Early Benchmarks Put Retina MacBook CPU Performance in Range of 2011 MacBook Air

[subsonix]

The Rosetta framework was included there during transition but when you are taking one machine code and running it on a virtual x86 machine you will probably at least halve the performance (maybe more). With the Power architectured applications they ran considerably slower on newer hardware (for almost a year). Now take that same concept and apply it to a low power CPU and you will have a machine that would likely have less performance than half the power of the latest iPads. It would take a year for some applications to sort out with the "fat binary" as you said.

Yeah, I but wasn't talking about emulation, just the fact that Win RT was different in other ways than the CPU it ran on. OS X on two architectures would be the same OS, with the same frameworks and so on.

Then you have the "fat binary" which is basically two copies of the same application, one for each platform - which makes applications.... fat... It was only meant for transition.

A fat binary is a feature of the Mach-O file format, http://en.wikipedia.org/wiki/Fat_binary

If they were to do it again, they would probably move to just compiling down to llvm byte code.... The effect is that it would effectively increase the testing budget significantly (in one case I was personally involved in at a large shop - it was 2 millions of dollars in additional costs to certify each - and that was just a difference in operating system versions not even architectures).

Then they also need a JIT compiler for that llvm code, I mean why not then compile it to multiple architectures and solve the same problem.

 

[Renzatic]

Macs aren't really taking the enterprise by storm or anything, Apple is a consumer company. Also, market share means nothing. Plus, I just meant this one computer anyway. If you think x86 is the best thing, buy an Air, Pro, iMac, or Mac Pro.

As has been stated before, having an ARM computer bearing the Macbook name will only serve to confuse consumers, and dilute the brand. You'd have a few people running out to spend $1300 on a brand new computer, only to find out that it's nothing more than a glorified iPad without the apps.

In other words, it'd be the Surface RT with even less appeal.

 

[Michael Goff]

As has been stated before, having an ARM computer bearing the Macbook name will only serve to confuse consumers, and dilute the brand. You'd have a few people running out to spend $1300 on a brand new computer, only to find out that it's nothing more than a glorified iPad without the apps.

In other words, it'd be the Surface RT with even less appeal.

If Apple were to have come out on stage at, say, the WWDC with their new ARM MacBook? Those problems would likely be gone. They'd have Adobe on stage, maybe somebody from Microsoft, spouting how great this thing is. They'd have ads telling consumers about how light and portable it is, and where this fits. They'd have the major things ported over within a year, maybe two. I don't think consumers would be confused, Apple is king when it comes to marketing their stuff.

Edit: As for being an RT device with less appeal, I disagree. For me, the biggest problem with my Surface RT is the processor. If it had something as powerful as the A8X, it'd have been awesome. Maybe not 1300$ awesome, but I doubt they'd charge that much for this (as they don't have to pay for the Core M, which isn't exactly cheap)

 

[Renzatic]

...Apple is king when it comes to marketing their stuff.

This is true, but ask yourself this: what advantages would an ARM Macbook hold over its current Intel based iteration? Would it be faster? Probably not. The A8 is quicker than the Atom chips, but only benches at roughly half the performance of the Core M. Battery life? Yeah, you'd probably get a couple more hours out of an ARM chip, but, as stated above, it'd be at the cost of performance. Price? It'd be cheaper for Apple, that's for sure. But Apple being Apple, it'd probably still cost us around $999.

As for higher level applications, yeah, it'd get them eventually, though it'd take a couple of years for all these companies to port them all. Then you'd have to take performance into consideration. The Macbook as it is, while far from offering cutting edge, screaming performance, is still a pretty capable machine.

On top of that, you'd have a Macbook that isn't to boot or virtualize Windows. Don't discount the need or appeal of this.

What you'd have with an ARM Macbook is a machine that sits between the iPad and the regular Intel Macs, but without any of the appeal of either.

 

[Michael Goff]

This is true, but ask yourself this: what advantages would an ARM Macbook hold over its current Intel based iteration? Would it be faster? Probably not. The A8 is quicker than the Atom chips, but only benches at roughly half the performance of the Core M. Battery life? Yeah, you'd probably get a couple more hours out of an ARM chip, but, as stated above, it'd be at the cost of performance. Price? It'd be cheaper for Apple, that's for sure. But Apple being Apple, it'd probably still cost us around $999.

