Apple's A7 Processor Truly 'Desktop Class', iOS Apps Don't Take Full Advantage

[jettredmont]

OK, now I see what you mean.

Let me phrase my argument this way: "Apple fanboys tend to state Apple did the right and best thing when only delivering even the latest iDevices with 1GB of RAM".

Again, this argument is technically nonsense. Anyone knowing the battery life of competing 2GB RAM-equipped WP products (the Nokia 1020) know the additional power usage of having to constantly power on an additional 1GB of RAM is negligble and in no way result in vastly inferior battery life. (Again, I purposefully not list 2/3GB Android phones because their OS is far less battery-friendly than either iOS or WP and, consequently, don't represent a battery-wise ideal OS.)

And there isn't a space / volume constraint either - after all, for example the Nexus 7 2013, which has a considerably smaller volume than the Retina iPad Mini / Air, has managed to pack in 2GB of RAM. The volume argument ("2GB of RAM would take far more volume") is also very often cited by Apple fanboys - by the above-cited "gaussian blur" over at DPR as well.

All in all, this is why I consider the argument "1GB is the best compromise" fanboyish. Because it's simply not true, neither battery life- nor volume-wise.

Much better, thank you.

However, your assertion that "1GB is the best compromise" is false is equally as unsupportable as that it is true.

I am not an Apple engineer, much less an iPhone hardware engineer. I don't know what the compromises are there. It is possible that they simply said, "Haha! If we fool everyone into buying this phone with 1GB of RAM, they'll be screaming for a new one with 2GB of RAM next year!". But, I somehow doubt it.

Apple has good engineers. Strike that: Apple has damned good, world-leading engineers. They aren't chumps. I trust them to weigh the costs and benefits and come up with the optimal configuration far more than any random forum poster on the interwebs.

 

[pubwvj]

More evidence that MacOS and iOS will merge. Hopefully we won't lose the power of the desktop machines and laptops and hopefully Apple will get the legacy support right this time.

 

[haruhiko]

1) More RAM is always better;
2) Apple devices (with default configurations) have been provided with less than sufficient RAM since the old Mac days - but you could upgrade the RAM yourself easily for a Mac, not the case after retina MacBook Pros and MacBook Airs.

 

[Menneisyys2]

Much better, thank you.

However, your assertion that "1GB is the best compromise" is false is equally as unsupportable as that it is true.

I am not an Apple engineer, much less an iPhone hardware engineer. I don't know what the compromises are there. It is possible that they simply said, "Haha! If we fool everyone into buying this phone with 1GB of RAM, they'll be screaming for a new one with 2GB of RAM next year!". But, I somehow doubt it.

Apple has good engineers. Strike that: Apple has damned good, world-leading engineers. They aren't chumps. I trust them to weigh the costs and benefits and come up with the optimal configuration far more than any random forum poster on the interwebs.

While I agree iPhones would have witnessed some battery life decrease from including 2GB RAM, the case definitely wouldn't have been the same with the larger-battery iPad. And, as I've pointed out in my prev. post, the iPad (because of the much larger-res screen and the needed screen buffers) is in a much larger need of more RAM to avoid for example Safari tab reloads.

In addition, Apple has always been aware of the UIWebView problem. All over the Web, its huge memory usage has been well-documented since the early iPhone OS 2 days. With the release of the XGA-screened iPad 1 and the Retina-screen iPhone 4, the problems have become much more pronounced. Apple should have fixed this problem at least in the iPads, where additional RAM, as I've explained above, surely wouldn't have caused much battery life degradation.

Leaving RAM out of the iPad just can't be explained by the lack of technical feasibility or decreased battery life (again, with the latter, unlike with the iPhone). Then, only one explanation remains: margins... That is, money, filthy money. As some others have also agreed in this #, Apple did sacrifice user satisfaction, lack of tab reloads / app unloading / crashes to secure their additional $5...6 / device income. While their business decision can be understood, it in no way can be called consumer-friendly instead of overly greedy and consumer-milking.

 

[Judas1]

A lot of people are missing the point. While the article praises the A7 processor, it pretty much knocks down iOS. Almost no app uses the potential of the A7. And why do you you think that is? The OS it's operated on maybe?

 

[SvP]

A lot of people are missing the point. While the article praises the A7 processor, it pretty much knocks down iOS. Almost no app uses the potential of the A7. And why do you you think that is? The OS it's operated on maybe?

No, i think it's because of the weather.. You know, global warming and all.. ;-)

 

[rudigern]

The real benefit of the transition (as shown in the PC space) is being able to address more than 4 GB of of RAM, which the iPhone doesn't currently have.

Can we stop with this 'only point of 64 bit is more RAM'. That is 64-bit memory addressing, 64-bit processing is separate and has huge benefits in computation depending on what you are doing.

