The 11-inch MacBook Air: Faster than the old 13-inch MacBook Air

On paper, the new 13-inch MacBook Air shouldn’t be any faster than the old 2008 MacBook Air - at least in CPU bound tasks. The 2010 model gets a faster GPU but the CPU is literally the same 1.86GHz Core 2 Duo. Memory sizes and speeds haven’t changed either. While the SSD is faster, if you’re running a CPU bound benchmark there shouldn’t be any performance difference. If you assumed the same thing I did, you’d be very wrong.

Take a look at Cinebench 11 comparing the 2008 13-inch MacBook Air to the two 2010 models:

Cinebench 11 - Multithreaded CPU Benchmark
  11-inch MacBook Air (2010) - 1.4GHz Core 2 Duo 13-inch MacBook Air (2010) - 1.86GHz Core 2 Duo 13-inch MacBook Air (2008) - 1.86GHz Core 2 Duo
Cinebench 11 Score (Higher is Better) 0.81 1.1 0.70

Not only is the 2008 13-inch MBA slower than the new 13-inch model, it’s actually slower than the 11-inch model running at 1.4GHz. Something is amiss. Perhaps it’s just this benchmark?

I turned to our Handbrake H.264 encode test to verify my sanity:

Handbrake 0.94 - H.264 High Profile Transcode
  11-inch MacBook Air (2010) - 1.4GHz Core 2 Duo 13-inch MacBook Air (2010) - 1.86GHz Core 2 Duo 13-inch MacBook Air (2008) - 1.86GHz Core 2 Duo
Average Frames per Second (Higher is Better) 1.14 fps 1.55 fps 0.96 fps

Once again, a similar situation. The old MBA is actually slower than the new 11-inch, despite the advantage in CPU speed.

It looks like what we’ve stumbled upon is a combination of Apple aggressively throttling the clock speed of the older MacBook Air CPUs to meet thermal requirements, and the CPUs used in the new MacBook Airs being far better behaved from a voltage/power consumption standpoint.

The 45nm process these Core 2s are built on is as mature as it’s going to get. I’m guessing yield on these parts is as high as can be and as a result, power consumption is probably consistently lower than the original 1.86GHz parts Apple shipped back in 2008. The peak thermal specs themselves haven’t changed, but the actual power characteristics have.


The 2008 MacBook Air under Load

Maximum temperature, at least reported by the MBA’s sensors, isn’t any lower on the new 13-inch than the 2008 I compared it to. Both CPUs hit roughly 84C (183F) under full load. But look at what happens to the chips after a minute at that load:


2010 13-inch MacBook Air

2008 13-inch MacBook Air

iStat Menus reports the 1.86GHz Core 2 Duo in the 2008 system consuming only 7.59W, while the same CPU in the 2010 machine is drawing 11.45W. The 2008 machine is throttling back to reduce overall temperature while the 2010 system keeps going.

As a result, even the 11-inch MacBook Air will probably end up being as fast, if not faster than the 2nd generation 2008 13-inch MacBook Air. And our performance results confirm that:

11-inch MacBook Air (2010) vs. 13-inch MacBook Air (2008)
  Application Launch Test Adobe Photoshop CS4 Aperture RAW Import Cinebench R10 - 1CPU Cinebench R10 - XCPU Quicktime H.264 Transcode
Apple 11-inch MacBook Air (2010) - 1.4GHz Core 2 Duo 27.8 seconds 72.4 seconds 1.29 PPS 1612 2967 33.1 fps
Apple 13-inch MacBook Air (2008) - 1.86GHz Core 2 Duo 26.4 seconds 71.3 seconds 1.20 PPS 2046 2882 30.1 fps

You'll notice the less CPU intensive tasks are quicker on the old 13-inch system as the CPU isn't able to get hot enough to trigger Apple's throttling. The single threaded Cinebench test is the best example of this. The 26% performance advantage jibes with the 33% increase in CPU clock speed (it's actually a little low, most likely because the old CPU still isn't running at full speed even in this test). But now look at the heavier tests - the multithreaded Cinebench test and the Quicktime encode. Both of these stress both cores and drive TDP up, which forces Apple to pull clock speed back down. We wondered how Apple was able to cram such a high speed CPU into such a thin chassis as early as it did, now we know.

In practice I found the 2008 13-inch MBA launched applications quicker (short bursts of full clock speed), but after prolonged use or completing CPU intensive tasks it was tough to tell apart from the new 11-inch. What's even more troublesome is that Apple's aggressive clock throttling went relatively undetected until now. This is something I'm going to have to devise tests for and pay more attention to in future reviews. Sneaky, Steve, sneaky.

External Temperatures and Noise

The old MacBook Air chassis had a few dozen slits cut out of the aluminum for ventilation. The new MacBook Air chassis hides the ventilation slits between the base of the unit and the hinge. You can’t see them, but they’re there.

The slits are smaller than they were on the old chassis, which means moving air through them at the same rate sounds louder than before.


The 11-inch MacBook Air under Load

The CPU in the 11-inch model, even when under heavy loads, likes to stay at or below 72C (162F). At that temperature, the internal fan doesn’t spin above 4000 RPM (usually down below 3400). The 13-inch MacBook Air however is far more likely to generate noise. Running our simple Handbrake test the CPU will peak at over 82C (~180F) and the system’s internal fan will ramp up to over 6K RPM to compensate. Not only does the chassis get hot, but the fan gets audible. It’s still too small of a fan to really be considered loud in the grand scheme of things, but it’s loud enough to be annoying.


The 13-inch Macbook Air under Load

The surface temperature of the new 13-inch MacBook Air easily gets as high as the 2008 model. I measured a peak of 38.9C (102F) on both the 2008 and 2010 13-inch models. The 11-inch never broke 38C (100F).

