ReadyBoost Flash Drives Lack Significant Boost
How We Tested ReadyBoost Drives
We designed our tests to address ReadyBoost's central claims. For example, Vista's SuperFetch feature is supposed to monitor which applications you use the most and preload them into your system memory, thus shortening application launch times. So to help our test systems "learn" our usage tendencies, we conducted multiple passes of our application launch test.
We tested ReadyBoost with two systems--one desktop and one laptop--both of which came loaded with 1GB of RAM. Microsoft imposes a minimum system requirement of 512MB of RAM for installing Vista Home Basic, but Vista Home Premium, Vista Business, and Vista Ultimate all require 1GB of RAM. From our previous tests of Vista systems, we've confirmed that 1GB is the bare minimum for running the Vista operating system. Fortunately, any system that meets Vista's minimum CPU specifications should be upgradable to 1GB of RAM.
The benchmarks we run include application performance tests based on PC WorldBench 6 Beta 2, a hand-timed application start test, and a file read and write (copy) test.
The three test systems we used were as follows:
Test Bed 1: ReadyBoost Test (Desktop PC)
Computer: HP Compaq dc5750 desktop
PC Processor: 2-GHz AMD Athlon 64 X2 3800+
Graphics: ATI Radeon X1150
Hard drive: 80GB Samsung HD080HJ/P
Operating system: Microsoft Vista Business (32-bit)
Test Bed 2: ReadyBoost Test (Notebook PC)
Computer: HP Pavilion tx1000 notebook PC
Processor: 2-GHz AMD Turion 64 X2 TL-60
Graphics: nVidia GeForce Go 6150
Hard drive: 160GB Seagate ST916082
Operating system: Microsoft Vista Home Premium (32-bit)
Test Bed 3: File Read/Write Test
Computer: CyberPower desktop PC
Processor: 3.4-GHz Intel Pentium D 950
Motherboard: Asus P5WD2-E Premium
RAM: 1024MB DDR2-667
Graphics: 256MB nVidia Geforce 6200 PCI Express
Hard drives: two Hitachi Deskstar HDT722525DLS380 drives, using Intel ICH7R SATA Raid 0
Testing Methodology: ReadyBoost Test
We created a disk image that contained a version of our WorldBench 6 Beta 2 benchmarking utility, along with Adobe Photoshop CS, Microsoft Office 2003, and Yahoo Jukebox (we also used Microsoft Internet Explorer for our tests). In its default configuration, WorldBench 6 Beta 2 disables prefetching in order to avoid obtaining inconsistent benchmark results; but the version we used for ReadyBoost testing kept prefetching enabled, because that was the feature we were trying to evaluate.
We launched two applications, one after the other, immediately after booting, timing how long each operation took. Then we waited 5 minutes before starting up the next two applications (again, sequentially), timing each operation. Then we closed all four applications and restarted the first two applications. We waited 5 minutes more, and then restarted the second two applications, again timing each operation.
We ran seven passes of this 30-step test script to allow SuperFetch to learn the applications that we were using most frequently. Before starting a new drive/memory configuration, we restored the system from the image we had created.
We waited 5 minutes between application launches because Windows loaded other applications at startup, and the results of the timed tests for booting up and for loading Photoshop CS and Excel before completion of the 5-minute waiting period varied significantly. Consequently, rather than count those timed results, we counted only the results after the first 5-minute wait.
The reported results of our hand-timed ReadyBoost tests reflect only the sixth and seventh passes of the test script, because those passes were the ones most likely to reflect SuperFetching of the application we used to test. We averaged the times for the sixth and seventh passes.
Testing Methodology: WorldBench 6 Beta 2 Test
We set up the test-bed system with the standard WorldBench 6 Beta 2 settings. Then we ran three passes of the test, and saved the results.
Testing Methodology: Read/Write Test
We copied a folder containing files and folders (approximately 747MB of total content) by dragging and dropping to and from the desktop. We timed both of these operations. The results of this test were determined by throughput and by seek time, as the files and folders were spread out on the drive. We performed each of the tests three times, with reboots in between, and then averaged the results.
We also copied a zipped, uncompressed 747MB file by dragging and dropping it to and from the desktop while timing the operation. This test mainly measured throughput, since a single contiguous file was involved. We performed each of the tests three times, with reboots in between, and then averaged the results.
ReadyBoost Flash Drives Lack Significant Boost