10-26-2012 08:39 AM
83 1234
tools
  1. freestaterocker's Avatar
    From what I understand, no current quad core processors have onboard LTE, so the LTE modem would have to be added separately. These separate LTE modems (such as the ones used in first generation LTE Android phones) use more battery power. Thus, the dual core/LTE chipsets are a power saving measure. At least that's what I've heard.
    The yet-unreleased S4 Pro does. But they're still working out power-management to make it functional for phones.
    10-02-2012 07:34 AM
  2. SnailUK's Avatar
    But if a killer app was designed to run on as many cores as possible you would see an immediate rush of people clamoring for a quad core phone.
    But number of cores is just like saying number of horsepower on a car is the only measure of performance.

    The iPhone 5 is "only" a 1.3ghz dual core, yet runs like an absolute rocket, due to using the right components, and spending the time to get the hardware optimised correctly.

    or put simply

    Cores != performance
    10-02-2012 07:40 AM
  3. power5's Avatar
    Rocket that boots slower than a S3.

    iPhone 5 vs Galaxy S3 boot up time - SiNfuL iPhone

    10-02-2012 07:49 AM
  4. AngryNil's Avatar
    Rocket that boots slower than a S3.
    Boot time is quite frankly one of the most awful "measures" of performance.
    10-02-2012 08:04 AM
  5. power5's Avatar
    Not sure why a synthetic benchmark is so much better than a real life benchmark. I hope you know that companies like Nvidia have released drivers SPECIFICALLY for doing better in synthetic benchmarks. So, would that be a better measure? Everyone uses 3dmark to test PC GPU performance. Its a set program that any company could deliver perfect drivers for, and they do. So, you will take that result as gospel, but not a real time boot test?
    10-02-2012 08:23 AM
  6. freestaterocker's Avatar
    Rocket that boots slower than a S3.

    iPhone 5 vs Galaxy S3 boot up time - SiNfuL iPhone

    And my 1ghz single core HD7 with mango boots faster than an S3...
    10-02-2012 08:37 AM
  7. power5's Avatar
    Awesome post the video. Would love to see that.

    But, we all know that a wp7 device has no chance of running wp8.
    10-02-2012 08:43 AM
  8. sentimentGX4's Avatar
    I appreciate your insight along with your always staunch opposition to Quad-Core phones.
    In 2-3 years when Quad-Core Mobile Processors (Smartphones) are widely prevalent, I expect you to lead the charge against Hexacore devices of the same type. Because surely we can't have our mobile computing devices breaking boundaries, getting smaller, or becomming more energy efficient.

    1912 Forever!! That is all.
    That's a misguided thing to say. Snapdragon S4 dual cores are currently just as fast as ARM Cortex A9 Quad Cores, if not faster in most regards.

    Demanding Tegra 3 quad cores over Snaodragon S4 is like asking for 1912 as well. Not only because the Tegra 3 isn't faster than Snapdragon S4 but because the Tegra 3 is built around older technology.

    I really wish spec junkies would inform themselves better. There is nothing particularly desirable about the current crop of quad core mobile processors.
    10-02-2012 09:21 AM
  9. CSJr1's Avatar
    That's a misguided thing to say. Snapdragon S4 dual cores are currently just as fast as ARM Cortex A9 Quad Cores, if not faster in most regards.

    Demanding Tegra 3 quad cores over Snaodragon S4 is like asking for 1912 as well. Not only because the Tegra 3 isn't faster than Snapdragon S4 but because the Tegra 3 is built around older technology.

    I really wish spec junkies would inform themselves better. There is nothing particularly desirable about the current crop of quad core mobile processors.
    What part of breaking boundaries, getting smaller and more efficient did you not understand? If you did understand, why would you put up a Tegra 3 40 nm against an S4 28nm or heaven forbid a S4 Pro 28 nm?

    What you fail to realize is that I, contrary to your opinion of "spec junkies", know that quad core is not useful NOW. But, stop living in the present or you may find you become one of those that stiffle innovation and developer drive due to the plethora of status quo. Sure, there is nothing the LG Optimus G can do now that any Dual core couldnt do. (Besides shoot lights out in benchmark tests thats not a useful application) But What If.. What if a WP8 Quad had native C++ code for video transcoding and fully used the Adreno 320? Could we hasten the end of the desktop? What if.. ..
    Last edited by ebsn; 10-02-2012 at 12:26 PM.
    10-02-2012 12:13 PM
  10. a5cent's Avatar
    All OS are more optimized than people think.
    Then I will state that some OS' are a lot less optimized than you think.

