Vista vs XP performance: Some informal tests

After posting about the inadequacy of a recent test report I thought it would be interesting to conduct my own informal tests of Vista vs XP performance. I do not run a computer laboratory, but I guess my tests have the benefit of being real-world. I tested several conditions on three computers. On two of them I was able to …

[Steven Hewittt]

Interesting

Good article - really impressed at how neutral it was! :-D

Startup aside (as MS expect people to use the power putton on the start menu, which doesn't shutdown but puts the machine to sleep instead - which is faster), the article essentially says that Vista is only slower (in this test) when rendering GDI graphics rather than the system as a whole...

Very interesting.

Not sure how the testing works regarding RAM. If it's a simple case of how much is loaded into RAM then Vista would look worse - but that's meant to be. If Vista see's unused RAM it will try to use it as it knows RAM is faster than HD's. As such much more of the OS is pre-loaded into RAM initally than on XP for performance gains. When an application requires more RAM Vista drops out parts of the OS from RAM to grant the app the required amount of memory it needs.

Good article - thanks!

[Anonymous Coward]

GDI v Direct X

"Second, an application written to use DirectX rather than GDI should perform better, other things being equal."

Well that's hardly a surprise - make GDI applications run like crap and people will abandon it and use DirectX instead. Typical Microsoft behaviour - kill the opposition no matter which system is better/more flexable

[Anonymous Coward]

Agreed

without any benchmark I thought XP was slower than 2000 and I simply removed the pre-installed Vista on a Dual Core 2GHz and replaced with 2003 ... lot faster to boot & to use.

[Anonymous Coward]

Quantify the hardware please

You mentioned a few of the tests had very little difference between XP and Vista. Were all the "low difference" results on the same machine, or was some hardware better than others at certain tests only, no silver bullet?

Either way I'd like to know which one of your computers you would use for Vista and its current hardware configuration.

[Anonymous Coward]

Now on to File Performance

Please continue on and test file performance....specifically the huge drop off in performance when Vista's shell functions are used (ie. Copying folders, extracting archives, etc). I know it has the new xfs features, but the loss in performance doesn't outweigh the benefits....and as far as I could find, you couldn't disable xfs.

[Anonymous Coward]

Now on to File Performance

oh dear no! the pain of trying to run an exe from a .zip on Vista is just unbearable! why quantify what is so obviously a massive drop in speed.

[M J Bickley]

Waste of time

It doesn't matter - Vista is the next OS. Seen it, done it:

DOS->Windows

1->2

2->3

3.x->95

95 ->98

98-> 2000/Mil

2000 -> Xp

XP->Vista

Just use the technology - MS doesn't care what you think. If you want a really fast machine, run Windows 98.

[Solomon Grundy]

Performance

This was a great article for The Register - considering their usual slanted view on things (no offense but that's the truth). Props to the author for straight fact reporting!!

PC performance is a funny thing though. While on some "benchmarks" Vista may be slower than XP, we recently deployed 32 Vista Business (pre-installed) PC's in one of our satellite offices and the resulting gains in productivity/reduced support calls has been quite impressive. Vista is easier to use for non-geeks and even with some specialized compatibility programs, is a good investment.

I feel a lot of people commenting on Vista performance really haven't used it in a business environment. Try a few and see what you find. I bet you'll be impressed.

[Anonymous Coward]

hmm

Am I missing something? Where are the rest of the test results? Only 2d graphics are mentioned.

Anyway seen as you didn't do any fixing of the XP system (anyone worth half their weight in salt kills all unnecessary services on that pile of dung too, removes the fisherprice and tweaks a few reg entries to make it run better) I don't see much relevance.

How does a fixed XP system run against a fixed Vista system on a range of tests on identical hardware though?

Interesting none the less.

[Ben]

plain english

Thanks , interesting in a windowy kind of way and clear enough

to understand even for me .

Where are all the posts from the IT angle jockeys ?

"OOOOOY! OVER HERE , THERE'S SOME OF THAT IT STUFF"

Probably the complete lack of biased and exaggerated claims ,

nothing to see , move along , why its so quiet i can hear the faint

squeak of the crash trolley going from Annon to Annon over in

the ElReg customer survey thread...............kinda eerie.

Good article .

Ticks box for "things that DONT need f*cking with one bit"

oh , wrong thread................llocks

[Michael]

Poor statistics...

"On both dual-boot machines, the best Vista performance was an amazing 70 per cent slower than XP. Put another way, XP was three times faster."

If to be slower is to take more time to complete a task, then this sentence is saying that it takes Vista 70 per cent more time to complete a task, which doesn't jive with XP being three times faster.

If Vista takes 100 seconds, then XP takes 30, which makes XP 70 per cent faster (takes 70 per cent less time) than Vista, while Vista is 300 per cent slower (takes 300 per cent more time) than XP.

