Reply
  • darkich - Wednesday, September 10, 2014 - link

    It is the GX6650.
    The sustained performance comes from the GX6650, the Imagination themselves wrote about that GPU keeping its cool even after 50 consecutive benchmark runs.
    Reply
  • anexanhume - Wednesday, September 10, 2014 - link

    The 50% boost is a result of the changes to the ALUs to have 50% more FP16 performance, so if Apple is using theoretical FP16 numbers, then yes, it could be GX6450 at the same clocks. Not sure why this is being ignored. I don't think we have a precedent in their claims that show they always use FP32. Reply
  • lucam - Thursday, September 11, 2014 - link

    I do agree, still don't get why it can't be the GX6450. Reply
  • dj_aris - Wednesday, September 10, 2014 - link

    What is the cost of adding another gig of RAM I wonder? After 4 or 5 tabs in Safari, each visit in every pages reloads it from scratch. That's unacceptable. So much for the "user experience". Reply
  • MonkeyPaw - Wednesday, September 10, 2014 - link

    Well, sometimes there's more too that than just available RAM. The OS can also aggressively free RAM to conserve power. I had a Galaxy Note 10.1 with 3GB of RAM, and Android Chrome reloaded tabs all the time, even after I disabled as many unused Samsung services as possible. I imagine iOS is pretty aggressive at this too, since even the iPads are limited to 1GB. Reply
  • lilo777 - Thursday, September 11, 2014 - link

    I have Galaxy Note 10.1 2014 and I have never seen tabs reloading (some pages do reload but that's because the web site forces an update). And reloading the page probably requires more power than storing it. I believe that looking at across the specs where we see iDevices having less RAM, less storage (at the same price), less antennas, less sensors, smaller batteries than high end smartphones it is obvious that it's all about money. Reply
  • ninjaquick - Wednesday, September 10, 2014 - link

    http://www.dramexchange.com/WeeklyResearch/Post/2/... 8 to 9 dollars per unit. Reply
  • JDG1980 - Wednesday, September 10, 2014 - link

    1GB of RAM on a flagship device in 2014 is just not OK. I don't want the flash memory (or worse, data plan) to get thrashed every time I switch tabs. They should have moved to 2GB at least a year ago - failing to do so now is inexcusable. I like the iOS ecosystem and the clean layout and lack of crapware compared to Android, but such a small RAM capacity completely rules it out for me. Reply
  • Laxaa - Wednesday, September 10, 2014 - link

    1GB last year was the thing that made me stay away from the 5s. As for this one, my girlfirend already signed up for it so I will get plenty of hands-on time anyway. Reply
  • Mycelus - Wednesday, September 10, 2014 - link

    While I also would have liked to see more RAM, you guys need to stop judging the phone by the spec.

    Apple doesn't cheap out or purposely hold out to rip off the user. They put what's fit for the phone's purpose and lifetime.

    Id be willing to bet the 1GB of RAM does a better job than 1.5 GB on an Android device.

    Phones are only meant to be used for 3-4 years. Doesnt matter whether it's an iPhone or a flagship Android phone.

    Stop obsessing over the RAM because its not going to be an issue. If it is, go buy an Android device with all the RAM you can eat.
    Reply
  • funnyferrell - Wednesday, September 10, 2014 - link

    But at this point, 2-3GB of RAM is standard on Android flagships. There's no way 1GB of RAM can do better than 3. Reply
  • extide - Wednesday, September 10, 2014 - link

    It IS and already HAS BEEN an issue! Reply
  • Stochastic - Wednesday, September 10, 2014 - link

    1GB of RAM is going to seem paltry come 2017. Reply
  • lilo777 - Friday, September 12, 2014 - link

    I think I have never seen so many technically flawed posts on AT before. Is this a result of new work that Anand is doing at Apple? I hope not. Safari and Chrome sharing a lot of the code base they would inevitably use the same amount of memory. Then of course iOS being a 64-bit OS, Safari will actually need more memory. Please stop with this nonsense of secret Apple optimizations. They are not even known for a good software in general. They are a hardware company. Reply
  • darkich - Wednesday, September 10, 2014 - link

    I'm guessing that Apple decided to leave more RAM for the iPads, so they can represent a more powerful iOS devices.