    I can leap into a lengthy post pointing out hundreds of situations where Android devices lack any type of optimization (on most handsets), but I'm not sure that would be appreciated.

    Instead, simply consider this: Any optimizations Google makes to Android must be hardware agnostic, as Google may not make any assumptions about what types of hardware manufacturers ultimately decide to slap together. That is why stock Android generally fails to exploit much beyond the general purpose CPU cores. That is why giving Android devices two general purpose CPU cores, instead of just one, results in such a huge performance gain.

    It's not even just about optimized hardware utilization, but also about optimizing the OS for the types of use-cases it is expected to encounter most often. Android utterly fails in this department too. Honestly, that we even need something like "project butter" on a dual-core 1GHz platform, just to get smooth UI performance, is so utterly ridiculous it is not even funny. Anyone implying this is anything but unoptimized "slop" is simply... I don't know... pick your own word ;)

    Edit: Personally, I think it is fair to label Android as bloatware.

    Watching the Nokia phone app that edits video and pictures in real time, that could very very easily work faster with a quad core. So look at the software that Nokia is making with DC, what do you think they could do with a hex core ARM?
    That would be utterly inefficient. That is what SOC's have specialized media processing cores for. WP8 devices actually have seven cores, not just two, but if the OS is to make use of them, it needs to be able to rely on them existing, which Android can not do to the same degree (see this post for details).
    But if a killer app was designed to run on as many cores as possible you would see an immediate rush of people clamoring for a quad core phone.
    Yes, but only if no dual-core SoC exists, that is capable of running that thread-friendly app just as fast, with the added benefit of running those apps with two or less threads even faster. As I've explained multiple times, that is the situation we can expect to encounter in the smartphone space for the foreseeable future, because all smartphone SoC's are constrained to the same power and thermal restrictions no matter how many cores they have. Read my thread above for more info.
    Last edited by a5cent; 10-02-2012 at 02:56 PM.
    power5 and Gaichuke like this.
    10-02-2012 02:21 PM
  11. cckgz4's Avatar
    It also doesn't have a lightening dock connector. We are so behind in this smartphone game
    a5cent likes this.
    10-02-2012 03:04 PM
  12. power5's Avatar
    Then I will state that some OS' are a lot less optimized than you think.

    I can leap into a lengthy post pointing out hundreds of situations where Android devices lack any type of optimization (on most handsets), but I'm not sure that would be appreciated.
    (see this post for details).

    Yes, but only if no dual-core SoC exists, that is capable of running that thread-friendly app just as fast, with the added benefit of running those apps with two or less threads even faster. As I've explained multiple times, that is the situation we can expect to encounter in the smartphone space for the foreseeable future, because all smartphone SoC's are constrained to the same power and thermal restrictions no matter how many cores they have. Read my thread above for more info.
    I am not overly experienced with android but the very few times I have used it, once being a S3 in att, it was not laggy at all. Other times were on Droid?? devices from a couple years ago. None were probably rooted or running too advanced versions of droidOS either.


    By more than DC I simply mean double what we have now on the chip, or triple for a hex core. So everything is doubled, not just 4 CPU cores. No way a faster clocked "DC" will outperform a slower "QC" if the app utilizes as many "cores" as it can use. So if it can only use the media processor then it will have double the amount in a QC than it does in a DC. I do not want a unbalanced quad with 4 CPU cores and forget the rest of the SoC architecture.
    10-02-2012 04:00 PM
  13. a5cent's Avatar
    By more than DC I simply mean double what we have now on the chip, or triple for a hex core. So everything is doubled, not just 4 CPU cores. <snipped> So if it can only use the media processor then it will have double the amount in a QC than it does in a DC. I do not want a unbalanced quad with 4 CPU cores and forget the rest of the SoC architecture.
    Hey Power5, I understand what you are saying. It would be great if that was how things worked, but it isn't. For example, all the DSP's and all the media processors are pretty much exactly the same across all of Qualcomms Krait based SoC's, no matter how many cores any specific SoC has.
    No way a faster clocked "DC" will outperform a slower "QC" if the app utilizes as many "cores" as it can use.
    This is exactly the main error in judgment many people are making. It is understandable, but it is flat out wrong.

