The Practice of Code

Updated with benchmarks for 2010 MacBook Air with Encryption as well as clearer charts and data

I purchased a 2010 11.6" MacBook Air earlier this year with the maximum, possible specs: 1.6 Core 2 Duo with 4 GB of RAM and a 128 GB SSD. I took a performance hit when I switched from my aluminum unibody MacBook (which had a 2.4GHz Core 2 Duo), but it never felt slower. As a matter of fact, it always felt faster, to which most, if not all, will agree was due to the switch from a traditional hard drive to an SSD. Shortly after I purchased it in February, there were some reports that Apple was using different SSDs across the line, some made by Toshiba, and some made by Samsung, and the Samsung ones were faster. I had a Toshiba. Regardless, while I was able to condense my entire life to fir on 128 GB, down from 250 GB, I had no breathing room, so when the 2011 MacBook Airs were announced, I upgraded again.

So I ordered a fully loaded 11.6" model, this time with a 1.8 GHz Core i7 with 4 GB of RAM and a 256 GB SSD. While awaiting its construction and arrival, some initial benchmarks had been released on the web for Lion’s full disk encryption, aka FileVault 2. This was something I was very interested in, but the results seemed inconclusive to me, and none were specific to the upgrade I was awaiting. So, when my new machine arrived, I decided to record and publish my own benchmarks.

Now, I am not a scientist, so please expect some human and environmental error here. First, the machine’s raw specs:

2010 MacBook Air
1.6 Core 2 Duo
4 GB of RAM
128 GB SSD (identified as an APPLE SSD TS128C)

2011 MacBook Air
1.8 GHz Core i7
4 GB of RAM and a 256 GB SSD
256 GB SSD (identified as an APPLE SSD SM256C)

I used XBench 1.3, only ran the disk test, ran it 3 or 4 times, and if any set of results was drastically different from the rest, I threw it away and averaged the rest.

In both cases, I had a pretty pristine admin account (I create an admin account at first setup, and basically never touch it again). In the case of the 2010 MacBook without Encryption test, and both FileVault tests, both drives had about 100GB of data on them. This differs from the 2011 MacBook without Encryption test as this was done relatively shortly after first turning on the new machine.

So, without further ado, the results:

Here is the summary:

Here are the details:

So the jump from the 2010 MacBook Air Toshiba to the 2011 MacBook Air Samsung is pretty sizable (almost 65% faster at random tests, 40% overall). The dip in performance from enabling Lion FileVault on the 2011, while not drastic, is also not insignificant (18% overall), so that makes me personally quite happy, as I was able to double my storage capacity, add encryption, and still have an overall performance improvement. However, the big story here is the dip in performance for the 2010 model with FileVault enabled, as the drop is much more sizable at 44%. Unfortunately, it’s difficult to know the exact cause, as there are a lot of variables at play here, including but not limited to:

1. General machine architecture (Sandy Bridge vs. its predecessor)
2. CPU type (The Core i7 has built-in technology to accelerate AES encryption)
3. SSD make and manufacturer (Toshiba vs. Samsung)
4. Unscientific testing

My personal take from the numbers is that Lion FileVault comes at a large performance hit on Core 2 Duo machines w/Toshiba drives; a performance hit that I would personally find unacceptable. Inversely, the performance hit on a Core i7 machine w/Samsung drive is 3x less drastic, and thus, for me, well worth it. But, as with all things performance related, perception is relative, so if you’re weighing up an upgrade, run XBench on your own machine and base your comparisons on that.

I will still have the old machine for another week at least, and I’m happy to do any other tests or benchmarks, as well as answer any questions.