    Especially given the announced productivity software alliance with IBM, I think it is safe to expect a beefed up iPads loaded with some serious apps and capable of real multitasking.
    Reply
  • watzupken - Wednesday, September 10, 2014 - link

    Totally agree on the disappointment on the ram. At this stage, nothing is confirmed but I certainly will skip this generation if this is confirmed. Not sure why is it that tough to add that extra 1GB of ram. They are pushing ahead with 64bit processor and better graphics and higher resolution, but all keeping the pathetic 1GB of ram. Reply
  • Gondalf - Wednesday, September 10, 2014 - link

    The average A7 GeekBench result on 64Bit iOS 7 is 1414/2564. The leak is under iOS 8, so che cpu is unchanged and the boost is largely OS related (100Mhz little bump apart).
    Disappointment.
    Reply
  • extide - Wednesday, September 10, 2014 - link

    Nah, that doesn't really make sense. Apple's stuff is all native code anyways, so it's not like there is an underlying java interpreter that could have seen significant performance improvements, which could cause that 25%. It's basically guaranteed to be CPU arch improvements. Reply
  • name99 - Wednesday, September 10, 2014 - link

    You can find the exact numbers in the post above, but the interesting points are that the A8 CPU was stated by Apple as being 25% faster than Cyclone. (No more detail given than that.)
    Put that figure aside and we have a (supposed) Geekbench 3 number from China released about 12 hrs before the keynote that was 25% higher than the Cyclone number --- at this time last year. That checks out, BUT there is one strange element in the mix which is that the Geekbench 3 numbers for Cyclone today are about 8% higher than they were a year ago --- presumably better LLVM, perhaps a very slight boost from the OS/runtime.
    [The version of Geekbench generating the earliest 2013 numbers presumably came out around Sept 2013, using LLVM 3.3. The latest version in the app store is dated April 23, 2014 which I'm guessing was LLVM 3.4. Presumably we've had improvements to LLVM since then, so we may see a further jump in scores once a version is released using the iOS8/LLVM 3.5 tools.

    Geekbench for Windows, as far as I can tell, uses VC++, so I assume there are no 8% or so annual improvements hiding in that compiler, whereas LLVM still has enough known flaws and imperfections that we may have a year or two remaining of big jumps in successive releases.]

    If we take all these numbers as legit (and leave aside the complaining/arguing about exactly WHAT Apple meant by 25% faster) then we get the kinda symmetrical result that the new CPU is 25% faster than Cyclone 2013 split basically equally three ways between higher frequency (1.4 rather than 1.3GHz), better compiler, and "other". "Other" COULD be slight modifications to the core, or it could be the consequence of a better uncore.

    What did Apple get for 20nm given this limited speedup? Basically half the power usage.

    One way to interpret this is that they're (by choice) doing an Intel --- they've hit a performance level they like and are now optimizing for power.
    Another way to interpret it is that they've settled into a tick-tock --- 2014 they optimized power while keeping everything else as similar as possible, 2015 will be back to optimizing performance.
    Another way to interpret it is that the stars of the CPU team were taken off the project to work on the S1 (the SoC for the Apple Watch, about which we know *nothing*) and so this is basically a cleaned up version of the A7, with any obvious problems fixed but no magic or daring applied to it.
    Yet a fourth way to interpret this is that they are now going to be more aggressive with binning, so the phone that was benchmarked gets the lowest binned chips (1.4GHz) while the iPhone 6 Plus (not benchmarked) gets maybe a 100MHz faster, the iPad faster again, and the mythical 13" iPad Plus (which may become real this year) gets the top of the line chips.

    [At 1.4GHz in a lowish-end machine, SB i3 gets single-threaded 1209 vs 1409 for 1.3GHz A7.
    Most of the difference appears to be lack of hardware assisted AES and SHA-1, but Intel fans should not get too cocky. A7 does a better job on things like JPEG and PNG compress/decompress, or Lua --- ie your classic branchy code dependent on unpredictable data.
    (And Geekbench does use these Intel instructions when available, so it's a legit comparison given that Intel chooses not to include these instructions across the entire line.)