 

[AppleScruff1]

Apple has good engineers. Strike that: Apple has damned good, world-leading engineers. They aren't chumps. I trust them to weigh the costs and benefits and come up with the optimal configuration far more than any random forum poster on the interwebs.

But more than likely that optimal configuration favors Apple dollars and sense wise. $5 on 100 million devices is a nice chunk of change in the Apple bank account. The experience is "good enough" with 1GB of RAM. And they know their customers trust them. So who gets the shorter end of the deal?

----------

Can we stop with this 'only point of 64 bit is more RAM'. That is 64-bit memory addressing, 64-bit processing is separate and has huge benefits in computation depending on what you are doing.

And it appears that thus far nothing being done on the iPhone or iPad requires it, so maybe more RAM would have brought more benefits to the consumer at this point in time. Just a thought.

 

[SvP]

No he's right. It's brand loyalists that aren't going to be taken seriously (and hitting? Seriously? We're a on a forum dude! ). I could name a few off the top of my head on this site. No real tech person would take anyone "brand loyalist" seriously because they can justify anything a company does even when it's below what the competition offers. All they care about is that it looks cool and has a logo on it.

Example: iPad mini was introduced with a non-retina display. The Nexus 7 had a retina display and was cheaper. Last year the mini gets updated with a retina and has a price increase. Heck the iPad Air is STILL at 16GB at $500! That's pathetic! The competition offers much more for much less. In many ways, the iPad is behind. Heck the iPhone is rumored to get a bigger screen this year. So what? It's been done. I'm already enjoying a 4.7" screen for almost two years now!

While I'm not for name calling, I don't consider that particular term to be name calling.

You avoid it tho; "brand loyalist" is a different word than "fanboy". Also; don't confuse your opinion with a truth, they're separate things.

 

[appledes7]

Nevertheless, while iOS is indeed very well optimized (along with WP, naturally; and as opposed to Android), there are some sore points. For example, UWebView's (and, consequently, Safari's) huuuuuge memory usage. While 7.1 has indeed been brought us some improvements, Web pages still take up far too much memory under iOS.

How much RAM an app uses is irrelevant so long as it doesn't hurt performance. My fiance's iPhone 5 (1GB, dual core, ~1.3 GHz) is much smoother when it comes to web browsing than my Nexus 5 (twice the RAM, twice the core count, nearly twice the clock speed).

Who cares if iOS takes up more RAM to browse the web when it still runs faster than competitors? RAM is meant to be used. Really this only matters if it affects multitasking performance. And I have found that iOS can at times switch between apps more seamlessly than Android. For example, my Nexus 5 seems to lag heavily when I tap the recent apps (AKA multitasking) button. Compared to my fiance's iPhone 5, my Nexus 5 is much slower at this.

So keeping in mind that iOS web browsing takes up a little more RAM than some of its competitors, who cares if it performs better and still multitasks smoothly? Isn't that the whole purpose of having RAM in the first place?

Again, I am not trying to justify Apple's hesitance to upgrade the RAM in their devices. I am simply stating the obvious: RAM is meant to be used. A statement such as "UWebView's (and, consequently, Safari's) huuuuuge memory usage" is pointless by itself. Put that statement in context and explain why it is bad, otherwise it is irrelevant.

Ultimately, our devices can always be smoother, faster, and more efficient. But with 1GB RAM, the iPhone is performing better than my top of the line Android devices with twice the RAM and twice the core count. How much RAM a certain app uses in iOS is irrelevant. What matters is performance, not numbers.

 

[subsonix]

And it appears that thus far nothing being done on the iPhone or iPad requires it, so maybe more RAM would have brought more benefits to the consumer at this point in time. Just a thought.

You could end up with a situation where some applications only run on the latest device because it requires more than 1GB of ram. On the other hand, on a small device with 1 screen that can only run 1 application at a time what exactly requires more that 1GB of resources in ram at one time. Games with large worlds load bite size chunks as needed from disk for example.

 

[SvP]

I'm not condescending. As I stated above, I just can't take anyone seriously stating 1GB is the sweet spot and that Apple couldn't have done better. Tons of competiting models on the market prove I'm right. Do these 2/3GB-equipped models have much worse battery life (apart from Android's toll on the battery - this is why I've contantly been referring to the WP8 Nokia 1020)? Do they have much more volume? Nope.

Apple made a, for us end users, very bad decision because of plain greed. To maximize their (Apple's) revenues (adding a 2GB chip would have increased the manufacture cost of each iDevice by $5-$6, which, as the end user prices wouldn't have risen, would be a $5-6 revenue decrease for Apple) and to make people more inclined to upgrade to the next iteration, which will surely include 2GB of RAM.

Yeah yeah, opinion, opinion. And all that opinion goes right down the drain, because you are arrogant ánd condescending in an ad nauseum discussion. You even manage to put it in one sentence. "I don't condescend, but I can't take them seriously". Not taking them seriously, calling them fanboys.. It's condescension.

