MacBook Air Performance: SSD vs. Mechanical HDD

We took the same MacBook Air and just swapped drives for running these tests, you can't get more scientific methody than that. First up are some synthetic tests to help set expectations, for that we turned to XBench 1.3.

XBench isn't a particularly good benchmark for OS X, but it does have a basic drive performance test that suits our needs:

XBench 1.3 256KB block Disk Tests 80GB 4200RPM HDD 64GB SSD
Sequential - Uncached Writes 28.4MB/s 24.6MB/s
Sequential - Uncached Reads 28.0MB/s 49.9MB/s
Random - Uncached Writes 20.6MB/s 17.8MB/s
Random - Uncached Reads 13.0MB/s 49.2MB/s

The trends are pretty clear here: write speed is about 17% faster on the mechanical HDD while read speeds are much higher on the SSD, particularly when they aren't sequential reads. Since the mechanical drive has to worry about locating data on a spinning platter, random data spread out over the relatively slow spinning platter takes time to access. The SSD benefits from having an equal access latency to data regardless of where it's located in the drive's flash memory.

While most desktop applications are quite sequential in nature, multitasking can change I/O access patterns considerably.

Our first set of real world tests on the SSD are basic stopwatch application launch tests. We took 7 applications and timed how long they take to start up on the mechanical drive vs. the SSD:

Application Launch Time Tests 80GB 4200RPM HDD 64GB SSD
Adobe Photoshop CS3 18.0s 6.9s
iWork '08 - Pages 11.0s 3.5s
iWork '08 - Keynote 13.0s 6.3s
iWork '08 - Numbers 7.1s 3.5s
Mail 4.0s 2.6s
Microsoft Word 2008 28.8s 11.0s
Safari 2.9s 1.0s
System Boot Time 54.4s 32.5s

The SSD completely destroys the mechanical disk in application launch times, and these results aren't just numerical, they are very noticeable in using the system. The SSD is noticeably faster in application launches, accessing files in Finder and spotlight searches. After I used the SSD MacBook Air, I tossed the standard HDD back in and honestly thought something was wrong - it felt significantly slower, despite feeling mostly "fine" before I was exposed to the SSD.

Most application level benchmarks however favor the higher write speeds of the mechanical disk instead:

Application Benchmarks 80GB 4200RPM HDD 64GB SSD
iPhoto Import 125.9s 128.8s
iPhoto Export 196.0s 201.0s
Pages Export 37.6s 41.1s
Keynote Export 25.0s 23.7s
Word 2008 - Compare Docs 107.8s 109.3s
PowerPoint 2008 + Word, Print PDF 149.0s 162.6s
File Decompression 103.7s 138.4s
Photoshop CS3 76.2s 79.0s
Quicktime H.264 Encode 5.6m 5.8m

Most of these benchmarks show the SSD as slightly slower than the standard HDD, but a couple are noticeably faster on the standard drive thanks to its write speed advantage. The SSD recommendation would be an easier one if the benchmarks clearly leaned in one direction or another, but it's not as difficult as you might think on first glance.

The split between reads and writes on a desktop system is biased towards read performance, so you're more likely to notice the SSD's faster read speeds than its slower write speeds. The theory was echoed in my real world usage of the machine, the SSD was just faster.

Installing the SSD The Impact of SSD on Battery Life


View All Comments

  • aliasfox - Thursday, February 14, 2008 - link

    It may be 50% faster than the 1.2 GHz ULV processors in most other ultraportables, but that also means it's also about 50% slower than most mainstream high end CPUs (2.2 GHz and up).

    Slow hard drive doesn't help either.
  • sprockkets - Wednesday, February 13, 2008 - link

    I don't get it. I can use my Pioneer laptop slot loading burner on an external enclosure and it be powered solely via the one usb port and works just fine, even while writing to dvds.

