Myth Busters: Is Windows Search a True Performance Hog?

You’ve probably heard about this optimization tip a thousand times—disabling the Windows Search index will significantly increase PC performance and reduce hard disk activity. Is this actually true? Or is it not worth giving up this nice Windows 7 feature?

How does Windows Search work?

Right after you install Windows or turn your PC on for the first time, Windows Search creates an index of specific files, folders and other items, such as Outlook emails or Start Menu entries, on your hard disk. This index is located within the hidden “C:\ProgramData\Microsoft\Search” folder and usually takes up about 10% of the size of contents that has been indexed.

When you look for a file using Windows Search, part of the search index is loaded into the main memory (RAM) of your PC making extremely fast searches possible. Why? Instead of browsing through your entire hard disk or huge folders, Windows Search simply accesses the index and produces results immediately.

If you’re constantly looking for data, such as your emails or files, then Windows Search is pretty helpful. The fact is, disabling the Windows Search index results in longer search times—but only (and here’s the important part) for folders that are being indexed. These include the entire user folder (pictured below), all Start Menu entries, offline files, Outlook contacts, appointments and emails (if installed and used), OneNote notes (also if installed and used), and Internet Explorer history.

To see how much and what is indexed on your system, click on the Start orb, go to Control Panel, and type in “Windows Search” into the search bar. Then click on “Indexing Options”.

You’ll see the following window.

Depending on how much data you’ve got, the index might grow relatively large. The question is: Is shutting down the Windows Search index worth the performance benefit, or is it just the placebo effect? How much of the entire index is actually being loaded into memory? Let’s find out!

Preparing the speed tests

I benchmarked my day-to-day work machine—a laptop sporting a Core 2 Duo with 3.2 GHz, 4 GBytes of RAM and a 256 GB SSD hard disk—and a lower-end machine with a Core 2 Duo 1.86 GHz, 2 GBytes of RAM and a much slower 5,400 RPM hard disk. For future reference, I’m going to refer to them as the faster notebook and the slower notebook.

The faster notebook had nearly 40,000 items indexed. However, I added all of the folders on all of my hard disks to the Windows Search index! That should give some impressive results, as more than 300,000 files were added to it; it should also give you a good feeling of how such a large index can affect the speed of Windows machines.

The slower notebook had about 20,000 items indexed; this number is about right for those typical PC users who store some pictures, e-mails and documents on their computers. This should give you an idea of how much the search index really reduces performance in its default setting.

Note: Both laptops are running Windows 7. This version of Windows has been developed to perform faster in many areas—especially in terms of file searching.

Conducting the benchmarks

Like all of our previous performance tests, we took a lot of time to get the most exact results. In each and every case, we repeated the benchmarks three times. If there were major discrepancies, we repeated them another two to three times and calculated the average; for example, some of the PC Mark Vantage scores and Windows boot times always showed slightly different results. So, let’s get moving!

Testing Windows Search index activity

I added a couple of hundred files to the slower notebook and about 300,000 files to the faster one to see how much these machines struggled with a growing Windows Search index. Now, Windows is actually smart about this—when a user is working on the machine, indexing speed is reduced so that is does not impact performance. But, let’s check the numbers to see what panned out.

*Note: I kept the huge index with more than 300,000 files on the faster Notebook but reduced the index on the slower notebook back to its original state with 20,000 files.

The results were fairly reasonable; it didn’t matter if we added 500 or 300,000 files. Windows never slowed down to a perceivable pace—even an average CPU consumption of 10% did not noticeably affect performance. I can only imagine a scenario in which the PC is working under 100% usage for a consistent period of time in which this might have a small effect.

Windows Search index peaked at about 13%; this likely occurred when it needed to add larger files. Even then, the impact was minimal, and most importantly, the indexing was done in a very short time. How often do users actually add 500 files and need 100% CPU power for 25 seconds? And how often does the average user add more than 300,000 files to a folder that’s indexed by Windows Search? For me—this might happen once a year. In this case, allow Windows Search to do its work overnight, especially if you need full performance.

I even used Process Monitor to check how many events (e.g. registry accesses, file system operations, etc.) SearchIndexer.exe produced, while Windows added thousands of files to the index. Over the course of about 20–25 minutes, the faster notebook produced about 2.5 million events—and all I did during that time was listen to iTunes, write this very blog post and take three screenshots. Out of these 2.5 million events, Windows Search only produced 50,000. I think that says it all.

The bottom line: the effect on performance while adding files is absolutely negligible. So, let’s focus on performance while the search index is actually enabled.

