Am I the only one that wants a single core A5?

[JulianL]

Lots of people are talking about a quad core iPhone but for me my iPhone 4 is already pretty snappy and right now I care a lot more about battery life than I do about speed. A dual core A5 is already in the iPad 2 and Steve Jobs said in the iPad 2 launch event that the A5 drew the same power as the old A4; the battery life specs for the iPad 2 seem to confirm this so a single core A5 (with a similar halving of the number of cores/shaders in the GPU and the shared cache sizes) should draw about half the power of the existing A4. Admittedly that won't double battery life because the screen is a huge factor but I would have thought that it would be a worthwhile improvement.

Additionally, an A5 core appears to have about 25% more performance than an A4 core at the same clock speed, and the GPU performance took such a huge leap with the A5 that even a half-sized A5 GPU will still probably be about double the performance of the current iPhone A4 GPU, so even a single core A5 will be a noticable performance boost from the current iPhone 4.

OK, all the above would look bad in the "my spec is better than yours" contest but for my personal feature set it would be a far better tradeoff of power versus battery life. Does anyone else think like me or am I alone on this one?

- Julian

 

[macingman]

Wirelessly posted (Mozilla/5.0 (iPhone; U; CPU iPhone OS 4_3_2 like Mac OS X; en-us) AppleWebKit/533.17.9 (KHTML, like Gecko) Version/5.0.2 Mobile/8H7 Safari/6533.18.5)

Your just about the only one that thinks this. There will be very little battery improvements. There is many more factors to battery life then just the processor+screen.

 

[MaxBurn]

I am thinking that second core is turned off most of the time in the ipad2 anyway, only brought on when needed.

I am sure it would be a fasinating read to see all the power management stuff that goes on for ipad2.

 

[CylonGlitch]

the battery life specs for the iPad 2 seem to confirm this so a single core A5 (with a similar halving of the number of cores/shaders in the GPU and the shared cache sizes) should draw about half the power of the existing A4.

This is a where you're logic isn't correct. There are two cores, but that does not mean that turning one off will save you 50% of the power; that's not how semiconductors (typically) operate. There are resources on the chip that will be used by one, both, or multiplexed between the two; only the components that are used by one will be able to be shut down. Yes, this will save some power, but it won't be as significant as you would hope because there is a lot of the "both" logic that has to be powered up and running. If you save 25% I'd be very surprised.

On top of all this, the software and the hardware (the A5 itself) are designed so that anything that isn't used is shut down. Thus any part of the processor that is not active will go to sleep to reduce power. This will reduce that 25% even further.

You might be able to see a 10% reduction in processor power requirement by going to a single core A5; but is that really going to make much difference? The power requirement of the rest of the phone is going to make up a much larger percentage and thus you'd not likely notice that savings or if you did, it wouldn't be significant.

 

[victorvu]

Lots of people are talking about a quad core iPhone but for me my iPhone 4 is already pretty snappy and right now I care a lot more about battery life than I do about speed. A dual core A5 is already in the iPad 2 and Steve Jobs said in the iPad 2 launch event that the A5 drew the same power as the old A4; the battery life specs for the iPad 2 seem to confirm this so a single core A5 (with a similar halving of the number of cores/shaders in the GPU and the shared cache sizes) should draw about half the power of the existing A4. Admittedly that won't double battery life because the screen is a huge factor but I would have thought that it would be a worthwhile improvement.

Additionally, an A5 core appears to have about 25% more performance than an A4 core at the same clock speed, and the GPU performance took such a huge leap with the A5 that even a half-sized A5 GPU will still probably be about double the performance of the current iPhone A4 GPU, so even a single core A5 will be a noticable performance boost from the current iPhone 4.

OK, all the above would look bad in the "my spec is better than yours" contest but for my personal feature set it would be a far better tradeoff of power versus battery life. Does anyone else think like me or am I alone on this one?

- Julian

I'm with the OP on the idea that I favor battery life more than any fancy specs. But I think using dual core / quad core CPU with appropriate power management will bring you a better battery life. Besides, doing a specific work twice faster is actually gonna help you preserve more battery to do something else.

Look at i3, i5 CPU in laptop nowadays, they have like 6 or 8 hours battery life, compare to 3 or 4 hours that dual core laptops have.

 

[Pink∆Floyd]

Yes, you are most certainly the only one who wants this.

 

[wordoflife]

Um, no? Don't dual cores use less power? As in, a 600mhz dual core will use less power than a single core 1ghz but still be faster?

I'm pretty sure the iPhone 5 will have the same processor as the iPad 2. Hopefully iOS 5 takes advantage of a dual core.

 

[AndroidfoLife]

The moto atrix uses a dual core and has better battery life then the iPhone. The iPhone battery life will not take that large of a hit with a dualcore. But it doesn't need it. The iPhone does not do real multitasking so its not really needed.