As for higher level applications, yeah, it'd get them eventually, though it'd take a couple of years for all these companies to port them all. Then you'd have to take performance into consideration. The Macbook as it is, while far from offering cutting edge, screaming performance, is still a pretty capable machine.

On top of that, you'd have a Macbook that aren't isn't to boot or virtualize Windows. Don't discount the need or appeal of this.

What you'd have with an ARM Macbook is a machine that sits between the iPad and the regular Intel Macs, but without any of the appeal of either.

The A8X, which I suggested would be used, has an 1808/4529. The thing they're bragging about is beating what's in the iPhone. That's not really something to brag about, really. So the AX has a slightly lower single core and a slightly higher multi core.

http://browser.primatelabs.com/ios-benchmarks

 

[Renzatic]

The A8X, which I suggested would be used, has an 1808/4529. The thing they're bragging about is beating what's in the iPhone. That's not really something to brag about, really. So the AX has a slightly lower single core and a slightly higher multi core.

http://browser.primatelabs.com/ios-benchmarks

Huh. That is a lot more impressive. It just about matches the Core M. Still, my other points stand. It'd be a lot work both for Apple and 3rd party developers to support another platform who's ultimate goal is to reach parity with what's already out now.

 

[Michael Goff]

Huh. That is a lot more impressive. It just about matches the Core M. Still, my other points stand. It'd be a lot work both for Apple and 3rd party developers to support a 3rd platform who's ultimate goal is to reach parity with what's already out now.

Yeah, but it has a lot of potential. Imagine how much they could optimize to the A processor. I'm not saying it'd be a perfect main computer, but as a secondary (maybe the cost of the lower end MacBook Air?) it'd be great. This could perform better than the slightly stronger Core M, theoretically.

 

[Renzatic]

Yeah, but it has a lot of potential. Imagine how much they could optimize to the A processor. I'm not saying it'd be a perfect main computer, but as a secondary (maybe the cost of the lower end MacBook Air?) it'd be great. This could perform better than the slightly stronger Core M, theoretically.

There's tons of potential. I just think Apple should spend its time realizing it with the iPad, rather than temporarily gutting the Macbooks to get there. iOS already has the developer support, a catalog of decent apps, and a huge user base. It'd be better to build on top of that, rather than start all over and hope for the best.

 

[Michael Goff]

There's tons of potential. I just think Apple should spend its time realizing it with the iPad, rather than temporarily gutting the Macbooks to get there. iOS already has the developer support, a catalog of decent apps, and a huge user base. It'd be better to build on top of that, rather than start all over and hope for the best.

I just doubt Apple is going to really push the iPad to where the A8X can take it. We're at this point where the limits of the OS are holding back their own chips. Or maybe it's their obsession with making the iPad thinner than it should be. Either way, the OS is not taking as good of advantage of the hardware as I think OS X would.

Then this year we're going to get the A9X, and it's going to be even more powerful with just as little to do with it.

 

[Renzatic]

I just doubt Apple is going to really push the iPad to where the A8X can take it. We're at this point where the limits of the OS are holding back their own chips. Or maybe it's their obsession with making the iPad thinner than it should be. Either way, the OS is not taking as good of advantage of the hardware as I think OS X would.

This is why I'm waiting for the iPad Pro. The only limits on iOS are the limits Apple themselves put into it. There's no reason why it or an iteration of it can't eventually become the touch equivalent to OSX.

Though the way things currently are...

Then this year we're going to get the A9X, and it's going to be even more powerful with just as little to do with it.

So, so true. Apple's got the most powerful mobile CPU on the market, that's being used for Facebook, Candy Crush, or a bunch of productivity apps that run almost as well on my 100x slower iPad 3. What's the point of having and advertising all that power if you're not gonna use it?

 

[diamond.g]

The A8X, which I suggested would be used, has an 1808/4529. The thing they're bragging about is beating what's in the iPhone. That's not really something to brag about, really. So the AX has a slightly lower single core and a slightly higher multi core.

http://browser.primatelabs.com/ios-benchmarks

Interestingly the fool link is only comparing the geekbench score of the 5Y10a which Apple isn't using... Clearly the chips Apple is using would be faster.

 

[bkkcanuck]

The single port (which is not a problem for me) and other changes are up to each to decide if they are a problem for them..... but it is interesting to note some of the first impressions on performance (which I did not believe would be a problem for the majority of users).