    FP results are pretty much equivalent except for FFT where A7 is about 50% faster. I assume they've coded up a fairly naive FFT and so this difference shows better A7 support for the sort of scattered memory access you get in the final stages of a large FFT. But it could alternatively show that (either the code or the compiler being really smart) do a radix-8 FFT on A7 because there are enough registers available, and have to use radix-4 on x86-64.
    A7 ray trace is about 80% faster than i3, but I don't know enough about the critical loop of ray-tracing to know what this tells us.

    Memory performance [at least the streaming access that GeekBench tests] is, of course, the one place where i3 is unequivocally superior, generally about 30% to 50% higher bandwidth than A7.

    One final interesting point is that the GeekBench multithreaded summary numbers are about 1.8x the single-threaded numbers for both A7 and A8. The A7 numbers for the computation benchmarks are pretty uniformly 2x, so we're testing very naive parallelism without even much memory contention. The only place where we don't get 2x is the memory benchmarks where two cores get about 10% higher throughput than one core. I assume this means that the memory subsystem for the A8 has not changed substantially relative to the A7. It's unfortunate that we don't get tests of other multi-core issues (eg speeds of various types of contended and uncontended locks; but being honest these are not especially relevant to mobile as it exists today).

    For comparison, a Sandy Bridge i3 gets about 2.1x scaling, mainly via the boost from hyperthreading whose speedups are surprisingly variable across the different multi-core benchmarks.]
    Reply
  • Alexey291 - Wednesday, September 10, 2014 - link

    So anandtech takes "leaks" and rumours to be real now? Seriously what the hell Anand hasn't been gone a month and its gone down the shitter already? Reply
  • Laxaa - Wednesday, September 10, 2014 - link

    I guess they have some reputable sources and that we will get a deeper analysis once the device is out. RAM size or not, the A8 is a very interessting SOC. Reply
  • Drumsticks - Wednesday, September 10, 2014 - link

    Wow, did you miss the multiple paragraphs where they said "We don't like leaks but we believe this has a strong chance of being true due to the fact that it matches up exactly with Apple's own paper specs"? Reply
  • Alexey291 - Wednesday, September 10, 2014 - link

    We don't like leaks but here is our analysis of this here leak. Because it matches up to some marketing blurb no less!

    Nuff said.
    Reply
  • extide - Wednesday, September 10, 2014 - link

    Do you not understand logic at all? AT's argument here is sound. Reply
  • Alexey291 - Wednesday, September 10, 2014 - link

    The logic is perfectly sound aye.

    "We don't like leaks but we'll do it for clicks!"
    Reply
  • lilo777 - Friday, September 12, 2014 - link

    It looks like you are disappointed with A8 performance. Well many Apple fans are. But that's not AT fault. Reply
  • Alexey291 - Saturday, September 13, 2014 - link

    Yeeeeeeah I'm an apple fan... That'll be why my current apple device ownership is precisely zero... In fact I've never personally owned an apple device though I do end up having to use them occasionaly.

    And how is me being or not being a fan of anything is relevant here? They are using a leak to essentially guess what's inside the a8 soc.

    The only purpose of doing that? Generate traffic. I mean they are even using a picture from mac-effing-rumours as evidence...

    Next article should have them guessing the specs of the apple watch based on the star alignment on the day of the announcement.
    Reply
  • lilo777 - Friday, September 12, 2014 - link

    Apple statements are not leaks. AT is using available data and trying to do their best (and way better than other web sites) and for this perhaps we should be thankful. Reply
  • Alexey291 - Saturday, September 13, 2014 - link

    They are guessing based on conjecture and a leak of a single benchmark. Yeah I'm truly grateful. Reply
  • akile wong - Wednesday, September 10, 2014 - link

    So you guys wanted more ram but i dont think it was worth it, considering the battery gains wouldnt be appearent since extra ram would give same or less battery life compared to previous gen.I'm i wrong? Reply
  • Spunjji - Thursday, September 11, 2014 - link