 

[JulianL]

To clarify, I wasn't talking about disabling the second core, I was talking about Apple producing an A5sc part for the next iPhone that physically didn't have a second core. Half as many transistors means a significant reduction in power drawn by the chip. The fact that the second core in a dual core CPU is often unloaded doesn't mean it draws no current, leakage current is a big issue in today's semiconductors.

It sounds as if I am in the minority here although not completely alone.

- Julian

 

[err404]

It sounds as if I am in the minority here although not completely alone.

Your desire for better battery life is valid, but just removing a core doesn't get you much loser to that goal. ARM CPU's are already very good at dynamically shutting down resources not currently being used. Simply removing a core would drastically reduce performance while doing little to reduce the power consumption.

On the other hand Apple can lower the power requirement by simply under clocking an Dual Core A5. Apple already does this with the A4. The iPhone 4 runs at 800mhz while the iPad(1) runs at 1ghz. Apple may be able to dynamically alter the clock (they may do this already). With the extra speed provided from the A5, they could significantly down volt the CPU when not running intensive apps and still provide good performance.

 

[Apple 26.2]

Your desire for better battery life is valid, but just removing a core doesn't get you much loser to that goal. ARM CPU's are already very good at dynamically shutting down resources not currently being used. Simply removing a core would drastically reduce performance while doing little to reduce the power consumption.

On the other hand Apple may be able to lower the power requirement by simply under clocking an Dual Core A5.

Exactly. Plus the masses would scream bloody murder. All to squeeze out a few extra minutes? Not worth it.

Well said, bro.

 

[Alaerian]

I really don't care. I'm not a spec-whore.

 

[amarcus]

In most circumstances two cores are more efficient than one. I'd expect to see both performance and battery life take a hit in this lobotomised A5.

 

[gpat]

Exactly, the chip is going to suck more power while being slower, just like the MacBooks when you tried to disable one core.

 

[ani23]

Yes

 

[vincenz]

The battery life on the iPhone is great right now, but I'd love for them to extend it even a few more hours.

 

[Krevnik]

The moto atrix uses a dual core and has better battery life then the iPhone. The iPhone battery life will not take that large of a hit with a dualcore. But it doesn't need it. The iPhone does not do real multitasking so its not really needed.

Wut? Seriously... Wut?

There are very good benefits to having dual-core, despite the iPhone's more limited multitasking model. Dual-core doesn't relate to multitasking per se, but rather being able to run multiple threads simultaneously rather than forcing everything through a time-share on the same core.

iOS/OS X apps are usually written in a way to keep heavy work off the main (UI) thread. This means you can split up the threads across different cores. Any multithreaded apps that push the limits of the CPU will get a noticeable speed boost from the second core.

And there are always processes running in the background on an iPhone. The kernel has work to do, there are a handful of daemons that do quite a bit of work in the background on behalf of other apps. Safari and Mail also run in the background (why Safari needs it though, I don't really know).

The key thing is that everything these days runs on at least 2 threads, even if the developer doesn't explicitly design it that way (thanks to the design of things like AppKit/UIKit and the Cocoa/Foundation frameworks). If I have work to do that spans multiple threads, I can get it done faster with a second core, and go into a lower power state faster. Web browsers are a good candidate for this situation, because they do a lot of rendering and fetching that can be split off into threads that will be balanced on the two cores.

And splitting a lower load across two cores helps since you can lower the clock speed on the CPU to something lower than the maximum and still get the work done in an acceptable amount of time. This saves battery life over a single core solution where the single core would have to be clocked at the maximum clock speed in order to handle the load.

 

[JulianL]

Simply removing a core would drastically reduce performance...

But that was half of my argument, removing an A5 core would not drastically reduce performance, it would almost certainly increase it at the same clock speed. Clearly you now think I'm mad but don't forget that my starting point for measuring comparative performance was the iPhone 4 with an A4 which is a single core chip so a single core A5 isn't removing a core at all when considered relative to the current A4, it's a case of maintaining the same core count but the new core is more capable at the same clock speed and draws about half the power.

I don't for one second believe that I'm going to get what I'd like, my guess is that the existing iPad A5 will be dropped into the next iPhone at the same clock speed as the current A4, which is widely thought to be 800MHz, and according to Steve Jobs that will draw the same power as the old A4 (assuming the frequency vs power curve is the same on the A5 as the A4) so any battery life improvements will need to come from boosting the battery capacity, optimising power draw of other components (and the screen is the big one there), and software optimisations. Still, Apple has progressively improved battery life on the last couple of iterations of the iPhone so it'll be interesting to see if they can keep that going this next time.

- Julian

 

[Krevnik]

But that was half of my argument, removing an A5 core would not drastically reduce performance, it would almost certainly increase it at the same clock speed.

Uhm... no?

That isn't really how it works. Since the core wouldn't change, that core would perform the same as it does now. You just have half the peak performance available. That would mean the core needs to operate closer to the max frequency a larger percentage of the time, and would wind up drawing more power than if it could run at the lower frequency more often.