    The main error people make, as always, is assuming all cores are equal, when they are anything but. It is entirely possible to have a quad-core and a dual-core CPU, both clocked at 1.5GHz, and each running the same, highly parallel app (which can use any number of cores), and end up with the dual-core performing much better. All this means is that an individual core in the dual-core CPU is doing much more work per clock than an individual core in the quad-core CPU. Simple as that.

    This benchmark comparison is an example of precisely that. In it we see the dual-core MSM8260 (the same SoC coming to most WP8 devices) executing the multi-threaded linpack benchmark almost TWICE AS FAST as its quad-core competitor, even though the quad-core is running on ALL FOUR CORES in this test! According to your statement this should be impossible, but it is clearly happening, not just in theory.

    I've hinted at the reasons why this won't really change much in the mobile space in my previous post (#20).
    Last edited by a5cent; 10-02-2012 at 05:47 PM. Reason: Spelling
    10-02-2012 05:36 PM
  14. cckgz4's Avatar
    The first lte us quad core phone will be the one x+
    10-02-2012 06:41 PM
  15. power5's Avatar
    That is 2 different architectures. That is the same as people comparing performance per Clock speed of AMD vs Intel. Intel is always faster in this test. Their single threaded performance always destroys AMDs. So AMD has started putting 6 core chips against intel quads and have gotten much closer to intel performance. I am not doubting that the S4 snapdragon DC is better than the tegra3 QC. But I will bet that the S4 pro is much faster than the S4 DC in that test. I thought I mentioned in this thread that architectures had to be the same, but that must have been the other thread talking about similar topic.

    Edit, in x86, AMD needs 50% more cores than intel to compete and its at a lower price point as well. Compare that to the tegra vs snapdragon and tegra would need to be a 6 core to compete with DC snapdragon. That is how inferior tegra architecture is. The point is to get it to truly be an experiment to see if more cores are better than fewer cores. So this is easy in the x86 world on desktop. I can turn off as many cores as I want. If I have a 6 core intel I can turn on 4 and have a DC intel and show how much of an improvement the extra cores make in multi threaded tasks like renderings and transcoding. Its a huge difference and the more cores ALWAYS wins. Its just simple reasoning. An S4 Pro would run faster than a DC S4 in anything that can utilize unlimited core resources. I would hope no one disagrees with that.
    Last edited by power5; 10-02-2012 at 08:05 PM.
    10-02-2012 06:45 PM
  16. a5cent's Avatar
    I am not doubting that the S4 snapdragon DC is better than the tegra3 QC. But I will bet that the S4 pro is much faster than the S4 DC in that test. I thought I mentioned in this thread that architectures had to be the same, but that must have been the other thread talking about similar topic.
    Yes, absolutely. Two different architectures. Obviously, I also agree that the 4c Krait will be faster than the 2c Krait, provided we get apps that actually know how to make use of those extra cores (a big "if"). It must be that way, since all the Krait cores are exactly the same. BUT! The 4c Kraits use twice the transistors and have almost twice the power envelope of the 2c Kraits. Qualcomm recommends against using the 4c Krait SoCs in smartphones for exactly that reason. It's also why none of the 4c Kraits come with built in radios... they're WiFi only, intended for use in tablets that can provide sufficiently large batteries to power them.

    Of course, that won't stop manufacturers from trying to put them in a smartphone anyway (having four cores sells). And luckily, they will get away with it, for the obvious reason that those two extra cores will almost never get fired up and thus never draw power for any notable amount of time.

    Assuming the purely hypothetical scenario, where all of us started installing web-servers or started doing CGI rendering on our smartphones, that house of cards would come tumbling down pretty fast and manufacturers know it. Obviously, manufacturers need a way out of that sticky situation should it ever crop up anywhere, and their solution is simple: add a "feature" allowing you to turn those two extra cores off completely. So, you get the bragging rights for having four cores, but only as long as you never find a way to actually use them, at which point you will be turning them off (unless you are happy with three to four hours of battery life per charge).

    Despite all that, one good reason remains to get a 4c Krait based WP device, namely the Adreno 320 GPU that comes with it. It is far superior to the Adreno 225, but that is another topic entirely.