    That would have to be some seriously power-inefficient RAM. Reply
  • SydneyBlue120d - Wednesday, September 10, 2014 - link

    I'd like to point out that the iPhone 6 is the world first consumer device supporting HEVC encoding :) Reply
  • MrSpadge - Wednesday, September 10, 2014 - link

    Considering the prices of Apple products I don't see how they could fit more than 1 GB of DRAM into the bill of materials... Reply
  • extide - Wednesday, September 10, 2014 - link

    Are you joking? Lol, the extra GB would cost them like $10-20 tops! Reply
  • ninjaquick - Wednesday, September 10, 2014 - link

    http://www.dramexchange.com/WeeklyResearch/Post/2/... 8-9 dollars per unit. But that would be, at 15 million devices sold, over 100 million dollars less profit. Apple only cares about that number, the profit. They can and will sell millions of obsolete devices because they can. And they will laugh all the way to the bank. Reply
  • Igroucho - Monday, September 15, 2014 - link

    Think again! Supposedly u can. If Apple would be that greedy, why did they leave the money on the table for so long time??? Of course they could've cranked out big ass phones months ago but no, they want to do it right! Reply
  • lilo777 - Friday, September 12, 2014 - link

    You use some strange logic. iPhone BOM is the lowest among flagship smartphones and their retail price is the highest. The only thing preventing Apple fromadding RAM is greed. Reply
  • SirMaster - Wednesday, September 10, 2014 - link

    I always thought Apple might not upgrade the RAM as often to reduce application fragmentation.

    Since they want to support the iPhone 4s and iPad 2 still (both are getting iOS 8) increasing the possible RAM would allow developers to make apps that would probably crash or just not run on these 512mb devices.

    Restricting the max RAM to 1GB makes developers take more time and care into keeping memory requirements low so that the 4s and iPad 2 users still have very capable and compatible devices which is an imporatnt aspect to a lot of consumers and to Apple.

    At least it's a lot more than can be said of 4+ year old Android phones and tablets.
    Reply
  • bloodterfly - Wednesday, September 10, 2014 - link

    Pffft I can still run most things fine on my ancient ass HP Touchpad. Granted no intensive games, but I'm not terribly interested in mobile gaming. Reply
  • JDG1980 - Wednesday, September 10, 2014 - link

    You can't stand still forever. Why should new devices be hamstrung for the sake of people who can't or won't upgrade? Reply
  • tipoo - Wednesday, September 10, 2014 - link

    See, that would be less of a problem if they weren't *STILL* selling things with the A5 and 512MB RAM. Reply
  • Speedfriend - Wednesday, September 10, 2014 - link

    A8 multi core score 2920
    Intel Atom 3770D 2899

    hmmm, what will the next iteration of Atom bring
    Reply
  • Achtung_BG - Wednesday, September 10, 2014 - link

    Single core AMD FX-8100 2800Mhz results 1622, single core A8 1400Mhz results 1633 :)
    GPU A7 GFXbench 3.0 off-screen 1080p 800fr->GPU A8 >1200fr. Intel HD 5000 1590fr :)
    Tegra K1 with Denver core 2.5GHz > 1800 single core Geekbench, GPU 1918fr :) :) :)
    Reply
  • darkich - Wednesday, September 10, 2014 - link

    Nvidia Tegra K1
    3400 Geekbench 3,
    4600 Geekbench 2

    Atom will be DOA as always, as well as the obscenely overpriced Core M abomination
    Reply
  • Wilco1 - Wednesday, September 10, 2014 - link

    Those are the quad-core 32-bit K1 scores, not the 64-bit K1 (why create so much confusion by using the same name for completely different SoCs?). Eg. http://browser.primatelabs.com/geekbench3/672893

    The latest Atom is indeed DOA once again as it cannot even beat the ageing Cortex-A15 on an old 28nm process. By the time it becomes available it will have to compete with Apple A8, Nvidia Denver, and Cortex-A57, which will have almost twice the single-threaded performance as Silvermont.
    Reply
  • darkich - Wednesday, September 10, 2014 - link

    Here's some mind blowing information..

