Qualcomm Employee: 64-Bit A7 Chip 'Hit Us In The Gut'

[tbrinkma]

Going 64-bit is like going with a multi-core CPU for mobile. Will it give you a speed increase in certain situations? Yeah. Will it be faster overall compared to its 32-bit counterpart simply because it has 32 extra bits? No, it won't.

Technically true because of the bolded part, but utterly misleading (and the next paragraph of your post is utterly incorrect).

If Apple were to release a 32-bit A7 alongside the 64-bit one, they'd both perform equally well in about 99% of all tasks normally performed on mobile platforms.

The move to 64-bit means a change to the chip's interface with the rest of the world. Instead of taking two (or more) operations to do 64-bit math, it takes one. (Many 32-bit math operations can be done two at a time, as well.) Additional registers open up. It's things like this that give the 64-bit processors a speed boost over otherwise equivalent 32-bit processors. Some of it *is* "simply because it has 32 extra bits", but most of it is because of other architectural differences between the chips (such as those extra registers).

Perhaps you recall the change from 32-bit x86 processors to 64-bit ones? You know, where the AMD processors of the day *were* available in both 32-bit and 64-bit versions, made on the same process, at the same clock speeds. Identical software compiled for both processors ran faster on the 64-bit chips due to these same sorts of architectural differences.

Of course, this all happened back when Intel was still recovering from the evolutionary dead-end that was the P4, and was betting that the incompatible Itanium & Itanium 2 processors would take over as the 64-bit chip of choice, so it's been a while (10+ years now?). Its entirely possible that you're too young to remember it, or that you weren't as involved or interested in hardware at the time.

----------

When are people going to get that no matter how optimized and improved a 64bit software is, it eats up more RAM than the 32bit counterpart, so it is very essential to add more RAM if you're going with 64 bit hardware and OS.

And no I'm not asking an iPad with 8GB of RAM. But 1GB certainly meagre.

Yep. Memory usage increases with the move to 64-bit pointers. Not as much as you seem to be implying, though.

Each *pointer* will use 64-bits instead of 32-bits. Everything else uses *exactly* as much RAM as it used to. On average, having 64-bit pointers increases RAM utilization by about 5-10% compared to the same application using 32-bit pointers.

Some apps will see pressure based on this increase. Most won't.

----------

Only for applications that are are being bottlenecked by 32-bit architecture, which are rare.

The best way to explain it is that jumping from 32 to 64-bit doesn't offer any performance gains in and of itself, but it's less likely to get bogged down, because it can handle larger amounts of data being thrown at it more gracefully. Like if you have a game that's been compiled for both architectures, the 64-bit version won't have double the framerate of it's 32-bit counterpart. But if you were to start throwing in complex physics models, and higher end AI, a 64-bit processor running the same amount of cores on the same CPU architecture as a 32-bit processor would be able to leverage all that information a little better, leading to slightly improved framerates.

It wouldn't be a massive, game changing, difference, but it'd be measurable. Thought that advantage could be matched by a slightly quicker 32-bit processor. Really, the best places to see the real advantages in 64-bit architecture would be in movie editing with large files, 3D rendering, and the like. Anything that has to do with tons of raw data being tossed around and worked on. Things you're not currently doing on iOS, in other words.

Do you think a 64-bit email app will allow you to type out words faster? Will it send your messages any quicker? What about 64-bit Angry Birds? Will it allow for...what...more birds onscreen? Will it suddenly be better because you can stuff twice the amount of data into a 64-bit register? Hell, these apps aren't even coming close to saturating a 32-bit chip. 64-bit won't do anything.

You should stop. It's pretty clear that you don't actually understand what you're talking about. There's no need to continue demonstrating it.

 

[LordVic]

With such a HUGE lead, the webshare issue still remains. Are these android dumb phones with no web access?

How is that a mobile platform with 80% marketshare ONLY has 33% web share????

To top it off, ICS and GB still dominate 43% of android. 2 revs behind....