Now if we changed that one core to have better performance by using a larger portion of the die, we wouldn't be talking about the A5 anymore, it would be an entirely different beast with a different TDP due to the larger footprint of the core.

the new core is more capable at the same clock speed and draws about half the power.

No again. As others have already pointed out, dual core chips share a lot of components which all need power. The frequency the chip is running at, and the scale of the die (45nm in this case) tends to drive power consumption more than the core count. If you take a look at the A5 floorplan in this link, you will notice that a single core takes up no more than about 1/3rd of the actual space on the die. This means the power savings are much more modest than your claim if we were to remove one.

http://www.chipworks.com/en/technic.../2011/03/apple-a5-vs-a4-floorplan-comparison/

EDIT: Also keep in mind this footprint is only for a single layer of the SoC package. It ignores the layers of shared cache (which needs the same amount of power with or without a second core), and the DRAM (which again needs power and can't even be fully powered down during sleep).

... according to Steve Jobs that will draw the same power as the old A4 (assuming the frequency vs power curve is the same on the A5 as the A4) so any battery life improvements will need to come from boosting the battery capacity, optimising power draw of other components (and the screen is the big one there), and software optimisations.

The simple fact that the iPad 2 actually tests better in battery life trials from places like Engadget tells me that the second core is actually showing a power benefit due to the reasons I stated above in my previous post. It isn't all about the TDP and assuming that the chip runs exactly the same way in both situations. The second core is allowing the iPad 2 to run at a lower clock speed more often than the original iPad. Assuming the same freq vs power curve, that means better battery life.

 

[sammich]

Uhm... no?

That isn't really how it works. Since the core wouldn't change, that core would perform the same as it does now. You just have half the peak performance available. That would mean the core needs to operate closer to the max frequency a larger percentage of the time, and would wind up drawing more power than if it could run at the lower frequency more often.

Exactly. If I might add a few real world examples to the mix:

An episode of Top Gear had a Prius driven around the track at best time speed, while a M3 was following behind it. At the end, when fuel consumption was tallied up the Prius used up many times more fuel since it was being used at it's designed limits.

Another example comes by way of a future bluetooth spec comment (can't remember if it was about 2.# or 3.0). It was said that while the newer version used slightly more power when operating, it was several times faster which would mean that it was operating for shorter periods than the old/current spec.

 

[SAIRUS]

I think Apple should take a hit and up the screen size to make the phone bigger and allow a bigger battery. It would of course have to compensate for the bigger screen though, so maybe they can adjust the back to something thinner and more durable than the glass on the iPhone 4. Just a little more space to get that battery going.

 

[Reach9]

I believe a Dual Core A5 chip is perfectly fine, and hopefully we'll see iOS 5 use some power and deliver a great OS.
The Dual Core is more productive, and it extends the battery life. If the iPad 2 didn't have all the need for power, then the battery life would've been longer.

 

[NT1440]

But that was half of my argument, removing an A5 core would not drastically reduce performance, it would almost certainly increase it at the same clock speed. Clearly you now think I'm mad but don't forget that my starting point for measuring comparative performance was the iPhone 4 with an A4 which is a single core chip so a single core A5 isn't removing a core at all when considered relative to the current A4, it's a case of maintaining the same core count but the new core is more capable at the same clock speed and draws about half the power.

I don't for one second believe that I'm going to get what I'd like, my guess is that the existing iPad A5 will be dropped into the next iPhone at the same clock speed as the current A4, which is widely thought to be 800MHz, and according to Steve Jobs that will draw the same power as the old A4 (assuming the frequency vs power curve is the same on the A5 as the A4) so any battery life improvements will need to come from boosting the battery capacity, optimising power draw of other components (and the screen is the big one there), and software optimisations. Still, Apple has progressively improved battery life on the last couple of iterations of the iPhone so it'll be interesting to see if they can keep that going this next time.

- Julian

Might want to look into how processors work before you post any more, its slightly embarrassing....

 

[chris2k5]

iOS is so snappy with the A4. I can see them putting in the dual core A5 for sure though. I don't think battery will take a hit.

If it does during tests, I am sure they will optimize it to use the second core only when needed.

 

[JulianL]

Might want to look into how processors work before you post any more, its slightly embarrassing....

For the record, I know very well how processors work. What a few people here seem to fail to understand is that all cores are not equal. The core in an A4 is almost certainly(*) based on the ARM Cortex-A8 and the cores in the A5 on the Cortex-A9.

There were substantial architectural improvements with the Cortex-A9 that boost IPC by something like 25% compared to a Cortex-A8 core, in real life performance and not just on synthetic benchmarks. The most significant architectural change is that the Cortex-A9 supports out of order execution which makes a significant performance difference but there are other changes too.

It's a shame a few participants sunk to personal attacks but I got the answer to my question so I'll leave it at that.

- Julian

(*) I say "almost certainly" because I don't believe that Apple ever definitively confirm this level of spec detail although the evidence uncovered by third party analysis seems to strongly support the hypothesis.