    Tegra A15 K1 is matching/beating the lowest power Core i5 4210Y(!!) in Geekbench, also besting it in GPU scores..at less than 50% the TDP, on a higher process, and for maybe 30% the price.

    Talk about the Intel illusion!
    Wake up, people

    For source, look up the wonderful processor comparison database on notebook check site (anandtech is blocking links to it, gee I wonder why)
    Reply
  • darkich - Wednesday, September 10, 2014 - link

    ..to be clear, I'm not implyong they are blocking it because of some crazy pro Intel conspiracy, but because the notebookcheck site is actually a phenomenal source of information and has the best technical reviews on Internet.. hence it could be seen as dangerous competition by anandtech Reply
  • Wilco1 - Wednesday, September 10, 2014 - link

    Yes it is no secret that the low power Haswells are soundly beaten on performance by SoCs at a fraction of the cost and at lower TDP. Note the i5-4202Y would be closer, that scores 4357 at 12W.

    To get to what a quad-core A57/A8/Denver would score you'd need at least i7 4600U at 15W which scores 6603. But then again that CPU alone costs a whopping $400...
    Reply
  • name99 - Wednesday, September 10, 2014 - link

    The Atom value is for 4 cores. That's only interesting in a world where apps utilize all four cores.
    This is NOT the world we live in today.
    Every damn six months I ask the multi-core supporters to give me concrete evidence of significant MOBILE apps that make use of more than two cores, and every six months I hear nothing.
    Reply
  • lilo777 - Friday, September 12, 2014 - link

    You must be iOS user. Perhaps iOS does not show all the processes that run in OS (iOS also historically has had very limited multitasking). In Android you can see dozens of concurrent processes. Is that enough reason for having many cores? Reply
  • kron123456789 - Sunday, September 14, 2014 - link

    Thing is, that I'm not yet found an app that can load CPU of Nexus 7(which is 4xKrait 300 cores at 1.5GHz) more than 50%. But, there are more powerful CPUs now. Reply
  • ciaphuas - Wednesday, September 10, 2014 - link

    For apple its all about profit...if they can extract every dollar they can from consumer they would. That is why they only have 1gb of ram. Here are some numbers crushed by the motley fool.
    http://www.fool.com/investing/general/2014/09/08/1...
    Reply
  • KPOM - Wednesday, September 10, 2014 - link

    I wonder if the iPad Air 2 will ship with an "A8X" chip with 2GB? It seems to me that most of the RAM complaints relate to the iPad, where people use more memory-intensive applications like Office, or leave more tabs open in Safari. Reply
  • gsalkin - Wednesday, September 10, 2014 - link

    Honesty, I would have liked to see Apple use the increased energy efficiency and room for transistors to increase the clockspeed of the cyclone a lot more and then add two A53 cores. A 2+2 chip could have helped in performance and idle efficiency a lot more. Reply
  • davendrak - Wednesday, September 10, 2014 - link

    If the benchmarks are true and the ipad use this soc, the nexus 9 with denver tegra k1 will smoke everything in the market Reply
  • Pissedoffyouth - Wednesday, September 10, 2014 - link

    I'm loving the "more RAM wastes power" argument being thrown around at the moment.

    Yeah because NAND reads and writes are way better on battery life than extra LPDDR3...
    Reply
  • Alexey291 - Wednesday, September 10, 2014 - link

    Ah well you know. For some apple can do no wrong. And all their devices are magical. Reply
  • Morawka - Wednesday, September 10, 2014 - link

    how come you guys didn't live blog the apple event? that was a huge disappointment. I thought the content of this site wouldn't be affected by anand going to apple, but it seems like you guys weren't invited or didn't go because apple thought it would be a conflict of interest? Reply
  • Ryan Smith - Thursday, September 11, 2014 - link

    Previously we only live blogged the iPhone announcements because there wasn't a stream. This year there was a stream, so we expected it to be redundant. We did not count on the stream crashing and burning though. Reply
  • lucam - Thursday, September 11, 2014 - link