Browser tests to determine who is using what platform isn't great. Many users do not use the stock browsers, or even the stock settings on their browsers. Many android browsers can be made to emulate a desktop platform and show up as such. My Chrome for example on my phone is set to show up as Chrome for the desktop. That doesn't include the users who are on dozens of alternative web browsers you can download and use.

This heavily skews numbers. Where Apple pretty much identifies itself no matter what as an iphone browser, you don't have that assurance with Android.

 

[tbrinkma]

No it doesn't! Especially when running something in shiny new 64-bit uses enough extra RAM to make you run out of it.

Here's also one case to destroy the whole "64-bit is better than 32-bit ALWAYS" argument:
int a = 0; for (int i = 0; i<99999; i++){ a++; }
I've measured it to use about double the RAM running 64-bit.

That's interesting, because 'int' is a 32-bit integer on either processor.

 

[kdarling]

Browser tests to determine who is using what platform isn't great.

For all sorts of reasons, including the ones you gave.

One example is if they're counting ad pulls or page loads.

Some browsers are known for constantly reloading pages when the user navigates back and forth.

That characteristic alone could hugely multiply the hits over another browser that displays pages from cache.

 

[LordVic]

For all sorts of reasons, including the ones you gave.

One example is if they're counting ad pulls or page loads.

Some browsers are known for constantly reloading pages when the user navigates back and forth.

That characteristic alone could hugely multiply the hits over another browser that displays pages from cache.

also, if you're counting all iOS devices v Android. iOS has a significantly larger install base if you also include tablets in the mix.

Basically, it's a cool stat to see what browsers and settings people use, But it can never, EVER (for both android and iOS camps) be used to say who's devices are actually more used, even online

 

[tbrinkma]

Isn't this the model that Intel/Microsoft used successfully to kill the vertically integrated models of Apple (desktop) and Sun (server) in the 80s and 90s?

No, the model Microsoft used was to attach themselves to an existing monopoly, and ride along until they were big enough to start beating up smaller players.

It helped that IBM didn't actually foresee how the PC market was going to progress. If they had, they'd have developed their own OS for it, rather than outsourcing it from someone else. That lack of foresight also meant that they didn't think it was a big deal that Microsoft didn't want an exclusivity clause in their IBM contract, so they were free to supply the same OS to other PC makers.

 

[PracticalMac]

LOL. I'm sorry, Not many apps make use of this. Plus, my Note III would eat it for breakfast.

On that, I do like my 5c though. It functions well and it's a pretty blue color that makes me smile.

Not yet.

I am sure few if any developers knew about this, and will take time to compile their software and tweak it for 64.

Samsung makes very competitive products, if only they did not also copy other companies technology and cheat on benchmarks.

 

[tbrinkma]

When are people going to get that 64-bit doesn't say anything about performance?

When are people going to get that with 64-bit certain calculations can be done more efficient, but others significantly less?

When are people going to get that marketing lines are not the same as truth?

Can you name an operation which is less efficient (cycles per operation) on a 64-bit processor? The A7 in particular?

 

[Tinmania]

I for one don't believe that their are paid trolls here. I'm talking about the over the top worshiping of all things Apple.

With the assumption that anyone who disagrees is a "paid troll" a clear indicator that you are likely right.


Michael

 

[tbrinkma]

That's like saying there's no point in putting in a 64 bit CPU in an iPhone until all the available apps are recompiled. Yet everyone is claiming there's a benefit just by having the CPU.

With Android, each manufacturer could start with recompiling the kernel and drivers for 64 bit. That's easy enough.

After that, the big advantage is that almost all Android apps are Java based. That means if the runtime gets recompiled and tweaked, ALL the apps immediately benefit... no recompile necessary by the hundreds of thousands of developers.

Not at all. Until you have an OS that is 64-bit, there's no benefit to having a 64-bit processor, because none of the apps will be able to be 64-bit either.

With Android devices there's several layers, some of which are 64-bit ready, others aren't.

Processor - The only 64-bit ARM chip on the market at this point is Apple's A7. That will change sometime next year.

Linux Kernel - 64-bit ready, not sure whether the 64-bit ARM architecture specific stuff is done yet, but I suspect work started when the architecture itself was announced, and testing can't really begin in earnest until there's real 64-bit ARM hardware out there to run it.