I understand you're mad because you can't play your game, and I really get the frustration. I sincerely hope it runs better on your current phone and I also hope the specs on all future devices get raised. I really mean that.

The sad bottom line is; you're chasing windmills. The only ppl that listen are on a rumours forum. I don't even understand your goal: did you take it upon yourselves to convince the "sheep" to wake up? Is it that you want to educate them?

 

[AppleScruff1]

You could end up with a situation where some applications only run on the latest device because it requires more than 1GB of ram. On the other hand, on a small device with 1 screen that can only run 1 application at a time what exactly requires more that 1GB of resources in ram at one time. Games with large worlds load bite size chunks as needed from disk for example.

I understand what you're saying. I was thinking more about web browsing where Safari or Chrome reloads pages even if you only have 3 tabs open. What is the reason for that? Memory management or something other than not enough memory? I'd like to know. It is a bit annoying.

 

[techwhiz]

ARM started on the desktop - it was ludicrously fast compared with the x86 processors at the time - 1987, when it appeared in the Acorn Archimedes. This machine was capable of running a full WIMP GUI in colour and did great 3D graphics too at a time when PCs were still stuck with DOS. There's nothing about the ARM processor that precludes it from being in a desktop other than the design decisions to make it low power and efficient which is where it found a niche. A full desktop class design is totally possible and it would certainly give other RISC processors a run for their money.

The problem is that while everyone else has been focused on performance ARM has been focused on low power with a compromise in the area of things like branch prediction, speculative execution, etc. Those things take lots of transistors and it was not the focus of ARM.

Cache performance and multiprocessor cache coherency are not things that ARM has been focusing on. ARM just got to 64 Bit. Multiple threads, shadow registers for fast context switching, etc....

Intel and even AMD are way ahead of ARM in the performance area.
Also trying to compete with Intel in processor development has put many a company out of business.

While it makes sense for Apple to use ARM in a mobile device it makes almost no sense for them to do it in a desktop processor.

 

[iRun26.2]

Start your therapy now. It's coming Spring 2016, and will be Fabed in Malta, NY.

2016 is fine with me.

I just want one more MBA (the new rumored 12" retina MBA) to be still Intel based. I can put off worrying about my x86 software no longer working on an MBA at least for another generation.

Maybe the 2016 ARM processor will be able to emulate x86 code as fast as my current MBA can run it natively. I'd then be OK with switching to ARM unless Intel has something even better that ARM can't touch. I don't care about price.

 

[jettredmont]

I'd like to see some figures. Assuming we have an 1500 mAh (that is, comparatively low-capacity) battery, how much effect the increased RAM would have on the battery life with current(!) RAM tech? 3%? 5%? 10%?

And the above only applies to the iPhone because of its comparatively small battery. On the iPad and its much-much larger-capacity (8000 mAh?) battery, the difference would be barely visible. With iPads, one really can't seriously state "they don't have 2GB of RAM because it'd have seriously decreased the battery life". Except for brand loyalists, that is - but no one should take them seriously in a question where they need to justify Apple's greed.

Well, there aren't a whole lot of studies on this, but here's one:

http://web.cs.wpi.edu/~emmanuel/courses/cs525m/S13/slides/smartphone_power_analysis_wei_wang_wk9.pdf

... which seems to be a summary (although including some additional details in a few areas, oddly) of:

https://www.usenix.org/legacy/event/usenix10/tech/full_papers/Carroll.pdf

And there's:

http://www.cs.arizona.edu/~gniady/papers/igcc11-phone.pdf

In the first, he looks at a 128MiB RAM device and sees ~40mW power consumption from the RAM during the heaviest-drain activity (viewing a video), much less at times of lower use. He looked at higher-RAM devices, but doesn't give the same details there.

In the second we get tidbits like:

According to [12], the memory subsystem in a smartphone can consume 34.5% of the overall phones energy when executing applications, resulting in a much higher energy consumption than the CPU which contributes 15.5% to the overall energy consumption. Therefore, it is critical to improve energy efficiency of main memory in smartphones.

Their "standard" DRAM configurations show up to 170 Joules to run 192MiB RAM while playing music.

http://www.google.com/url?sa=t&rct=...496LBMgjUvAVDjR9oqcLwOg&bvm=bv.63934634,d.cWc

(sorry for the link length; it's a direct download of a PDF)

Lists a rate of 560mW to save 1MB of data into RAM on a phone (a Nokia N95), but an energy cost of just 2.2 joules. So, indeed, RAM can take up a lot of energy, if run full-bore.