This is an all too common mistake. When dealing with percentages, "A is 70 per cent faster than B" does NOT translate to "B is 70 per cent slower than A"

[Jim]

@Poor statistics...

I don't get your example.

Vista=100s, XP=30s

Let's say that Vista is 90s instead (for convenience). That means that XP is 300% (3 times) faster, ie does the job in 1/3 the time. It also means that, conversly, Vista is 300% (3 times) slower. You see it IS true that if A is x% faster than B then B is x% slower than A.

To explain, it is necessary to examine the logic of these two statements and understand that the percentage is derived from the ratio of the values.

When you switch the order of comparison OR switch the comparison type (faster/slower) then you invert the ratio. With these two statements you switch both these components so that you invert the ratio twice (a/b -> b/a -> a/b) so you end up with the same answer.

The fact that the difference is 70% of the higher value has no real relevence to the calculation.

I do agree that the original statement is rather confused. If Vista was 70% slower, that is it took 170% of the time XP took (30s), then it would take 51s not 100s. 51s is clearly not 3 times 30s.

[Anonymous Coward]

@Michael

Back to school for you: speed is usually measured in number of (somethings) per unit time. E.g. number of clock cycles per second; number of miles travelled per hour, ...

So if A is 70% slower than B then A is doing 70% fewer basic operations per unit time, which does indeed mean that B is (more than) three times as fast as A.

[E]

GDI

I recall many moons ago that hardware accelerated GDI calls were the cat's pajamas. MS & the video card vendors made noise over it.

Video drivers have been made simpler in Vista by slowing down GDI... years after they were made more complex to speed up GDI.

I think this is 100% marketing bull-crap - an attempt to make a silk purse out of a sow's ear. MS has a history of labeling bugs to be features. I cannot really see this as a feature because AFAIK if I make a C/C++ call to a Windows function that opens a window, then I am using the GDI, not DirectX. I could be wrong - it's been several years since I programmed Windows windows.

What next? MS will tell us that it had to disable OpenGL capability in NV & ATI drivers because that makes MS' driver model simpler? Wait and see.

Another possibility is MS replacing the API in order to keep the larger development world confused and focussed on Windows. If John & Jane Programmer have to worry that suddenly they'll need to understand DirectX inside out instead of the good old GDI API, then they're not going to spend much time looking at MocOS or Linux or UNIX.

Several times in the past MS renamed/rebranded/changed the API for OLE/COM/DCOM/ActiveX, for no better apparent reason than that it made for good marketting (ActiveX better than OLE! [but really it just sounded better]).

[Anonymous Coward]

Just for fun...

I'd like to see Windows 2000 compared against XP and Vista. With every Windows release comes more bloat and less attention to the basics, like filesystem performance for example. And please do a retest on a machine with abundant memory. While I don't think anyone is yearning to go back to Win2K, I support all manner of Win2K and XP clients, and without having performed any benchmarks, I can say that the Win 2000 boxes just 'feel' faster. In our Active Directory environment, the Win 2000 machines routinely login faster, file transfers seem faster, bringing up file permissions and changing them is faster, etc. And what a joy waiting for XP to 'build the domain list' Fortunately we're sane enough to have not deployed Vista---I can only imagine the bottlenecks.

[J]

Cool, but...

Interesting. As the author already was at pains to note, far from perfect. But, as someone already observed above, hardly a fair comparison. Vista was optimised for speed in every possible little obscure way (because the regular install was being massacred in other tests, if I understood correctly). XP was left alone. And XP was STILL faster in general.

What would the numbers look like if XP was also optimised for speed in the manner Vista was? That is, assuming it can be done as well, besides turning off unused services. Logic would suggest it would be even more of a difference in XP's favour. But computer issues don't necessarily care about logic... :-)

[Anonymous Coward]

@Just for fun

I had Windows XP Professional on a Compaq Presario 930 laptop. It was starting to feel slow.

I recently got rid of XP and dropped 2000 onto it. Instant speed boost. It now not only starts up faster, but feels more snappier too.

[Nano nano]

Vista Networking is broken

A new laptop with Vista is unable to connect to a bog-standard home network with existing XP and W2k PCs.

Networking with Windows until Vista worked pretty much out of the box, now it is implied that in order for Vista to work, our "outdated" (="obsolete"!!) ADSL gateway is at fault.

So what part of the Internet Protocol does Vista not now understand ?? And why does it arrogantly assume it has no responsibility to work with users' kit ?

[Tim Anderson]

Quantify the hardware

I put the details of the machines used in the comment to the post here:

http://www.itwriting.com/blog/?p=430

> Either way I'd like to know which one of your computers

> you would use for Vista and its current hardware configuration.

The lowest spec machine spends most of its time in XP or Ubuntu. The best spec machine runs great with Vista.