Dalvik VM - Still 32-bit. The last roadmap I was able to find showed 64-bit in 2016. I strongly suspect that work has been bumped up in the queue, but until that's done Android apps won't be able to take advantage of 64-bit.

Android APIs - Still 32-bit. Can't start changing that until there is a 64-bit Dalvik VM.

Apps - Still 32-bit. Can't start changing that until there is a 64-bit Dalvik VM and the Android APIs are updated. This includes both the 'core' Google apps *and* all the 'special' front ends and applications used by the various device makers. The (relatively) few apps which bypass the Dalvik VM layer may be able to take advantage of 64-bit once the kernel changes are finalized, but that's a *small* subset of apps.

Apple's biggest advantage in this space is their ability to coordinate all of the work they needed to do to make the transition.

----------

Technically, only Samsung's fab division is supposed to know about Apple's designs due to heavy confidential information protected by contracts, laws, and so on.

The rest of Samsung aren't supposed to know anything, they are supposed to keep each division isolated or they'd risk losing a lot of money because beside losing Apple, they'll lose a lot of customers who will not use their fabs in order to protect their IPs.

There were news last year that Samsung made this clear to everybody, that their foundry division is fire-walled from their devices division.

Remember, though, there's also been news which made it clear that Samsung's 'fire-walls' aren't as sturdy as they're supposed to be. Remember hearing about how they used confidential contract information only released to Samsung's *outside* counsel in a case as negotiating leverage with a third party? That's a *major* no-no in legal circles.

----------

It's not that simple. There are also some drawbacks. If your apps don't need 32 bits, then going to 64 is effectively having the cache size.

So it can be detrimental to performance as well.

Which is why 64-bit processors are designed with larger caches and wider buses.

 

[gnasher729]

So now we all of a sudden care about specs? Does anyone not see the hypocrisy?

You are misreading it. We are not arguing "the speed increase of 64 bit is important" vs. "no, it is not important". That would be arguing whether we should care about specs or not. No, we are arguing "64 bit is faster" vs. "64 bit isn't faster". It's a totally different thing.

 

[Michael Goff]

That's my point it will never be perfect. You guys spend so much time pointing out what's wrong not really enjoying your products.

----------



Are you talking about the paid Samsung trolls??

I enjoy my products, but that won't stop me from telling these companies when they screw up.

 

[gnasher729]

You're outing yourself, as you are not reading. He specifically mentioned "of the same caliber", as in apples to apples instead of apples to oranges what would be the case when comparing to different generations as you do.

The 64-bit A7 will be more power hungry than a hypothetical 32-bit A7.

It's like most people in this thread are saying the 64-bit A7 is faster than the 32-bit A6, ergo 64-bit is faster than 32-bit. No. The A7 is faster than the A6.

You are of course wrong. For example, AArch64 has thrown out most of the predication, which limits clock speed, costs power, and achieves less than people thought 20 years ago (while leaving some special cases that gain 99% of the advantages at minimal cost in the processor). There are plenty of other changes. There are changes that can't be made in 32 bit because of compatibility reasons, but they can be made in AArch64. String processing is faster (and therefore uses less power) due to 64 bit instructions. Public key encryption / decryption is massively faster due to 64 bit instructions. And some encryption / decryption now has hardware support, so you get massive power savings there.

----------

Because for the most part it doesn't.

Unless you are dealing with big numbers over 4 billion, The performance gains from the chip is often loss to the Slower Bus having to read in 64 bit commands vs 32 bit commands.

Now if you device can use more RAM, that the 64 bit system can address then you get the performance back.

AArch64 uses 32 bit instructions, just like ARM, so that argument is nonsense. The number of bytes read in one cycle is totally independent of the bitness. Exact numbers for ARM are hard to find, your typical Intel processor reads and writes 256 bit at a time.

----------

This is exactly what ARM did. Remember the ARM 32-bit instruction set is nearly 20 years old (1995). The 64-bit instruction set was announced in 2011 and released in 2012. It's the first majorly used CPU instruction set in decades designed from the ground up to be low power without trading performance.