    Dear Anandtech, how can be sure its not the GX6450 GPU? This GPU should 50% quicker than G6430 according to IMG. Not sure about your assumption though. Reply
  • mic-RO - Thursday, September 11, 2014 - link

    Does anyone else think there are a lot of unaccounted for transistors on this new die? My thoughts lead me to think there may be some massive new L4 cache lurking in there. Reply
  • errorr - Saturday, September 13, 2014 - link

    For all the crap about only 1GB of RAM and being cheap this is my guess too. That kind of density at 20nm takes some really amazing drawing or something really dense like a MASSIVE cache. I think we saw how the A7 used their on die cache to reduce memory access to improve battery life.

    My guess is that the cost of making the cache far exceeds the cost of additional RAM in a package.
    Reply
  • mic-RO - Tuesday, September 16, 2014 - link

    I've added up the major segments from the A7's poly silicon images, via Chipworks' die photos, and can account for about 1.4Billion transistors. Assuming that the A8 has a 50% bigger GPU via the GX6650's 2 extra core blocks and a slightly beefier pair of Cyclone cores there is still a lot of transistor real estate to put in a bigger L3 SRAM cache and possibly a L4 cache.

    Without knowing, or being able to find any openly available details about, the transistor counts of a high end ARM cores I'm just at a loss to figure out what else but a bigger cache (or 2) is lurking withing those >2 Billion transistors.
    Reply
  • Haroon90 - Thursday, September 11, 2014 - link

    So if it uses half the power yet 25% more cpu and 50% gpu speeds i presume the saved energy is needed to power the bigger higher resolution screens,so where does that leave ipad?

    Ok i know they are going to make it thinner,perhaps the same as the iphone 6 but the ipad is't getting a screen size increase so doesn't that mean that it's totally feasible to expect there to be a return of the AX ipad chips? I can see no reason for apple not to double the clock on the new ipad air,it's even more needed with the upcoming k1 based nexus 9 coming out next month which is 2.4x as powerful as the a7.

    Fingers crossed for a A8X next month,don't let Nvidia take the mobile performance crown Apple!
    Reply
  • lucam - Thursday, September 11, 2014 - link

    Lol, I like you last sentence! :)
    Despite I would like to see an A8X too, I don't think the Ipad Air 2 will have a different soc. But there is possibility to have an A8 with higher frequency.
    I do expect to have an A8X on the Ipad Pro on 2015 though!
    Reply
  • Haroon90 - Thursday, September 11, 2014 - link

    Well if they don't put those power savings into making a faster chip then i would expect the ipad air to be extremely thin and light,perhaps even beating the xperia tablet z2... Reply
  • Haroon90 - Thursday, September 11, 2014 - link

    Also with each release of the ipad it has also been soc performance wise at the top so it would be strange for apple to release a new ipad with it being underpowered compared to the competition. Reply
  • lucam - Friday, September 12, 2014 - link

    Indeed. It did sound strange for me when last year there was no presentation of any A7X. But then thinking back that soc was great (and still it is) for that time being far superior than others around. And also I reckon Apple had invested a lot of resources on A7 and it tried to capitalise this big investment chip in so many devices as it can. Now since the K1, it makes more sense having a powerful soc in the future Ipads

    Now if for argument sake the A8 has the GX6650, then an A8X which GPU should have? The GX6650 is the top of the PowerVr list. Maybe an high frequency GPU? Otherwise I don't have the clue what GPU could be then.
    Reply
  • eric1803 - Thursday, September 11, 2014 - link

    I wonder why so many people are shitting around because of RAM.
    I am doing a hell lot with my iPad Air, which also got only 1gb of RAM and I never ran into issues.
    I had RAM issues with the iPad 1 and some performance issues on the iPad 3 (but that was also because of the weak CPU etc.).
    And I am also not seeing much things being reloaded over and over again on my devices.
    Just safari loves to reload websites, but the same I had on my Note 3 with 3GB of ram.

    I can understand, when the iPad Air 2 will gets 2gb of ram. But I don't care about that 1GB ram on an iPhone. I got some apps, which are laggy, but those are just results of very bad programming.

