Tuesday, December 15, 2009

Macbook Air Part 1 - 128GB New SSD Benchmarks


This is the first article in a three part series on MacBook Air Solid State Drive Benchmarks.

I wanted a laptop that was light, portable but more powerful than a traditional netbook which typically have sub-par screens and underpowered graphics capabilities. I did my due diligence and I landed on a MacBook Air or (MBA). My MBA's specs are:

  • 13.3-inch LED glossy widescreen
  • Max resolution 1280 by 800
  • Intel Core 2 Duo 1.86 GHz processor
  • 2GB of 1066MHz DDR3 SDRAM (not SODIMM; built onto the motherboard)
  • 1066MHz Front-side Bus
  • 256MB of Shared DDR3 NVIDIA GeForce 9400M graphics card
  • Built in Bluetooth
  • 128 GB Solid State Drive (13GB used by the system) – which according to Anandtech is a Samsung device.
    • Model APPLE SSD SM128, Revision VAM0BA1Q
  • Installed Mac OS X 10.5.6 (Leopard) and the upgrade for Mac OS X 10.6 (Snow Leopard)
  • Built-in iSight camera, 1 USB port, 1 mini displayport, 1 audio port, 802.11n, etc.
Before purchasing a MBA I tried to find Apple SSD Benchmarks for the included 128GB SSD hard drive but my search didn't turn up anything relevant. I had previously read the Anandtech review of Solid State Drives, most interesting was the part on Samsung drives installed by OEM's (like Apple) which are usually the worst in terms of random write performance once the Solid State has been used.

It turns out new versus used performance is something very important when evaluating solid state drives. Almost all SSDs perform well when they are brand new but because of their design (Solid State Drives read in 4KB "pages" and write in 512KB "blocks") once the drive has been filled they are forced to perform more complex read-modify-writes (instead of read-writes) which causes performance to decrease. Since I couldn't find anything out there I thought I'd do a little benchmarking myself.

I've never done benchmarks for a hard drive (or anything for that matter) and I found it very difficult to locate suitable hard drive testing software for Macs. I decided to use a few simple scenarios for testing:

  1. New Drive – running OS X 10.5.6
    1. Cold Boot Times, Shutdown Times, and Time Launching Apps including iTunes, Safari and Firefox.
  2. New Drive – running OS X 10.6
    1. Cold Boot Times, Shutdown Times, Time Launching Apps including iTunes, Safari and Firefox.
  3. Used Drive – running OS X 10.6 (elapsed time: 1 week)
    1. Same as #2
  4. Formatted Drive – Dual boot with OS X 10.6 and Windows 7 (elapsed time: > 4 months)
    1. Same as #2 minus the Benchmark Apps, Windows 7 Performance Index
In order to have true benchmarks I needed something to compare, so for my SSD benchmark tests I compared basic times of the new versus used drives for the operations listed above. I also compared the used times versus a reformatted system to see if formatting an SSD returned it to a "new" state.

In order to recreate a "used" drive and its performance measurements, I copied a bunch of files to the MBA's hard drive until it there was no space left (actually there was about 500KB available). Then I deleted the files. I thought this would ensure the SSD's performance would match a drive that had been completely written to since each flash block had been filled with data.

The Initial Results:

  1. Brand New Drive - running OS X 10.5.6
    1. Cold Boot Time was about 27.3 seconds
    2. Shutdown Time was about 4.1 seconds
    3. Time Launching Apps
      1. iTunes was about 4 seconds
      2. Safari was about 1 second
      3. Firefox was about 1 second
  2. Brand New Drive - running Snow Leopard - OS X 10.6
    1. Boot Time was about 29.3 seconds
    2. Shutdown Time was about 2.3 seconds
    3. Time Launching Apps
      1. iTunes was about 1.5 seconds
      2. Safari was about 1.4 seconds (40% increase)
      3. Firefox was about 2 seconds (100% increase)
  3. Used Drive - running Snow Leopard (elapsed time: 1 week later)
    1. Boot Time was about 31 seconds
    2. Shutdown Time was about 1.9 seconds
    3. Time Launching Apps
      1. iTunes was about 2.1 seconds
      2. Safari was about 1.75 seconds
      3. Firefox was about 2.3 seconds
I also ran a few benchmark programs on the new hard drive but didn't run them on the "used" (and still haven't).

  1. CineBench - Rendering (1CPU) 1572 CB-CPU, Rendering (x CPU) 1143 CB-CPU, Multiprocessor Speedup: 0.73x, OpenGPL Standard 2512 CB-GFX
    1. OpenGL Standard Test CB 2512
    2. Single CPU Render Test CB 1572
    3. Multiple CPU Render Test CB 1143
  2. GeekBench - Geekbench Score: 2661
    1. Processor integer performance: 2095
    2. Processor floating point performance: 3710
    3. Memory performance: 2357
    4. Memory bandwidth performance: 1583
As you can see from the initial results the "used" SSD (after about a week of testing) didn't have a very big performance decrease. In hindsight I probably should have tried to encode a video or mp3, something that would require writing to the hard drive in large amounts of data. Although the times for launching a few applications increased it wasn't a big slowdown.

After I formatted the drive, put Windows 7 on it and used it for a few months I noticed a much more dramatic decrease in performance. Check out the subsequent results.


Related Articles:
Macbook Air SSD Part 2
Macbook Air SSD Part 3
Macbook Air vs. Macbook Pro SSD

4 comments:

Clint said...

I stumbled on your blog while searching for a comparison on a Pro with the 9400 graphics card vs. a MBA (macbbok air) with the same card and same amount of ram. I cant find anything. Your thought?

Wouldnt the geekbench be the same to the same specs or am I missing something?

Chris Kenst said...

Hi Clint,

Yeah the Geekbench would be the same, which is why I won't run them again. Let me see if I can get my friend to run a few benchmarks on his Pro with dual 9400's so you'll have a comparison.

Clint said...

Thanks for your quick reply. While theres no dual 9400 in any pro that Im aware of (SLI?) I would need someone to run geekbench on a pro and mba with simular specs, especially ensuring each machine has the 9400 and 2gb ram.

Chris Kenst said...

I didn't mean dual 9400s, I meant 9400M + 9600M. The Macbook Pro I would compare against has both graphics cards because it is the top of the line model. It would be hard to make a comparison because of the cards plus his regular hard drive, while I have one graphics card and a solid state on my MBA. Oh well.