The kicker is it will not benefit old apps still compiled against the 32-bit instruction set at all.

In maybe two years time I'd expect a chip where 32 bit apps will actually slow down. There are plenty of 32 bit instructions today that are limiting the clock speed, and which have been removed from 64 bit. If these instructions are changed to use two clocks instead of one, then the clock speed of the processor can be increased without any change in the technology used. Today that wouldn't go down well, but in two years time 32 bit code will be a tiny fraction of all code running.

 

[kdarling]

With Android devices there's several layers, some of which are 64-bit ready, others aren't. ... (snip)

Thank you for taking the time to lay them out.

Apple's biggest advantage in this space is their ability to coordinate all of the work they needed to do to make the transition.

Yep.

This conversion discussion is relatively short term, though, as Samsung says they'll be selling 64 bit smartphones by this time next year.

Then everyone can debate the relative specs. Whoa! Deja vu. I feel like this all has taken place before on the Internet. Like a bazillion times? lol

Regards.

 

[AppleScruff1]

With the assumption that anyone who disagrees is a "paid troll" a clear indicator that you are likely right.



Michael

There are a few here who throw that term around loosely. They seem to think that anyone who says anything critical of Apple works for Samsung, Microsoft or Google.

----------

yep just like Android users do not use the web. I don't buy it.

----------

but there are pleanty.

Have not seen any of those over the top Apple worshipers, u know and I know they don't exist just plain old Apple enthusiast here.

I have no doubt that some users are here just for reaction and to stir the pot, but I don't think that Apple, Samsung, Google, or Microsoft pays anyone to be here.

 

[nviz22]

That 64 Bit SoC is something that was unexpected and unnecessary. Apple's benefit is that it can have a device that can run efficiently with little need of processing power, hence why no 2.3 GHz processor with 2-3 GB of RAM. The 64 Bit SoC is quite powerful, but it can never be utilized properly until it is the standard for devices. When it's the standard in a couple of years, Apple might need to upgrade the RAM just to smooth out the experience more. I think the iPhone just needs a bigger screen, a much larger battery capacity, better resolution on its screen, a better camera, 2 GB of RAM in the iPhone 6, and some sort of sound quality like Boomsound/ Beats Audio of the HTC One. The HTC One is the exact Android version of the iPhone in my eyes, superior build quality, great looking phone, not the fastest processing power, but efficient and smooth, has unique features, great audio sound for music, etc.

 

[giantfan1224]

I have no doubt that some users are here just for reaction and to stir the pot, but I don't think that Apple, Samsung, Google, or Microsoft pays anyone to be here.

You might be right but it could also be a little naive since Samsung has been outed for this already.

 

[subsonix]

and some sort of sound quality like Boomsound/ Beats Audio of the HTC One.

The size and placement of the speakers in a phone is an assurance that sound quality will never be anything to write home about. It's there as a necessary utility, but to expect any kind of sound "experience" is quite ridiculous.

 

[TouchMint.com]

"Apple kicked everyone in the balls with this..."

Awesome haha.

 

[intrepidsoft]

It's more than that...

View attachment 452221

... but it's all meaningless, because web (or ad) share has no correlation with ownership numbers or usage.

1) Such stats include tablets. Heck, if you compare iPad vs iPhone web usage, you would think that iPad sales outnumber iPhones by several times.

In other words, iPad owners do surf more than anyone else, and they skew the results.

However, for smartphones, the usage is equal.

The initial "report" that started the whole "iOS users are on the web more" meme was this one from Chitika. Inside it was this key paragraph, which was totally ignored by reporters:



2) Apps and widgets don't show up in such stats. In many places around the world, apps and games are used far more often than surfing.

1) It's highly unlikely that over 25% of users are changing their user agents. In all android browsers, the default is set to mobile android or mobile etc. The analytics can still parse the device being used. Sites are capable of identifying the platform used in their analytics.

2) With 80% marketshare vs roughly 13% marketshare, there is still no clear correlation between the 2 numbers.