Tim

[Tim Anderson]

Optimising XP

> Vista was optimised for speed in every possible little

> obscure way (because the regular install was being

> massacred in other tests, if I understood correctly).

> XP was left alone. And XP was STILL faster in general.

That's correct, though I ran the test with non-optimized Vista as well. The reason was twofold:

1. I was sceptical about the claim that XP is 2x faster.

2. Given that Vista is slower, I'm interested to know where it is slower, and what is most effective at improving its performance.

I agree that more tests would be interesting.

Tim

[Steven Hewittt]

@ J

True, but also let's remember we are comparing a 6 year old OS on brand new hardware to a brand new OS on brand new hardware.

Run Vista under MS's recommended minimum specifications and the same with XP would be a much more fair...

[Kyle Miller]

Our Own tests

As a developer of a highly graphical 2D CAD application, we have performed a number of tests and found that Vista is about 20 times slower for some operations (drawing rectangles for example).

For us it is a real headache because some of the code is 25 years old and we can't just change over to DirectX, partly because of the effort involved and partly because we have to support different platforms (yes there are other platforms!).

If you want more details, post a comment here and I will show more numbers.

[Billy Verreynne]

@Kyle

Understand your pain Kyle, but it really does not make any sense to muck about with the GDI even for "just a" 2D graphic type applications. Been there. Done that. Still have the burn marks (and the code!) :-)

OpenGL. It should be not that difficult to switch. With immense benefits not only in performance, but also in portability to other platforms.

[Louis Cowan]

reactions

DOS->Windows "OOOOoohh"

1->2 "oOOOOOooh"

2->3 "oOOOOOOOOOooo"

3.x->95 "Woooooow this is awesome"

95 ->98 "Yay"

98-> 2000/Mil "Excellent!"

2000 -> Xp "err.. ok.. it's a bit San Fran though, innit? I'll try anythin once though"

XP->Vista "Nooooo!"

[Robert Long]

@Steven Hewittt

In what way would that be fairer? If I buy a machine tomorrow it will be new hardware and I'll still have the choice of Vista or XP. If Vista sucks then it sucks; I'm not going to give it a by just because XP would have been slower 6 years ago on some machine I'm not buying!

[Allan Rutland]

Memory usage

Been using Vista (and yes flame away all you like, but I've had no issues other than drivers) for over 18 months now, and have found a few intriguing things.

Personal favourite has to be booting the machine, and dropping it to sleep again. Haven't yet figured out what’s not restarting (everything runs correctly, all services are still there etc), but you will regain between 20-30% free RAM back. And a bit of a performance improvement.

And Readyboost is also quiet interesting, depending on machine seen next to no change, up to 15% speed boost from it. Although this does confuse some users having a USB drive stuck in the back of the machine permanently.

Guess a lot does depend on what’s being run, how lazy the dev's were who made it when they just kept hacking things together and using code going back 10-20 years or so. Things progress, and really performance can't be really regained until backwards compatibility is dropped (yes I know the jokes here about Vista's compatibility).

As others have said above, this was the same with going from 2000 to XP, it was slower. And the jump from 98 to 2000, or from 95 to 98, it's always been so. Every time I upgraded my Mac (which makes a nice doorstop) yes, it got slower also. It's always happened, and always will. Adding more to something makes things bigger as odd as that may seem to many.

Great artical though. Anyhow, flame away :P

[alistair millington]

@Solomon

I / we use vista at work (I don't risk it at home so dual boot) I haven't noticed any improvement in the use, merely a lack of support for our third party drivers that happen to be old and therefore disabled. (pc anywhere) meaning we can' supply certain clients who can't afford or don't want to upgrade everything.

An increase in calls from people who say "where is the..." Basic level stuff for staff who used to ask, "How do I create a shortcut." (Yes not all people are IT literate)

Slower turn arounds for calls because of the time taken to load up basic app's, like our call logging software. (15 minutes from power on to be able to answer a call.)

Not to mention delays in connecting through RDP to sites that use windows server 2000 / 2003. Where it seems to complain because they aren't using Vista. Anyone else get that really anoying message that says "The computer you are trying to connect to is not running Vista." Durrrhhh. Makes yo uwant to shout at MS. It's a f*cking server dumb ass. Why would they have vista running their servers?

Using the little doodad on the gadget bar in Vista that shows mem usage, 80% of a 2GB RAM in use when doing nothing but looking at the desktop. Most of our machines have 768 MB RAM.

We support large financial institutions that still run win 2000. Vista doesn't like it.

Nice report though Mr journo', kind of clearly stated what we all knew.

[James Pickett]

Speed

"If you want a really fast machine, run Windows 98."

I do, and it is.

[Michael J Smith]

Going to Gaol?

