Many Windows users have noticed that their computer feels fast and more responsive when it‘s brand new, yet after some time it seems to get more sluggish. While this could be some kind of placebo but the effect could also be a real phenomenon. If it is real, one possible cause for the gradual slowing down of Windows computers could be the slow buildup of left-over data.
While this topic has been interesting for me for obvious reasons, I never had the time to really look into this in the form of systematic benchmark tests to answer the question: Can temporary files and registry junk slow down a Windows 10 computer?
But let‘s take a step back.
Like many stories, this one also starts on Reddit.
If you don‘t know, Reddit can be a weird place. It‘s the Internet‘s truck stop where you can find almost anything. You can find a lot of helpful people and discussions, but you will also find a lot of toxic things.
In either case, one day I was on Reddit browsing for interesting discussions when I stumbled upon a thread at /r/pcmasterrace/ where someone was asking for an alternative for CCleaner. The thread is here, in the case you want to see it:
The discussion quickly escalated into an interesting debate whether temporary files and unneeded registry data can even slow down a modern Windows 10 based computer.
A Redditor claimed that no amount of temporary files or registry junk is going to ever slow down your computer:
His argument was that since left-over, temporary files by definition are left-overs, they are not being used by the operating system and therefore they will not have any effect on the computer‘s performance. No use, no harm, right?
That was an interesting argument and it does make sense. Especially in the context of modern systems with SSD drives, where the drive read speeds remain high even if the drive contains a lot of unused files. With older style mechanical drives, one could make the argument of too many temporary files possibly leading into data fragmentation and therefore loss of performance.
However, even with the argument making sense, I had a strong gut feeling that he was wrong.
I thought, surely, a large number of temp files and registry junk is going to slow down a computer.
But it was, after all, only a gut feeling. As an engineer, I don‘t really work by gut feelings or opinions. I trust data. And that is why I designed this benchmark.
The first benchmark: Does junk slow down your computer?
It‘s important that we answer exactly what we are trying to measure here. I designed a benchmark to answer two simple yet very specific questions:
- Can temporary files and registry junk slow down a modern Windows 10 PC? And,
- If they can, is the effect large enough to be noticed by the user?
My original hypothesis, which I also shared in Reddit, was that there needs to be a lot of junk data for it to have a clearly measurable or noticeable effect. I was thinking possibly million or so temporary files and perhaps even more registry data.
While on the surface the original argument makes sense – that unused temporary file and registry junk is, by definition, unused and therefore no amount of such junk would slow the system down.
However, it‘s not really true at all.
Even if some data is unused, and for example if Windows doesn‘t use the specific temporary files per se, Windows is still going to be maintaining and constantly updating the file system, for example. And while doing that, it is going to iterate through every single directory and every file, even if it doesn‘t technically use such temporary files.
The same applies to the Windows registry. When your computer starts up, Windows is going to load up the entire registry, and iterate through all the registry data. A common misconception is that the registry is only a few megabytes and therefore, simply loading up those few megabytes from the disk into RAM is going to take only a very short time. But this is not the entire picture. When Windows starts up, it doesn‘t just load some data from the disk to RAM.
The Windows registry is used to store system and third party application data and settings. Windows is going to iterate through all that data at some point during the system startup with some data being read multiple times. While this is optimized and possibly multi-threaded on Windows 10 and 11, it is still going to use CPU time and it shouldn‘t be trivialized into being simply a matter of loading a few megabytes of data from disk to RAM.
That is why I thought, surely a large amount of junk data is going to slow your computer down.
Therefore, I designed a benchmark: A freshly installed, activated copy of Windows 10, fully updated and with only Firefox installed to the system.
As it is obvious from these parameters, the goal was not to simulate any kind of real world scenario, or a computer that an average user might have. Instead, the goal was to eliminate variables. The less software there is in the system, the more likely we are able to see the effects of temporary files and registry junk data on Windows itself.
Now, it‘s not trivial to define what “slow your computer down” exactly means. One could be referring to the overall performance, such as the computer feeling sluggish, or some kind of system benchmark results.
Or, as I think most of us could agree, a very easy and usable metric for a computer‘s overall speed is how fast it turns on and allows the user to start to use the computer. Some theoretical benchmark result is one thing. How fast your computer starts up and is fully loaded and ready to use is another thing entirely. And that is what I focused on.
I used a stopwatch to measure three data points before and after the computer was filled with junk:
- Time to desktop. How many seconds it takes for the desktop to load up to the point that the wallpaper is visible on screen.
- Time to idle. How many seconds it takes for the computer to fully start up, in a way that the entire system is loaded? I define this by using the Windows Task Manager to see when the overall CPU usage drops to under 5%.
- Time to open a website. How many seconds does it take to start the computer, wait for the CPU usage to drop under 5% and then launch Firefox to open Reddit.com‘s front page and wait for the page to fully load and render it.