    Of course, should we ever get smartphone batteries that are capable of driving a 4c Krait "on all cylinders" for any reasonable amount of time, then any CPU designer worth their salt will be able to take that same amount of transistors (as used by a 4c Krait) and invest them into a new 2c CPU design, a new CPU architecture if you will, that will easily keep up with the 4c Krait, with the added benefit of being much faster with apps that use two or less threads. That's simply a repeat of the situation I refereed to in my last post, but with the 4c Krait taking the place of the 4c Tegra3.

    You understand what I'm getting at?
    Last edited by a5cent; 10-03-2012 at 12:40 AM. Reason: Oops... house, not deck of cards
    10-02-2012 08:14 PM
  17. power5's Avatar
    Of course and it makes sense and is the same in x86 land. DC of the next architecture (not SB vs IB which just changed process size) is usually at least as fast as the QC of the previous gen. But what usually happens is they do not get the same binning. By the end of the process the binning is much better and the chips are capable of running faster and faster clocks at the same voltage. So it is not unusual for the early chips to require much more power and thus conversely heat to reach the same clocks as the later binned chips. So, you can either take a higher binned QC krait and take some power away from it, and turn the speed down a few hundred mhz and have the same TDP as the first gen S4 DC Krait. Obviously the single threaded win will go to the DC with faster clock, but the multi threaded win will of course go to the QC. And in the end, it used the same power as the DC because its a better piece of silicon vs the immature DC. Now of course the DC chips will get better and better as the process matures as well. So a mature process DC will have lower TDP than the same maturity QC and we are right back to what you said. Best option for QC is to throttle it way way down when not being used and the standby time should be the same as the DC. Then when its needed it throttles up. No need to have manual user interaction with the number of cores running, Qualcomm should be able to make that an automatic part of the bios in the chipset bios.
    10-02-2012 08:31 PM
  18. independentvolume's Avatar
    Unless playing intense games, I think an 8s will handle day to day tasks as good as an 8x. This isn't jumbled android were trying to run, it's streamlined wp8.
    10-02-2012 09:04 PM
  19. PG2G's Avatar
    Despite all that, one good reason remains to get a 4c Krait based WP device, namely the Adreno 320 GPU that comes with it. That is far superior to the Adreno 225, but that is another topic entirely.
    There is a dual core variant of the S4 Pro also, which I think is what will be powering some of the WinRT tablets.
    10-02-2012 09:10 PM
  20. CSJr1's Avatar
    It's also why none of the 4c Kraits come with built in radios... they're WiFi only, intended for use in tablets that can provide sufficiently large batteries to power them.
    Qualcomm confirms S4 Snapdragon Pro SoC in LG Optimus G | ZDNet

    "In addition to this, the S4 Pro supports cameras up to 20 megapixel, and includes 3 camera support and stereoscopic 3D capture support. On top of that, the S4 Pro will allow seamless toggling between 3G, 4G and LTE networks, reducing dropped calls when on the move, a feature that is exclusive to the S4 Pro."
    10-02-2012 09:20 PM
  21. a5cent's Avatar
    It's also why none of the 4c Kraits come with built in radios... they're WiFi only, intended for use in tablets that can provide sufficiently large batteries to power them.
    Qualcomm confirms S4 Snapdragon Pro SoC in LG Optimus G | ZDNet

    "In addition to this, the S4 Pro supports cameras up to 20 megapixel, and includes 3 camera support and stereoscopic 3D capture support. On top of that, the S4 Pro will allow seamless toggling between 3G, 4G and LTE networks, reducing dropped calls when on the move, a feature that is exclusive to the S4 Pro."
    Okay ebsn, I'll take you up on this one. You're obviously implying that Qualcomm's Krait based quad-core SoC (the author of the article you've linked to mistakenly calls it the S4 Pro) comes with built in radios and is thus targeted at smartphones. I apologize in advance if I've misread your intent, but otherwise I have this advice to offer:

    Don't rely on tech bloggers. Most make more mistakes in a single article than you will in a lifetime. The author of the article you've linked to is apparently one of these "unprofessionals", as he constantly confuses marketing labels (used by Qualcomm to segment their product offerings) with actual SoC's. I'll use the quote you supplied as an example:

    "On top of that, the S4 Pro will allow seamless toggling between 3G, 4G and LTE networks"

    As I mentioned, the term "S4 Pro" actually refers to a group of SoC's. The "MSM8960 Pro" is one of the SoCs in that group, and it does exactly what you've quoted. However, it is not the quad-core SoC the author mistakenly thinks it is. The only quad-core SoC in the the S4 Pro lineup is the APQ8064, which comes without radios because it is targeted solely at tablets, just like I said.