I think the end story is that RAM can matter, a lot, in overall power consumption. How that compares to the battery size would require knowledge of the average voltages across the RAM while it is consuming this power, and even then I'm just basing this off guesswork. The main question is, how much more power does 2GB of RAM consume than 1GB? I can't find a good answer on that. It's probably a safe-ish bet to say it is a linear relationship, as I would imagine both sub-linear and super-linear tendencies. Which would bean that you'd multiply all the above by 8 to get to the ~1GB range, and by 16 to get to the ~2GB range.

 

[iRun26.2]

I would have doubted this even a year ago, but for MacBook Air and Mac mini, I think it could happen with A8. Wonder if they could manage a 20+ hour battery life MacBook Air Retina w/ A8, 12" screen, reeeaaally thin, maybe 8GB RAM, 256 and 512GB solid state storage options. That would be an exceptionally tempting machine!

That can't run any of my current x86 software???

No thanks!

 

[Menneisyys2]

So keeping in mind that iOS web browsing takes up a little more RAM than some of its competitors, who cares if it performs better and still multitasks smoothly? Isn't that the whole purpose of having RAM in the first place?

Again, I am not trying to justify Apple's hesitance to upgrade the RAM in their devices. I am simply stating the obvious: RAM is meant to be used. A statement such as "UWebView's (and, consequently, Safari's) huuuuuge memory usage" is pointless by itself. Put that statement in context and explain why it is bad, otherwise it is irrelevant.

Tab reloads.

 

[jeremysteele]

That can't run any of my current x86 software???

No thanks!

Love how people keep saying this... they also said it during the PPC->X86 transition... yet 95+% of the software ran fine.

People also thought this about intel-based android devices not being able to run software compiled for android ARM devices (real compilation, not Java software which worked fine without anything extra)

Yet in both instances it works fine thanks to JIT binary translation. Amazing tech that has been around since the 80s.

So long as the underlying OS is similar and the hardware is relatively similar (you aint gonna make a TI-85 run x64 software!), binary translation is pretty straightforward.

That said, ARM is still slow as molasses (although very energy efficient). Even if they were to make an ARM air in the next 1-2 years - it would be slow. Granted, most people probably wouldn't even notice.

 

[Michael CM1]

I can definitely tell the power boost from my recent upgrades from an iPhone 4S to 5S and iPad 3 to iPad Air. The 4S kinda slowed down with iOS 7, but the 5S just zips along. The iPad 3 was insanely sluggish, at times waiting to act when I would tap something on the screen. It was kind of pathetic at times when I tried to watch a 1080p video, such as "Gravity."

The real test is how these devices are performing in a couple of years. I'm on an iMac that I think is 3 years old. It's in no way sluggish on everyday tasks unless I go nuts on the amount of tabs I have open. Even then, I have 12GB of RAM to mitigate that. I'm sure a brand new iMac at the same price point would maybe encode video faster just as this did compared to my previous MBP, but I doubt I would notice much difference otherwise.

That's what I think when I hear "desktop class." Because they don't have to worry about power and space issues as much, they can be more powerful than mobile processors. Thus they can survive newer software over a longer period of time without showing age. If in three years these two iDevices I have are running about the same aside from when I play some brand new game that takes advantage of the A10 or whatever, then that will prove the claim of desktop class.

 

[kiantech]

Setting aside your strawman argument, a RAM bottleneck refers to the speed of the RAM, not the amount of memory.

You're wrong and clearly didn't read the article. Straight from Anandtech's article.

The other problem I see is that although Cyclone is incredibly forward looking, it launched in devices with only 1GB of RAM. It's very likely that you'll run into memory limits before you hit CPU performance limits if you plan on keeping your device for a long time.

This clearly implies size limitation not bandwidth. If it was bandwidth he wouldn't have referenced size but speed.

 

[appledes7]

Tab reloads.

Still smoother on my fiance's iPhone 5 than my Nexus 5 and Nexus 7.

 

[Tknull]

Let me phrase my argument this way: "Apple fanboys tend to state Apple did the right and best thing when only delivering even the latest iDevices with 1GB of RAM".

You seem to get a hard on from use of the word Fanboy. Just curious, are you capable of making a reasoned argument without feeling the need to put down those whom you see as your opponents? You might find a much more receptive audience.

 

[MattInOz]

More evidence that MacOS and iOS will merge. Hopefully we won't lose the power of the desktop machines and laptops and hopefully Apple will get the legacy support right this time.

but also evidence they'll diverge in other ways at the same time.

 

[bdkennedy1]

It's easy to see the road map. Apple likes to make their own hardware. Making their own processors allows quick turnaround by not having a middleman (Intel). I think they will roll out Macbook Air's with the A8 or A9, iMac's and Mac Mini with A10 and then MacBook Pros with the A11.

Apple always hides stuff in their hardware. This is the perfect example.

Is there going to be a desktop computer with a Qualcomm chip in it? Not for a long time or never. This is why Qualcomm almost **** their pants.