I would argue these three metrics are a fair way to establish how fast a computer is in everyday use.
Also, it‘s worth noting that I‘m now referring to a full system restart time. That is, without the use of anything like hibernation. With Windows 10 hibernation or Fast Startup related features, it‘s possible for the system to appear to be starting in a matter of a few seconds.
Baseline – before temporary files and registry junk added
In the first test, I established the baseline. How fast does the computer run when it is clean and without any junk?
Here are the results:
Time to desktop: 23 seconds.
Time to idle: 41 seconds.
Time to open a website: 57 seconds.
Windows 10 is very optimized to start up quickly. This is done by loading only the most important modules first and loading everything else on the background after the desktop has been shown to the user. This is why there is a noticeable delay from the “time to desktop” to “time to idle”.
Benchmark – the same computer filled with junk data
To perform this benchmark, I used two scripts.
Firstly, I wrote a Windows batch file that creates 30 000 temporary files of three different types to three different temporary data folders in the system. That is a total of 270 000 files.
Since we don‘t want this test to be related to hard drive usage, each of the test files is small, about 8 bytes or less per file and therefore reading data from disk is not going to be the bottleneck.
Then, I created a registry import file that adds about 120 000 registry keys and entries to the Windows registry under a few different locations.
I‘m offering both of these files for download so anyone can reproduce these benchmarks.
After the same Windows 10 system was filled with these temporary files and registry junk, I took the exact same measurements of the computer‘s speed as before.
The results were:
Time to desktop: 62 seconds.
Time to idle: 102 seconds.
Time to open a website: 118 seconds.
As we can see, with the temporary and junk data the system took over twice the time to start!
Not only was the computer clearly slower, there were actual problems with the system: the Windows Task Manager was glitching with the Performance tab seemingly not working at all and the main process listing tab looking very unusual.
The system also felt sluggish. This feeling of being sluggish is, of course, impossible for me to objectively measure, but the system did feel overall slower. For example when simply trying to use the Windows Task Manager or launching Firefox.
I wanted to see how much temporary files and registry junk data slows down a modern, Windows 10 system. I was expecting to see some measurable effect but I was surprised how massive the effect was.
Having a lot of temporary files and registry junk clearly slows down your computer.
It‘s worth noting that this benchmark is not about how much your typical computer system gets slowed down over time with the accumulation of junk data.
The point of this benchmark test was to address the original claim that no amount of junk files or registry junk data can slow down a modern Windows 10 based computer. And that, as my gut feeling told me, is absolutely incorrect. As we can clearly see from this test: too much junk data slows down your computer dramatically.
To show that I did this benchmark the way I described it here and that the results are as I say, I performed the entire test while recording my screen. The video of this test being performed is published in my company’s Youtube channel:
I‘m also releasing the two script test files which I used to generate the test data. Feel free to use them to reproduce the test. You can download the script files from here: https://jv16powertools.com/downloads/special/create_test_junk_data_v1.zip
The test was performed with VirtualBox 6.1 running on Linux Fedora 35 workstation host. The system is running on Intel Core i9-10900T CPU, 64 GB RAM, Nvidia Quadro P620 and Samsung SSD 980 Pro 2 TB NVMe hard drive. The virtual machine was dedicated with 2 CPU cores and 4 GB of RAM as recommended by VirtualBox.
Follow-up benchmark: Same idea, but better testing
After the original benchmark test results were published, we got a lot of feedback from the community. The feedback can be divided into three different main points.
Firstly, it was suggested that this benchmark should have been done on a real computer, instead of using a virtual machine. This is a good point and I fully agree with that.
The reason why the original test was done using a virtual machine was simply because it is much easier to do. For example, setting up a screen capture to video in a way it works even if the computer is restarted during the test is trivial using a virtual machine but not so trivial with a real computer.
Secondly, some users asked whether specific Windows 10 features, namely the File Indexing or the Windows Security Antivirus could be affecting the results.
If you suddenly create a large number of temporary files, it could trigger either the Windows File Indexing service to start to index those files and/or the Windows Security Antivirus to scan those files and either of these could cause the computer to slow down as seen in the original test results.
While I don‘t know whether these are valid points, they seemed like something that could have an effect. Therefore, I wanted to repeat the test with the File Indexing disabled and after running the Windows Security Antivirus simply to eliminate the possibility that either of these could be the root cause of the test system slowing down in the original benchmark.
And thirdly, there were concerns whether the testing was appropriate, because I am not an independent third party.
While I’m not an independent third party because I’m the developer of jv16 PowerTools Windows utility suite and one of its many features is the cleaning of temporary files and registry junk data, these benchmark tests are not in any way relating to my product. My product is not used in these tests nor even mentioned, other than this mention in order to be transparent about this connection.