    For that matter, I can do the same with notebook laptop drives. Rarely do I need both usb ports connected in order for it to work.
  • sprockkets - Wednesday, February 13, 2008 - link

    Only about 16mbits per second on the xfer rate on wireless? Wow. Were you using n? Reply
  • JarredWalton - Thursday, February 14, 2008 - link

    I don't know what Anand used, but I've seen everything from 3MBps to 12MBps on an 802.11n network... all with laptops in the same room, and many using the same chipset (Intel 4965AGN). Overall, N tends to feel about half as fast as 100Mbit Ethernet - or about twice as fast as 802.11G. Router choice unfortunately still has a major influence on 802.11N performance. Reply
  • Imaginer - Wednesday, February 13, 2008 - link

    It is what the overall feel and message of the article indicates to me. It is a bit expensive, non-expansive, lacks most utility to be used in most mobile situations, and it is purely for computing on the bare minimum.

    Kind of like that so called weekend car. I don't think I myself will have ANY need for such a device. Give me a powerful desktop and a versitle yet remaining non cumbersome notebook anyday.

    Most people in the market for a laptop usually would use it like their normal away from home computer and because of this, the air really disappoints. (not that I would invest in a new computer anytime soon).
  • jedmitchell - Wednesday, February 13, 2008 - link

    hey, so good review overall -- kept it very even handed considering the difficulty of reviewing a niche product like this. one hardware point I'd like to mention though is the info you give on the X3100. certainly the idea behind it is that as an integrated controller it won't provide very fast graphics, but there's a trick here: most of the things it's not rated to run... run. at least on the older macBook (santa rosa). final cut pro, maya, and photoshop actually all run pretty seamlessly on the X3100, both in OSX and windows (fcp is more memory/drive limited there than GPU). the only small problems are in windows where the X3100 drivers by intel are actually lacking several openGL 2.0 features present in apple's version.

    the X3100 even plays older games on windows without much trouble -- I can run the Orange Box games at 1024x768 with high quality settings and see a fairly regular 30fps, less a few texture memory glitches. anyway, it would be interesting to see how that performance in the same chipset scales from the macBook to the air.
  • jdwango - Wednesday, February 13, 2008 - link

    However I wish you had also tried to install Windows XP/Vista via boot camp and reported your thoughts. Reply
  • joey2264 - Wednesday, February 13, 2008 - link

    This would be a fairly good review if you would just mention the fact that most of the sacrifices Apple made to create the Macbook Air simply weren't necessary. If you look like at a notebook like the Fujitsu Lifebook S6510 of the Lenovo X300 this becomes clear. Looking at these two notebooks, it is obvious that each of the manufacturers could have come up with a 13.3 in, 1 spindle notebook that didn't make hardly any other compromises (decent keyboard, decent port selection, replaceable battery, upgradeable memory, standard 2.5" hard drives (Lenovo could have probably fit a 2.5" hard drive in there if they had used a 13.3" screen, with the requisite larger footprint, although it would have been a little heavier), etc). Reply
  • michael2k - Wednesday, February 13, 2008 - link

    The S6510 you mention is heavier (by a pound) and nearly twice as thick! It is much more comparable to a MacBook (5 pounds and an inch thick vs 4 pounds and 1.42 inches thick).

    The X300 is also not available yet, so a comparison will have to wait until we find out about price and build quality.
  • mlambert890 - Thursday, February 14, 2008 - link

    OK, so then the Sony TX, the Fujitsu P7k, the Toshiba Portege, the Dialogue Flybook, the Panasonic Toughbook, the Dell XPS1210, the Sony SZ, The LG XNote....

    There's a pretty long list of notes that are smaller and ligher or as light or slightly heavier with a lot more features than the MBA.

    The MBA is THINNER. Last I checked thinner is a BS feature. When someone can explain to me WHY thinner means ANYTHING beyond looking cool at Starbucks, maybe Ill be interested.

    The Sony X505 was pretty much the same situation as the MBA except it had a removable battery and more ports and that was 3 years ago. I think the MBA was like .2" thinner than the Sony *at its thinnest point* and about the same at the thickest.

    The MBA is big news for the cult of Mac which lately is including PC sites like this.

Log in

Don't have an account? Sign up now