Comparing the time to startup and shutdown

Does Windows Search have a significant impact on boot and shutdown performance? In theory, at least, it should slow things down a bit given that the service and the index needs to be loaded into main memory.

Windows actually booted just a bit faster with the higher-end notebook. Using Windows’ built-in event management tool, I was able to figure out that Windows Search delayed one out of every five boot processes by about nine seconds on the faster notebook and about 15–17 seconds on the slower notebook. I was not able to figure out why this happened only once in a while, but I guess it has something to do with maintaining and rebuilding the index during startup. On the other side, shutdown noticeably improved, which I didn’t expect at all.

Looking at virus scan performance

To check if and how Windows Search adds any extra load to typical file operations, I measured the time it took to scan a 1.2 GByte folder containing uncompressed TIF images and Word documents. Here’s how Windows performed with and without the search index.

There was no difference when it came to the virus scan performance (the activity is very intense on your hard disk and CPU).

Testing application performance

Outlook 2010 is a nice benchmark for testing application start-up performance, as it produces heavy file accesses, both on a single huge file (Outlook.pst) and several smaller files (the Outlook system itself and all of its plug-ins). Its resource consumption is comparatively high, so start-up time varies depending on the kind of PC you have.

On the faster notebook, Outlook did not start up faster after I disabled Windows Search. However, on the lower-end notebook, I could see a slight difference after I disabled Windows Search. Success!

Cinebench R11.5

Cinebench, based on the very popular CINEMA 4D animation software, is a leading benchmark for 3D animation performance. The benchmark generates the average frames per second or FPS (for OpenGL) and a score (for CPU) that represents the performance of a computer, which can be used to compare performances between different configurations. The higher the numbers, the better the performance of the PC.

There wasn’t any difference. I’m beginning to see that Windows Search isn’t such a performance hog after all.

PCMark Vantage – Productivity

This is the first and our favorite benchmark of the entire PCMark benchmark suite that we often use to test performance of software and hardware. The “Productivity” benchmark tests the performance of actual Windows applications, such as the responsiveness of multiple browser tabs, notepad pages and malware scans. It simultaneously performs dozens of these tests and calculates how fast your machine is when conducting these typical tasks.

The Productivity suite is the one and only benchmark so far that has shown me a significant increase in performance. An increase in 400 to 500 points is quite impressive—at least on paper. Windows Search apparently affects only the multi-tasking user who starts and closes many applications many times.

PCMark Vantage – Gaming

The “Gaming” test from PCMark Vantage heavily relies on the graphic cards. However, processor, RAM, and memory also play a big role. And since Windows Search is supposed to be such a memory-eater, it should, in theory, slow things down a bit.

Interestingly enough, hard disk accesses were just a bit faster on the slower notebook after I disabled Windows Search. Due to the fact that the faster notebook is equipped with an SSD, the performance difference isn’t noticeable. But overall performance in terms of FPS did not increase. I suspect that if gamers disable Windows Search, they might notice about a 10–20% decrease in loading times between maps and levels, but the game play won’t suffer from having Windows Search running in the background.

PCMark Vantage – Memory

This is another test for memory speed. The only problem here is, I couldn’t get it to run on the slower notebook—it crashed in every run. However, the results on the faster notebook speak for themselves.

Nothing—again, I ran each test three times and calculated the average. It always varied by a couple of points, but there is no real performance difference with Windows Search enabled or disabled.

Conclusions about Windows Search

I’ve arrived at one conclusion—Windows Search actually has an effect on performance. Yes, startup time and shutdown times did decrease a bit, and I also noticed a difference when running PCMark Vantage’s Productivity test, but this impact just isn’t noticeable when it comes to gaming or application performance.

I personally feel that responsiveness did increase a bit on the slower notebook; however, this could just be wishful thinking. It is probably just the placebo effect. I suggest to all Windows 7 users the following:

• Windows Search should always be enabled, if… you need to find files, emails, programs, and contacts, among other things, once in a while. Do yourself a favor and don’t disable Windows Search. If you rely heavily on faster searches, there’s no question—don’t touch it!

• Windows Search should be disabled, if… you just use your machine for one purpose only, such as for gaming or as a Windows Media Center-based PC, if you never search for files, or if you just want to squeeze the very last ounce of performance out of your PC.

So, is it a myth?

No, it is definitely not a myth. Windows Search may slow down things a bit, especially on lower-end machines—although that’s barely noticeable in some areas. The myth probably just continued onto Windows 7, as it originated with Windows Vista, where search performance used to be quite an issue.