At the end of the day, everyone can have biases or conflicting incentives. A person doing a hardware review might have sponsorships from companies who are competing against the reviewed product or a person doing software reviews might be part of affiliate programs of that or competing software products. These are the kind of potential conflicts that exist everywhere in the industry.
These benchmarks are not related to subjective opinions but objective things we can measure and these benchmarks are done with the highest amount of transparency I can think of. That is: all the technical details, test data generation scripts as well as screen capture videos of the testing are released which means anyone can repeat the testing to confirm whether the results are in line with these results.
This is, as far as I know, a much higher level of transparency than most other parties doing any type of benchmark tests would normally do.
That being said, let‘s start with the follow-up benchmark.
A follow-up benchmark: File Indexing and Defender Antivirus disabled
To test whether Windows File Indexing or Windows Defender Antivirus could cause the effect of Windows slowing down with many temporary files, I prepared a new virtual machine by first disabling the File Indexing.
Then, I did the baseline benchmark. The benchmark results of freshly installed Windows 10 in VirtualBox, with all the updates, only Firefox installed and with File Indexing disabled are as follows:
Time to desktop: 20 seconds,
Time to idle: 35 seconds,
Time to open a website: 49 seconds.
After this, I ran the test scripts to generate the temporary files and registry junk data.
Then, I ran the Windows Defender Antivirus in its full scan option.
Time to desktop: 59 seconds.
Time to idle: 87 seconds.
Time to open a website: 107 seconds.
It is also worth noting that during the testing, Windows Task Manager was again glitching and not fully working.
Similar behaviour was also seen in the original benchmark test. Therefore, we can also conclude that a large number of temporary files and registry junk can make a Windows 10 computer slower, they can also cause actual system problems.
This test was performed with VirtualBox 6.1 running on Linux Fedora 35 workstation host. The system is running on Intel Core i9-10900T CPU, 64 GB RAM, Nvidia Quadro P620 and Samsung SSD 980 Pro 2 TB NVMe hard drive. The virtual machine was dedicated with 2 CPU cores and 4 GB of RAM as recommended by VirtualBox.
The final benchmark: Does junk data slow down Windows 10 in a real computer?
The main concern about the original test was whether the slowing down was due to a large amount of temporary files and registry junk, or did it happen because the testing was performed on a virtual machine.
That is why in the final benchmark, I performed the same testing using a real computer.
I used an office desktop PC with Windows 10 Pro fully updated, running on AMD Ryzen 7 5700G, 16 GB of RAM with 500 GB Western Digital NVMe SSD system drive.
In its default state, that is, before adding any temporary files or registry junk data, the system performed a full restart as follows:
Time to desktop: 28 seconds.
Time to idle: 55 seconds.
Time to open a website: 68 seconds.
It is worth noting that these numbers refer to an actual, full system reboot and not any type of hibernation.
To eliminate the possible effect of the Windows services, I disabled Windows Search Indexing before running the test scripts to generate the temporary files and registry junk data. After the data was generated, I ran the full scan of Windows Defender Antivirus.
With the system full of temporary files and registry junk data, I performed a full system restart and the results were in line with the previous benchmarks performed in virtual machines:
Time to desktop: 42 seconds.
Time to idle: 74 seconds.
Time to open a website: 97 seconds.
The following chart contains both the before and after results:
These numbers show that a large number of temporary files and registry junk data can slow down a modern, Windows 10 system and this effect is not related to Windows File Indexing or Windows Defender.
Furthermore, this shows that the results of the original benchmark were not any artificial effect caused by the fact the test was performed in a virtual machine.
As with the original benchmark, the full testing sequence was performed while screen recording it to video on the virtual machine. A video camera was used to record the real computer testing process. These videos were created to show that the testing was performed the way it is described here, and that the results are as mentioned. You can see the video here:
It was claimed that no amount of temporary files or registry junk data can slow down a Windows 10 computer. I used two virtual machines and one real computer to test this claim.
Based on these test results, it seems evident that having a large number of temporary files and/or registry junk data can, in fact, have a major performance impact on a Windows 10 computer. A large amount of junk data made the test systems start up to 100% slower than without junk data.
It‘s also worth noting that on both virtual machines, the temporary data or registry junk data also caused the Windows Task Manager to start to glitch, which also implies that this type of data can also cause Windows system errors.
This test was designed to determine whether it is possible for temporary files and registry junk to slow down a modern Windows 10 system.
The purpose of this specific test was not to determine whether this is something that happens in the real world, nor was the purpose to see exactly what type of data nor what amount of data can cause this. Furthermore, both temporary files and registry data were tested at the same time to reduce the amount of work needed in the testing.
Disclaimer: I‘m the lead developer of jv16 PowerTools, Windows utility software whose one feature is cleaning Windows computers, including cleaning unused temporary files as well as registry junk. My username in Reddit is /u/JouniFlemming/ and we also have an official subreddit at /r/jv16powertools/