F
Field Nurse
Guest
Last week we timed the Windows 7 Preview, XP, and Vista's startup times on a single PC triple-booting the three generations of operating system. However, our tests had their flaws, we made a big boo-boo by not including Vista's SP1 in the mix, and you all suggested ways to make them more thorough and accurate in the comments. Taking your lead, we evened up the playing field, installed the Vista service pack, added a set of tests that take Windows all the way to a fully-loaded user desktop, and updated our numbers. Turns out that the Windows 7 Preview is indeed faster than both Vista and XP, but not at all by the margin we originally published. Get our updated numbers and test details here.
table.comparison { border-width: 1px 1px 1px 1px; border-spacing: 2px; border-style: solid solid solid solid; border-color: gray gray gray gray; border-collapse: separate; background-color: white; } table.comparison th { text-align:center; border-width: 1px 1px 1px 1px; padding: 1px 1px 1px 1px; border-style: inset inset inset inset; border-color: gray gray gray gray; background-color: #cccccc; -moz-border-radius: 0px 0px 0px 0px; } table.comparison td { text-align:center; border-width: 1px 1px 1px 1px; padding: 1px 1px 1px 1px; border-style: inset inset inset inset; border-color: gray gray gray gray; background-color: white; -moz-border-radius: 0px 0px 0px 0px; } table.comparison td.has { background-color:#8EBB8E; } table.comparison td.label { font-weight:bold;text-align:center; } <div> The hardware: My test system has an Intel Core 2 Duo processor running at 3.16GHz with 4GB of RAM. Windows XP, Vista, and the Windows 7 Preview are all installed on the same, physical SATA hard drive (unlike the last test, which had Win7 on an IDE drive). The version of Windows 7 that I'm using is the Preview release from last month's PDC conference, an incomplete, earlier-than-beta, for-testers-only edition, which automatically puts it at a disadvantage. Unlike the last set of tests, this time around I'm using Vista with Service Pack 1 installed (which it was not before—my apologies for that big, honking oversight).
The test: Since I'm triple-booting the three operating systems, I used a simple handheld timer to get my numbers, starting at the "Choose your operating system" screen (which comes after the BIOS startup) and ending at a fully functional desktop in my first set of tests, then the user login screen in my second. I timed each OS startup three times and averaged the results to account for my fat fingers (which I did not in the first set of tests).
Test Set 1: From Boot Choice to Fully-loaded User Desktop
For this test set, I created a non-Administrator user account with nothing in startup, and set Windows to automatically log into it on boot. These numbers start at the OS boot choice screen, and end at a fully-functional, user desktop. This isn't a true test of how long a desktop would render in real-world usage, because most people DO have programs in their startup. But these tests are to compare Windows version performance, not third-party software. That all said, the numbers:
Operating System Average time to reach desktop Windows 7 Ultimate
(Preview, PDC edition, 32-bit) 32 seconds Windows Vista Ultimate (32-bit, SP1) 33 seconds Windows XP Professional (SP3) 40 seconds Test Set 2: From Boot Choice to User Login Screen
This test measures the amount of time from the boot OS choice screen to the user login prompt only, hence the shorter times.
Operating System Average time to reach login prompt Windows 7 Ultimate
(Preview, PDC edition, 32-bit) 23 seconds Windows Vista Ultimate (32-bit, SP1) 24 seconds Windows XP Professional (SP3) 29 seconds The Results
As you can see, the Windows 7 Preview is just a HAIR speedier than Windows Vista, not the 20% we saw in the first set of tests (I attribute this to Service Pack 1), and way, way faster than Windows XP (8 and 6 seconds, respectively). While it's not the 20% speed increase I saw in my first round of tests, it still bodes well for Windows 7, which still has a lot of growing up to do from its Preview status.
Thanks to everyone who made suggestions on the first round of tests on how to make this more accurate and fair all around.