    Right now the mobile phone world feels like a stupid spec race, instead of doing some development and research on things, that are REALLY needed -> Battery life.
    Reply
  • lilo777 - Friday, September 12, 2014 - link

    Note 3 never reloads web pages. iPhone does. Note 3 also has better battery life (compared to iPhone 5s). What's not clear here? Reply
  • eric1803 - Friday, September 12, 2014 - link

    You know the Note 3 is a huge phone, so OF COURSE it has a better battery life. It will be interesting to see how the iPhone 6 plus compares to the Note 4 in battery life.
    Maybe it depends on the browser if it reloads webpages or not.
    At least I don't care about reloading webpages.
    Reply
  • Torrijos - Friday, September 12, 2014 - link

    I might add that on Anandtech test you can see that battery life / battery size iOS is way more efficient.

    Also, since the number of transistors augmented so much, they might have added more cache, or more channels to all the kind of memories? (using the device's flash more to keep things like web pages (the only instance people cite for the need for more RAM))
    Reply
  • lilo777 - Friday, September 12, 2014 - link

    Increase of cache does not in any way help with the issues caused by RAM limits. Firstly the size of cache is insignificant compared to RAM size. Secondly when data is copied to cache it still remains in RAM. Reply
  • Osamede - Friday, September 19, 2014 - link

    Example: you are at the airport and open some tabs to read when you get on the plane. Open phone on plane after takeoff and RAM-anaemic phone has to reload and you can't read what you has. Happened to me on iPhone 5s....

    This kind of scenario gets even more relevant and frequent when you have bigger screen phones like iphone 6 and 6 plus, because you tend to do more web browsing on a better screen.

    Apple surely knows this, but they are too busy running a McDonalds style business gussied up as a "five star" restaurant.
    Reply
  • bigstrudel - Tuesday, September 16, 2014 - link

    Battery savings 100% of the time by not powering extra ram > Page reloads in one app. Which only effects messy users who never close tabs. Reply
  • lilo777 - Saturday, September 13, 2014 - link

    GSM Arena (http://www.gsmarena.com/iphone_6_a8_gpu_benchmark_... discovered some A8 GPU benchmarks which are just marginally better than A7 scores. Based on this they speculate that A8 might have PowerVR G6630. This is for iPhone 6. 6 Plus might be different. Reply
  • lucam - Saturday, September 13, 2014 - link

    I would be careful with this benchmark. The A7 never was so brilliant in this test resulting behind to the Adreno 330 several times. And even the A8 in this test results slower than the Adreno 330.
    Having said that I have suspicion its not the GX6650, but the GX6450 since it seems to support the ASTC. The G6630 doesn't.
    Reply
  • Embedded Steve - Saturday, September 13, 2014 - link

    Hardware Dudes!
    As the man said: "It's the Software Stupid!"
    LLVM 3.5 was release on 9/3 - Coincidence My A$$ - Check here for all the goodness. Yea, Apple has a HUGE role in this, especially w.r.t. the 64 bit architecture.
    http://llvm.org/releases/3.5.0/docs/ReleaseNotes.h...
    Reply
  • lilo777 - Saturday, September 13, 2014 - link

    Obviously end users don't care where the improvements are coming from but that would mean that Apple were really misleading about GPU performance in their presentation. Reply
  • GC2:CS - Saturday, September 13, 2014 - link

    Series 6 XT offer 50% higher efficiency eg 50% higher performance at the same power plus 6650 would add another 50% thanks to being wider. They could just swap the 6430 for 6450 or they could also use the hexa-core and down clock it by 33% as lower clocked wider GPU is better in terms of power compustion, 20nm is not enough for 50% higher overall efficiency Apple is claiming.

    So far again the A8 looks like to be very focused on lowering the power comsumption which is extremely important in mobile business.
    Reply
  • darwiniandude - Tuesday, September 16, 2014 - link

    REALLY happy we're still going tk get articles like this and chip works teardowns after Anands departure. Keep up the good work!! Reply
  • AppleCrappleHater2 - Wednesday, September 17, 2014 - link

    Worship the holy apple.