Darrell Etherington, TechCrunch:

I came to the MacBook with certain expectations; specifically, that it would not be able to meet my more "pro" level needs, in terms of Photoshop, Final Cut Pro and Logic Pro. Luckily, the MacBook defied those expectations and performed well with each of the above applications.

Which isn't to say performance is on par with, say, the brand new 13-inch MacBook Pro -- it isn't. But pre-launch concerns of this machine being seriously hampered by its low-power Intel M processor were, in my experience, very premature. The new MacBook handled the tasks I threw at it so well that I am no seriously considering whether or not I can adopt one full-time, as a replacement to my original 2012 15-inch Retina MacBook Pro road warrior.

 

[KPOM]

Yeah, but it has a lot of potential. Imagine how much they could optimize to the A processor. I'm not saying it'd be a perfect main computer, but as a secondary (maybe the cost of the lower end MacBook Air?) it'd be great. This could perform better than the slightly stronger Core M, theoretically.

There is more than just GeekBench, however. The Core M performs reasonably well considering it puts out 1/3 the power of even the processor in the MacBook Air. It just throttles more, which hampers speed if the OEM hasn't designed the thermal management system properly.

 

[bkkcanuck]

There is more than just GeekBench, however. The Core M performs reasonably well considering it puts out 1/3 the power of even the processor in the MacBook Air. It just throttles more, which hampers speed if the OEM hasn't designed the thermal management system properly.

The stock Core M is NOT 1/3 the power of the stock Macbook Air processor... it is closer in power to the stock Macbook Air processor than you give it credit for. Even the headlines saying it was the same power as the 2011 Macbook Air actually used the top of the line i7 as a comparison. The average increase in "stock" processors performance inclusive of the 2011 upgrade is 30% (25% in 2011, 2% - 7% for every year after that), and they did not even bother to use the stock processor from 2011. The Atom processor (the one well below this one is maybe 1/3 the performance for single core - that is the Surface 3 processor).

Even early reviewers of the Macbook are not saying that it is "throttling".... they mention it gets "warm" (but not hot) under load - no need to throttle at that temperature (yes, I can remember the older laptops frying your unmentionables).

 

[KPOM]

The stock Core M is NOT 1/3 the power of the stock Macbook Air processor... it is closer in power to the stock Macbook Air processor than you give it credit for. Even the headlines saying it was the same power as the 2011 Macbook Air actually used the top of the line i7 as a comparison.

What I meant is that it is a 5W processor while the MacBook Air uses a 15W processor. That difference is nearly enough electricity to power the average CFL light bulb (which is typically 13W). I think the Sandy Bridge processor from 2011 might actually be a 17W chip.

 

[SlCKB0Y]

Weak sales of the Surface ended with the SP3, sales have really picked up and remained hot. The upcoming Surface 3 for $499 will be an even bigger success.

I was not referring to Surface Pro, only the two Surface tablets running Windows RT.

----------

The A8X, which I suggested would be used, has an 1808/4529.

Comparing Geekbench results for a device running iOS to one running OS X is meaningless.

 

[SlCKB0Y]

If Apple were to have come out on stage at, say, the WWDC with their new ARM MacBook?

What's the point? What the benefit of ARM over Core M for a Macbook?

Is the rMB too thick? Too noisy? bad battery?

The answer is no to all three questions, so then why do this with all the downsides?

1. No x86 virtualisation.
2. Not powerful enough for x86 emulation so no Rosetta equivalent
3. No boot camp.
4. No drivers for 3rd party devices
5. No backwards compatibility for any apps.
6. Massive overhead for Apple having to maintain ARM OS X for one device.
7. Consumer confusion over all the above
8. Developer frustration.

Throughout all of this discussion, you never address the downsides and you're yet to put forth a single compelling reason as to why they should do this.

----------

Huh. That is a lot more impressive. It just about matches the Core M.

Without the overhead of full multitasking and OS X...

 

[bkkcanuck]

Without the overhead of full multitasking and OS X...

You might want to reword that one. The operating system itself on the iOS devices ARE multitasking operating systems. Apple just restricts 3rd party access to it because of the way the device is used and the drain on the battery from not restricting it.

 

[SlCKB0Y]

You might want to reword that one.

* No user-installed Apps get unrestricted CPU access and multitasking, potentially lowering CPU overhead significantly.

Happy?

 

[bkkcanuck]

* No user-installed Apps get unrestricted CPU access and multitasking, potentially lowering CPU overhead significantly.

Happy?

