The low power consumption and utter silence of the Zonbu are very appealing.
Amusingly, the Zonbu has caused me to notice how ungodly loud the water cooler is in my dining room.
But do these environmental benefits come at a significant performance and boot time cost over a regular PC? I set out, stop watch in hand, to find out how the Zonbu compared with my desktop PC in this regard…
Start your engines
The first test was to measure the length of time it takes to boot both machines from a cold start – literally from when I pushed the power button.
My Zonbu boots from power-on to the login screen in 1m 30s. My login screen has my username saved, so after typing my password I can continue and it takes an additional 28s to get to a usable desktop, for a total time of 1m 58s.
Not exactly speedy.
Pitching a lower powered system as a replacement for traditional PCs means fighting common perceptions about the sacrifices that will have to be made. Demonstrating rapid boot time will go a long way to overcoming that objection.
Isn’t it interesting how much of our observations relate to psychology more so than technology? Marketing is a funny thing.
I would suggest this be an area of focus, I have a hunch there are some optimizations in the boot process that could lead to shaving a significant amount of time off that start-uptime.
Desktop XP PC
By comparison, my desktop PC went from power-on to a usable desktop is 1m 10s, 48s faster than the Zonbu. To be fair, I don’t have login enabled on my XP desktop, but then neither do a lot of people. Also, my XP machine doesn’t connect to any network shares, while the Zonbu is (I assume) establishing a connection with the S3 storage servers.
For the purpose of this experiment, the machines were using the same physical ethernet cable, which I moved between boxes to test. Everything downstream of the ethernet cable was identical.
Winner: Windows Xp – 1m10s versus Zonbu 1m 58s
The Checkered Flag
When it came to shutting the machines down, there was a significant difference as well. It took the Zonbu 39s to go from the desktop to being powered off, while it only took the XP machine 16s to clean up and shut down.
I’m not sure how important this particular metric is, as I usually walk away as soon as I click the shutdown button anyway, but I did want to be thorough. The things I do for you, my readers.
Launch your browsers
Next up I wanted to compare application launch times. I configured Firefox to open to a blank start page and cleared the cache.
The Zonbu launches Firefox in 1.7s, while XP launches Firefox in 1.5s. While my XP box has a slight edge, it is virtually imperceptible to me as a user.
My Zonbu has 512MB of RAM and a 1.2 GHz processor whereas my desktop has 2GB of DDR400 RAM and an AMD Athlon 3800+ 2.4GHz processor. Given this, The Zonbu’s comparitive performance is excellent.
Extra! Extra! News site load times
Next up I wanted to compare page load and render times between the boxes. I spend the bulk of my day with multiple tabs open in Firefox, and the Zonbu is really meant as an Internet terminal/appliance, so this is a key metric.
To ensure consistent performance, I pre-loaded all the pages in Firefox, which caused the key components to populate our ISP cache at the cable TV office. Doing so ensures a consistent load and render comparison on both platforms, and also ensures consistent DNS performance/caching.
I then flushed the locale Firefox cache (CTRL-SHIFT-INSERT) and shut down Firefox, before re-launching it to the blank start page.
The results are formatted as Site/Zonbu/XP. Times in seconds.
As you can see, the XP machine has a slight edge, usually besting the Zonbu by anywhere from .3s up to 1.5s with Google. The larger discrepency on the Google site is due to compression we utilize, see the note below. Being text heavy the Google sit is disproportionately affected by it. It looks like the normal delta between the two machines is around .6s at most.
We make aggressive use of GZip compression in our web caching architecture, and as such the browsers have to dynamically un-zip the text portion of web pages, which uses processing power and adds a bit of delay but generally makes plain text content move extremely fast on the network.
I did not expect the Zonbu and the XP desktop to be quite so close on page load/render times. I think the difference will probably not even be noticeable to most end users.
The boot process does require some futher scrutiny and optimzation to get it closer to the XP start-up time. Almost 2 minutes from power on to usable desktop is not ideal but not horrific either.
One thing the Zonbu team can immediately investigate is shortening the start-up audio file that is played when the desktop launches. For some reason playing this file back sucks up 100% of the CPU, stopping you from launching any applications until the sound finishes playing.
Going to a lower bit rate for the clip and shortening the length (which is about 40% too long, time wise) would buy you a few seconds immediately. Just a thought.
The Zonbu is still happily sipping just 9W of power as I write this, running in complete silence; I can honestly say I’m happy to give up .5s to 1s of page load/render time for the savings in power and noise.
Most importantly, I don’t feel like I’m really sacrificing material performance to get those economic and environmental benefits.