    The apple way, selling over expensive crap to stupid consumers that like to
    get robbed.

    This has been a disastrous launch in every respect. The iwatch is such an
    ugly piece of crap, it is truly unbelievable how a company, formerly known for
    its remarkable design, dares to put out such a crap ton of shit. Some
    characteristics are glaringly obvious and inherent to it: over expensive,
    hardly innovative, limited functionality and usability (need of an iPhone to
    make it work), looks exactly like a toy watch and so on.

    There are of course way better smart watches out there, especially form the
    likes of Samsung, Sony, Motorola, Asus, LG, simply put, there is no need for
    another piece of over expensive junk.

    The iPhone 6 is technologically stuck in pre-2011 times, a base model witch
    a capacity of 16GB without the possibility to use SD cards isn't even funny
    anymore. The screen resolution is horrendous, it isn't water proof, shock and
    dust resistant, it offers nothing innovative, just some incremental
    updates over its predecessor, both lacking severely behind their competitors at
    their respective launch dates.

    Now the Iphone 6 Plus offers a „Retina HD“ screen, full 1920x1080p, oh wow,
    where have you been for the past 4 years apple, talk about trailing behind.
    That’s pathetic. The interesting thing about that is the fact that apple
    always manages to sell backwards oriented, outdated crap to its user base, all
    while pretending to be an innovative technology leader. The similarities
    regarding any form of sectarian cult are striking.

    You gotta love how Apple always comes up with new marketing bullshit terms,
    aka "Retina HD", with the intention to manipulate its users while preventing easy
    comparisons with its competitors by withholding the actual specs. Apparently it’s
    not enough to have an 1080p screen, you have to call it "Retina HD" to make those
    suckers buy it, otherwise someone could look at the 4K Amoled and Oled screens
    form LG and Samsung devices and get outright disappointed. Same goes for
    everything else. Every outdated „feature“ needs to get its own marketing label
    to persuade buyers with crappy „experience“ and „usability“ ads, while covering
    the truth with marketing gibberish, knowing full well that only a fraction of
    aforementioned buyers cares to look at the facts and dares to compare them.

    Car engines come to mind. For comparisons shake let’s look at a 1.0 liter, turbo
    charged petrol engine and a V8 compressor. What’s better should be obvious, but
    by calling the former an „ecobooster“, thus giving it a special marketing label,
    this joke becomes a „feature“, something positive that can be added tot the list
    of features of a car.

    By doing so a negative aspect is transformed into a positive one, the
    reality is distorted, non tech savvy buyers are manipulated and comparisons are
    made more difficult (another layer of marketing bullshit to overcome), well done
    marketing department. You see , if something is seriously lacking (of course for
    profit, what else), don’t bother explaining, just give it a nice marketing term, distort
    reality, make it a feature and call it a day. Fuck that!!

    The Apple Iphone 1 and Ipad 1 might have been innovative at their time,
    but since then, the bitten apple has been continuously rotting from the inside
    outwards, always swarmed by millions of Iworms which regale themselves with its
    rotten flesh, not forgetting all other Americans who support apple by means of
    their tax dollars to finance its bought US Treasury/Government bond interest rates.

    Last but not least, every Apple product includes a direct hotlink to the nsa,
    free of charge, something that might make it a good value, after all.

    Ceterum censeo Applem esse delendam.
    Reply
  • Achtung_BG - Friday, September 19, 2014 - link

    Apple A8 CPU test:
    SunSpider ~ 350ms
    Geekbench ST~1600, MT~2900
    Linpack MT~ 1380MFLOPS
    GPU test GFXbench 3.0 off-screen ~ 18 Fps
    Reply
  • alwasih - Wednesday, September 24, 2014 - link

    I've always been curious, why is there a very large gap between the Apple A7 and Snapdragon 800/801 when it comes to single core geekbench scores? Is the Apple A7 really that much more powerful than the Snapdragon 801? Or is it because of the new ARM v8 instruction set used by the A7? The A7 actually attains these high scores while operating at much lower frequencies too. Its astonishing. Reply

Log in

Don't have an account? Sign up now