True reason ™: it comes with a countdown, to make sure you have to upgrade.
Search the web for it!
The truth is out there!!
(Actually, cynic as I am, I have the feeling that there might be a tiny sliver of truth to this - why make things last forever lest users pay but once?)
I really wish there was a more factual answer to this instead of educated guesses. I’m sure someone in the world at some point has done a formal study on this. I notice it too, but much more so in the pre-Windows Vista days than now.
There are factual answers. but there is not a single factor involved and the mix of factors won’t be the same for each case.
Uninstalled or updated applications leave libraries and services in place and Windows continues to load them, consuming resources.
Startup entries for missing files (again, left behind by uninstalled applications) still try to load, then time out -primarily affecting boot or logon delay.
The registry gets bigger and messier, making it slower for applicatioms and the OS to access.
(The reason these things don’t clean themselves up completely on removal is not just bad uninstaller configuration, it can also be caused by ambiguity over which application still needs that component, or by the resource being locked when the uninstaller tries to remove it)
Malware and adware consumes system resources and even after cleanup, can still leave bits behind.
Double-hidden (show hidden does not reveal all nonpresent devices) device manager entries for all of the devices you have ever plugged in, slow down the handling of devices in general.
And so on. If you’re looking for a single, simple answer, there isn’t one.
Um, people don’t seem to be guessing all that much in this thread. Almost everything has been quite factual.
Except post #7. That one was a bit guessy. 
Sorry, I’m just looking for proof of what things actually slow down a system vs. what we logically think would slow down a system.
Maybe someone has a link to an experiment where someone times say the opening up of Photoshop, writes some software that installs thousands and thousands of entries into the registry in various branches, adds a set amount of data to the registry (maybe several hundred MBs), and then times the opening of photoshop again to see if there is any difference. They then could remove the entries and data and repeat the whole thing a few times.
This way there would be actual documented proof of what slows down a system.
The registry is a pretty complicated piece of thing. It is continually changing, never idle.
Also, imagine a malware infection, that only modifies a few lines of code, yet you get a system that’s trashed. Only have to change a few low level OS component settings. What are you gonna do? Reinstall windows to a default setting.
It’s perfectly possible to do laboratory tests like that, but that doesnt mean people are just making shit up in the absence of precise metrics. If you know how a computer does a certain thing, you can very reasonably figure out what kind of changes to that thing will impact performance.
IT professionals generally don’t need the precise quantification of these factors, just the knowledge that they are factors.
For example, the nonpresent devices you can’t see in device manager… Windows checks each one to see if it’s there, and this affects boot time. How much does it affect boot time? Can’t say. It depends what they are and how many, but if your computer is booting slowly, thats one of the factors I would check.
To answer the question, “Why does a Windows install get slower with age?” I think it is possible to give an exact answer, and I would be surprised if someone hasn’t studied and published it somewhere. These things should be known in detail by now, not guessed. Anyone that has worked with computers for a long period of time can tell you why they think Windows slows down over time, that it’s due to a number of factors that sort of all snowball together to cause a noticeable difference in speed, but they don’t have factual answers, just logical guesses that make a lot of sense. The dialogue has been the same since at least the early Windows 95 days, and I just don’t understand why it hasn’t been answered more completely by now.
The reason I’m being admittedly stubborn about this is that I had an XP machine at my house that only had the OS, MS Office, Visual Studio, Chrome and some open source MP3 tag editor installed on it, ever. I used it for about a year and noticed a slow down and then reinstalled Windows XP with the few programs mentioned reinstalled and it ran noticeably faster. Why? I don’t know… I figured it was due to Windows Update’s updates, but when I re-installed I assume the same updates were applied just in a quicker fashion, so who knows. So in the spirit of fighting ignorance I wish there was a verifiable, factual answer to explain why.
You’re wrong. We actually know many of the factors that slow Windows down, I listed some of them above - these are not guesses, they are facts.
Nobody (as far as I know) is interested in measuring precise amounts by which any single factor (e.g. what is the impact in microseconds wasted for a specific broken DLL) will slow down a machine because there are too many variables in the mix - it becomes a gestalt of the individual issue, the environment, the context of other prevailing problems on the machine, etc.
It’s not impossible to measure that - but it would be an enormous waste of time. I could spend a couple of days measuring the exact impact of, say, a problematic driver on a machine, but that impact may vary, depending on factors including, but not limited to:
[ul]
[li]The exact hardware specification of the machine[/li][li]The OS version (including patching) and detailed configuration[/li][li]The driver version[/li][li]The firmware version of any devices[/li][li]The identity, version, health, permissions configuration etc of any installed software trying to use the driver[/li][li]The order in which services and drivers are loaded at boot[/li][li]The load being placed upon the machine either interactively by the user or by other installed software doing stuff in the background[/li][li]Any and all prevailing inherent issues such as memory faults.[/li][li]Any and all other of the multitude of other issues we could be precisely measuring instead.[/li][li]etc…[/li][/ul]
At the end of this exhaustive (and exhausting) analysis, I would have the precise measurement you are asking for, and it would be useless everywhere outside of the above context.
That’s why in IT, we don’t typically waste our time measuring the specifics when we know the broad strokes: your machine keeps freezing / this started last Wednesday / that’s when the system updated the graphics card driver / let’s try rolling it back / OK now? / who’s next?
If there was a single verifiable, factual answer to explain why, Microsoft would have published a fix for it.
My Win98 machine does actually start verifiably factually slower: a MS update to handle newer computers introduced longer timeouts.
My XP machine does actually start verifiably factually slower: There is a dropbox timout at startup.
Similar kind of stuff on many computers I’ve used.
One notable exception to what I said above; companies building a production run of computers (especially for corporate sale) may spend time lab testing their build to ensure they have the best performing drivers, there are no conflicts, etc. I know HP does this, but once those machines get out into the wild and are exposed to different collections of users, environments, peripherals, software, updates, malware and usage, pretty soon, no two will be the same.
I’m a software developer and, trust me, almost no one is writing software designed to fail after a while (there are a few notable exceptions, generally related to DRM or other legal restrictions on software)
The reason that software sucks is that we, collectively, are not very good at it yet. We’re still in the very early stages of understanding it, and it’s fantastically complex. If we were engineers, we’d be at the point in understanding of our craft where buildings fall over all the time.
No one looks at a collapsed bridge and thinks “Ah, the engineers probably designed it to fall over after a few years. After all, that way they get to design and build another bridge.”