    The author confuses himself into that same mistake over and over again throughout the article. Instead of dissecting it point for point, I'll just give you the facts:

    The LG Optimus G is based on two separate pieces of silicon:
    1) Qualcomm APQ8064 (quad-core Krait SoC without radios, not intended for use in smartphones)
    2) Qualcomm MDM9615 (LTE baseband modem)

    Here is a somewhat more respectable/reliable source on the LG Optimus G. Here is a spec sheet directly from Qualcomm with a high level overview of their S4 lineup.
    Last edited by a5cent; 10-03-2012 at 12:04 PM. Reason: Spelling
    10-03-2012 12:25 AM
  22. a5cent's Avatar
    There is a dual core variant of the S4 Pro also, which I think is what will be powering some of the WinRT tablets.
    I haven't heard that yet. It's certainly possible, but I would be surprised because tablets that include modems (which dual-core S4 Pro's do) are usually of the high-end variety, and those I would expect to come with a high resolution display that really shouldn't be driven by an Adreno 225.

    I'm expecting all Windows RT tablets using Qualcomm silicon to be built around the APQ8064, but that is just a guess.
    10-03-2012 12:36 AM
  23. a5cent's Avatar
    Best option for QC is to throttle it way way down when not being used and the standby time should be the same as the DC. Then when its needed it throttles up. No need to have manual user interaction with the number of cores running, Qualcomm should be able to make that an automatic part of the bios in the chipset bios.
    No. This "feature" has nothing to do with standby time. Qualcomm has that more ore less covered. The ability to turn those two extra cores off is included due to issues with power draw under full load. Should anyone actually have an app that is able to saturate all four cores, then the smartphones battery won't last long enough to be considered useful. The SoC obviously shouldn't power down under full load, but you will be given the ability to shut two cores down anyway.... in the interest of getting acceptable battery life. Of course this will be sold as a feature, which is kind of comical.

    Of course and it makes sense and is the same in x86 land.
    Okay, I'll take your word for it... ;) ...even though your views are centered around process maturity, clock frequencies and binning which are certainly important issues, but not relevant to the argument I'm making. I will dispute one thing though:

    Not much about mobile ARM CPU's is the same as in x86 land! Like I said... Intel targets the largest possible die-size they can economically mass produce, and cut away cores and cache to get to cheaper SKU's from there. That is why x86 CPUs with fewer cores are always cheaper, always have less transistors and always have a lower TDP (that is precisely why they exist as a CPU in the first place). None of that applies to ARM CPU's. Using the same manufacturing process, and given the same power restrictions (a.k.a the same transistor budget), two and four core designs will have similar prices, similar TDPs and similar overall performance (provided apps can use all four cores). Hitting different price points with mobile ARM CPU's is achieved by paring them with different GPUs, including modems, and of course binning... that last point is really the only similarity to the x86 world.
    Last edited by a5cent; 10-03-2012 at 01:51 AM. Reason: Spelling
    mmacleodbrown and power5 like this.
    10-03-2012 01:30 AM
  24. power5's Avatar
    Thanks for the chip list. Thats great.
    Got any picture links of the S4 silicon, I cannot find any? I just don't see why ARMs would not be able to laser cut bad areas or just deactivate them in the chip bios.
    10-03-2012 07:51 AM
  25. PG2G's Avatar
    I haven't heard that yet. It's certainly possible, but I would be surprised because tablets that include modems (which dual-core S4 Pro's do) are usually of the high-end variety, and those I would expect to come with a high resolution display that really shouldn't be driven by an Adreno 225.

    I'm expecting all Windows RT tablets using Qualcomm silicon to be built around the APQ8064, but that is just a guess.
    Both the Samsung Ativ Tab and a Dell tablet are using Snapdragon S4. Though you might be right and it is not a Pro. Can't find any details
    10-03-2012 08:13 AM
83 1234
LINK TO POST COPIED TO CLIPBOARD