Even casual use can ramp up temperatures pretty quickly. Just having a few websites open in the background that use Flash or other CPU intensive elements can slowly cause the MBA’s internal temperatures to rise. And now you’re beginning to see why Apple doesn’t install Flash on these things by default.

The danger zone is the upper left corner of the system, near the hinge. There’s only a single fan that cools both the CPU and GPU in the Air’s very tight enclosure. This is where that fan sits.

The Screen: Very Good Performance
Comments Locked

185 Comments

View All Comments

  • khimera2000 - Wednesday, October 27, 2010 - link

    I love me m11x R2 :D looking at this thing... im happy that its still speedier, and i dont care who you are something as thin as the macbook would freak me the !@#$ out if i droped it... more so then the one i have right now. (mono frame = expensive fix)
  • The0ne - Wednesday, October 27, 2010 - link

    "I really like the form factor of the 11-inch MacBook Air. It's great to carry around. It's like an iPad for people who have to get real work done. I just wish it was faster. If Intel made a 32nm Core 2 Duo, clocked high enough the 11 would be perfect. I guess that’s what Atom is eventually supposed to be, but right now the performance is just too low."

    Essentially this means the 11" MacBook is NOT suitable unless you want to wait and wait. You won't get any "real work done" by any means as it is. So why even bother to praise it and at the same time downgrade it.

    A netbook is both usable AND CHEAP. These are not and thus should not be called or even be consider netbooks. That's just crazy talk there. Might as well call all the rest of the ultra light notebooks netbooks.
  • kmmatney - Wednesday, October 27, 2010 - link

    It has an SSD, so no, you don't need to wait and wait. I agree with Anand's review - it is like a netbook in terms of portability, but its much faster, and has a much better screen. It's certainly more usable than the average netbook.
  • KarateBob - Wednesday, October 27, 2010 - link

    Good review so far, until you listed the system temperatures in Fahrenheit. I understand Apple lists the temps in (F), but the industry-standard units for temperature reading is Celcius. It's what most enthusiasts can understand. (ie. We 60C is getting hot for a pre-i7 CPU, but I couldn't tell you what 60C is in F)

    Can you please add Celcius numbers to the review, perhaps next to the Fahrenheit numbers, it will make the article much more comprehensible. Thanks
  • Sufo - Wednesday, October 27, 2010 - link

    I was just about to make this point - good thing i read through the comments before hand heh.

    I was very very disappointed by the use of F here at all, let alone it being the only scale used. Technical hardware reviews, for me, fall into the realm of science (albeit loosely) and C is the de facto standard in the scientific world (well, at least while it isn't K). Many component monitoring applications do not even have the option to display temps in F, and if they do, it is rarely (if ever) the default selection.

    So +1 to the request for at least displaying both numbers in future, and perhaps you could consider dropping F completely as i'd argue it has no context within the world of computing hardware.
  • Anand Lal Shimpi - Wednesday, October 27, 2010 - link

    Temperatures in C as well as F are both present now :)

    Take care,
    Anand
  • SraCet - Wednesday, October 27, 2010 - link

    To the reviewers and many commenters, it is tedious to listen to you divide the world into people who do "real work" and people who just fool around with IM and check Facebook.

    I do software development and scientific computing and for my purposes, the 11.6" MacBook Air is more than powerful enough.

    Reading your review, it sounds like the only things you consider "real work" are editing 12 megapixel photos, doing 3D renders, and (ironically) playing 3D video games.

    5 years ago, a 1.4GHz Core 2 Duo would have almost been the fastest CPU money could buy. Are you saying that people only started doing "real work" with their computers sometime in the last 5 years?

    Sorry, let me go back to writing code and running simulations, and stop interrupting your "real work"--i.e., resizing your pretty pictures because you took them at 20 times the resolution you actually needed for web publishing.
  • Sufo - Wednesday, October 27, 2010 - link

    "5 years ago, a 1.4GHz Core 2 Duo would have almost been the fastest CPU money could buy. Are you saying that people only started doing "real work" with their computers sometime in the last 5 years?"

    Well, only if you interpret his comments as suggesting that these tasks are _impossible_ to perform on the 11" MBA. Of course, that would be a gross misinterpretation - as i'm sure you can see. A more relevant extrapolation might be that 5 years ago, performing said tasks was a sluggish and intolerable chore - and on that we probably see eye to eye.

    As for the "real work" slur - i can understand your frustration, however you must realise that a large part (probably the largest part) of the non-casual MBA-buying demographic will be people who consider the "real work" of the article as well... "real work" - and you can't get angry at the reviewers for trying to include usage statistics tailored to the people most likely to be buying the device. If anything it is to their credit. Similarly, it is unrealistic to expect them to cover every single usage scenario.
  • SraCet - Wednesday, October 27, 2010 - link

    I suspect a very small percentage of professionals do anything that would stress out a Core 2 Duo. Most people do word processing, spreadsheets, presentations, e-mail. Editing and compiling code (developers). Accessing other computers remotely (IT types). Web browsing for business, like arranging travel. etc. etc.

    According to the reviewers, and apparently you, all of this stuff can be lumped in with "casual" (your word) use and is not "real work" because it can be done without taxing a dual core 1.4GHz processor.
  • Sufo - Friday, October 29, 2010 - link

    Ah, you've somewhat misinterpreted me here - the term "non-casual" was merely meant to identify people who use their machine for work purposes. Perhaps i should have used "professional". My overall point (as misguided as it may be) was really only alluding to the generalisation that most people who buy macs are artsy, journalist types - for whom editing pictures and obscenely flash-heavy websites etc is their normal, "real" workload. And yes, i realise this thread is dead :)

Log in

Don't have an account? Sign up now