Think your argument for different architectures / operating systems is a much better argument. The overhead of multitasking is still there, the iOS device still uses it for it's own uses and own applications. The only thing is remove the potential of 3rd party applications running in the background, but given that the CPU even on the "OS X" is within striking distance of 100% idle if you are not running any applications.... it is not going to have a substantial impact. I generally recommend running benchmarks after all applications are terminated to make sure applications are not interfering with the benchmarking -- which is easily to check out if you filter for the top range of benchmarks for any device.

 

[Michael Goff]

What's the point? What the benefit of ARM over Core M for a Macbook?

Is the rMB too thick? Too noisy? bad battery?

The answer is no to all three questions, so then why do this with all the downsides?

1. No x86 virtualisation.
2. Not powerful enough for x86 emulation so no Rosetta equivalent
3. No boot camp.
4. No drivers for 3rd party devices
5. No backwards compatibility for any apps.
6. Massive overhead for Apple having to maintain ARM OS X for one device.
7. Consumer confusion over all the above
8. Developer frustration.

Throughout all of this discussion, you never address the downsides and you're yet to put forth a single compelling reason as to why they should do this.

----------



Without the overhead of full multitasking and OS X...

Apple would get to control one more piece of the pie. There's also the idea of optimizing better, as well as there not being the potential for Intel to delay an update because of their chip issues. And Apple could likely charge less for this, as it would cost them less. As for the cost, they were testing an Intel OS X while selling computers with PPC.

 

[bkkcanuck]

Apple would get to control one more piece of the pie. There's also the idea of optimizing better, as well as there not being the potential for Intel to delay an update because of their chip issues. And Apple could likely charge less for this, as it would cost them less. As for the cost, they were testing an Intel OS X while selling computers with PPC.

They would get control of the pie for one more device, an orphan of sorts. When the ARM is sufficiently powerful to drive devices across the whole range of "OS X" range of devices -- then they will have "control".

You can be assured that they do have an equivalent Macbook running in-house that actually runs on the ARM. What they don't have yet is an ARM processor that can run on devices of the Macbook Pro and Mac Pro line yet (without a significant drop in performance). The ARM line is getting more powerful rapidly, while Intel has for the most part stagnated up until this point... which means some day it may happen..... we just are not there yet.

Apple is still dependant of companies that do the fabbing. Each generation becomes more and more expensive to fab. Lots of technical difficulties to overcome and there are very few companies that can fab at the leading edge (14nm/16nm) chips .... ones that seem to be still competitive at this stage are: Intel, GobalFoundries/Samsung, and TSMC. These companies put HUGE investments (ever increasing) in making the jump to the next size. Samsung lost 1 billion last year at their fab plants, the Apple order should give them enough volume to actually break even this year (maybe even name a profit). So they would still be dependant even with this illusion called control.

They would of course be able to save some money per device because of removing Intel from the equation -- but they would still have to pay for fabbing (and that is not necessarily cheap since companies that do fabbing spend billions and billions of dollars for each jump).

I am waiting to find out what is in store with the A9 processor...

 

[bkkcanuck]

they also need a JIT compiler for that llvm code, I mean why not then compile it to multiple architectures and solve the same problem.

They don't need a JIT compiler since llvm is not a vm but really a portable assembly language -- which has a "portable machine code" byte code. It would just need to translate this portable machine code into actual machine code at installation time .... not just in time.

With fat binaries you are carrying around binaries for multiple architectures that only serve to waste storage space.

In 2006ish there was no llvm to speak of, all they had were "big boned" (fat) binaries.

 

[KBO]

You must have never run a business. Take a look at this thing and tell me what wasn't re-engineered. Probably exactly one component: the audio jack.

There has been talk/rumors in iPhone forums that the audio jack will be dropped in the iPhone/iPad lines and incorporated into the lightning port.

Actually makes some sense but would require an adaptor or EarPod with a built in lightning jack.

 

[subsonix]

They don't need a JIT compiler since llvm is not a vm but really a portable assembly language -- which has a "portable machine code" byte code. It would just need to translate this portable machine code into actual machine code at installation time .... not just in time.

Yeah, but since it would require a re-compile anyway, my point was that you may as well just use the simpler fat binary. It's not like all code is contained in the binary anyway, dynamic linking means that a lot of stuff is contained in libraries and data can also be shared. Finally, it's not like binaries are a problem in relation to storage space anyway, data is.