3) Your final comment seems to simply be speculation without facts. Even then, Apple shows twice the revenue of android in the app store.

 

[Tinmania]

You might be right but it could also be a little naive since Samsung has been outed for this already.

And if you dig around you will see that is about phony user reviews. Think it through. What good would it do to pay "trolls" in an online forum? To spend time bickering with people for no reason? Not to mention few people crawl forums to make a purchase compared to those who rely on user reviews.

Personally, I think it is very naive to think that it is happening here--if not plain old wishful thinking. What, did Samsung set up sleeper cells half a decade ago only to activate them now? Or a decade? I have seen registered users here accused of that who joined that long ago.


Michael

 

[giantfan1224]

And if you dig around you will see that is about phony user reviews. Think it through. What good would it do to pay "trolls" in an online forum? To spend time bickering with people for no reason? Not to mention few people crawl forums to make a purchase compared to those who rely on user reviews.

Personally, I think it is very naive to think that it is happening here--if not plain old wishful thinking. What, did Samsung set up sleeper cells half a decade ago only to activate them now? Or a decade? I have seen registered users here accused of that who joined that long ago.



Michael

From yahoo story I linked:

Commission has slapped Samsung with a fine worth about USD $340,000 for paying a pack of trolls to bash HTC products and praise Samsung products in online forums

Samsung wouldn't have to pay very much, in my opinion, considering there are plenty willing to bash Apple and praise Samsung for free (and vice versa, of course). So the fact that Samsung has been caught redhanded leads me to believe it's at least a possibility on a forum like this that has thousands of users.

As far as what good does it do, it's all about branding and creating an image, much like the millions (or billions) of dollars being spent on Beer or Coca-Cola or McDonalds ads.

 

[Tinmania]

From yahoo story I linked:



Samsung wouldn't have to pay very much, in my opinion, considering there are plenty willing to bash Apple and praise Samsung for free (and vice versa, of course). So the fact that Samsung has been caught redhanded leads me to believe it's at least a possibility on a forum like this that has thousands of users.

As far as what good does it do, it's all about branding and creating an image, much like the millions (or billions) of dollars being spent on Beer or Coca-Cola or McDonalds ads.

"Online forums" is as far as you got. Come back to me when you dig deeper and find out there are no "examples" of that and that the original allegation was online reviews.

But if you want to believe MR is littered with "paid trolls" go right ahead. I find it humorous.


Michael

 

[kdarling]

The size and placement of the speakers in a phone is an assurance that sound quality will never be anything to write home about. It's there as a necessary utility, but to expect any kind of sound "experience" is quite ridiculous.

True, but some effort is better than none

My old 2005 Samsung i730 smartphone had stereo speakers with 3D surround sound enhancement from SRS Labs. Just getting halfway decent bass from a handheld was pretty neat.

-

HTC got one thing right, and it's something that has always bugged me:

They at least put the speakers facing forward.

It seems like most every other device has the speakers facing outward from its edges. This means you always have to cup your hands around the edges to try to direct the audio towards the viewer.

-

Ironically, when Apple sued Samsung over the original Tab tablet design in Europe, Samsung got around it by moving the speakers into thin forward facing strips on either side of the display, an arguably better design anyway.

 

[Renzatic]

Yeah, we probably do get the occasional paid shill around here. Though the vast majority of them, as in 99.99%, are usually one-off posters who registered the day before to post things like "OLOL APPEL SUX N U R STUPID HAEV U TRIED THIS NEW SAMSUNG FONE IT IS SO RAD LINK LINK LINK", and then get banned about 3 posts later.

But people who have been around for a bit, and post regularly in the other forums down below? I seriously doubt they're paid shills. Yeah, it's possible Samsung, Microsoft, Lenovo, LG, or whoever paid some random guy to quietly infiltrate a community and slowly turn opinions away from the focus of the group to their own products over the course of years. Anything's possible. But I kinda doubt the effort and money spent would be worth the results. In the end, all they'd do is convince someone who was already considering something else anyway, which random advertisements and tech reviews on other tech sites would do just as good of a job of. It'd be a lot of effort for very little gain.