Hi Tim

Don't the Vista and XP licenses have draconian restrictions on writing reviews about the software? Mr Balmer will be coming for your hide. :-)

Michael

[Svein Skogen]

@Solomon Grundy, and a few others.

I can acknowledge your findings. I have a handful of non-it folks running windows because they need to run some windows-only software. There is notably less support calls from those who've been upgraded to Vista, and given a crash-course in using the help system on Vista. Too bad this will be ruined, because some smart-ass search-giant wants to replcae the search system with one that can profile the user for directed ads (run a Vista machine on the same network as a network sniffer, after you've answered "no" to Microsoft's question about sharing usage data, and you will see that Microsoft actually does as it promises. Only thing spooky in that traffic, is something that looks like an SHA hash of the windows serial number, combined with the serial of the HDD, mainboard, and cpu. I guess this is because I'm running an OEM version. This traffic pattern is the same for Home Premium, Business, and Ultimate.) The data sharing of Windows, even with usage-data sharing _ON_ is far better than the load of traffic generated by google "desktop search". While Microsofts search-solution indexes and keywords your hdd content in a file on your computer, it seems that the comparable google product leaves the same index file on their servers aswell. Yet I don't see the "Windows is bad"-crowd yelling about googles data-collection. Maybe they don't care about the possibility that google's "no" to share officially with a TLA government organisation, may be a smokescreen over their non-official sharing? (The government org gave in far too easy, compared to what they've done to others)

Back to Vista:

I've found that Vista performance in day-to-day use is atleast comparable to XP, if not better, on the same hardware. Vistas performance on hardware XP alas lacks the driver for (adding a robson stick) is a LOT better on Vista. Period. Again I have to wonder how many of those being negative to Vista in the comments have actually USED Vista, and how many are regurgitating "facts" from "/."?

as for disk io being slowed in Vista, I doubt it, if you're running with the latest patches. However, I've found that some Antivirus programs will slow _ALL_ shell disk operations down to a snails pace in Vista. I'm guessing some of these AV manufacturers didn't quite want to make a real vista version of their product, but wanted a quick hack that would atleast install on vista. I've seen AV software steal as much as two thirds of the disk speed, without giving any indication in cpu load. In my book this pretty much says that the software is piping disk io through their process, which runs at "idle cycles" priority. I think Microsoft would do wise to release a patch that blocked these hacks from running again until the user had answered yes to a question stating something like "This software is NOT compatible with Windows Vista. Running this software will slow your machines disk operation by as much as two thirds. Do you really want to continue? Yes/No/Tell me more". Might be more fair towards Microsoft aswell, and let the end user make a more educated answer. This type of dialog should pop up on ALL software that uses system calls that isn't really compatible with (for instance) win32. (Yes, there are programs today that insists on using the old win/16 apis, and people are blaming vista for the fact that this software is slow on modern computers. Look above. To help you, search for the phrase "some of the code is 25 years old". What's the chances that code uses anything but the win/16, or even dos/16 system calls? I'm guessing that same user is just as hostile to a 64bit windows with 4 gigs of ram, compared to a 486 DX50 with 128MB ram, because their 25 year old code runs faster on the 486 using Windows 3.X? Maybe such users have so many speed issues on modern computers, Windows Vista is the least of them? (Notice, I said DX50, not DX2/50. Big difference.)

So far I've seen Vista as atleast comparable to the older windowses in performance, and FAR more end-user friendly. Of course, speedwise it's still windows (and FAR from FreeBSD), but it really works. Really. Those of you who comment on it without ever having tried it: Try to comment based on your own experiences, not something you've heard from some bloke down at the pub you can't remember the name of.

//Svein

[Steve Evans]

Reactions @Louis Cowan

I think you need a couple of minor edits...

DOS->Windows "OOOOoohh"

1->2 "oOOOOOooh"

2->3 "oOOOOOOOOOooo, you mean I can open more than

one window at a time!"

3->3.11 "Wow, networking, who fancies a game of hearts?"

3.11->95 "Woooooow this is awesome"

95 ->95 osr2 "Cool, now my USB works!"

95 ->98 "Yay"

98-> 2000/Mil

2000 - "Excellent!"

Mil (6 Months later) - "WTF, it's all borked!"

2000 -> Xp "err.. ok.. it's a bit San Fran though, innit? I'll try anythin once though"

XP (1 week later) - Turned off all the tarty bits, now looks like

and runs like 2000

even the find files now finds things following registry

tweaks

XP->Vista "Nooooo!"

[Rob]

RE: Memory usage

@Allan, your not the only one, my Vista install is running smoothly and performs better than my last remaining XP machine.

One definite improvement is my Vista hasn't fallen over or hung on applications for over 3 months now and the frozen apps that happened before hand always crashed without taking the rest of the system or other processes with it.

Have only experienced 1 BSOD and that was the AV's fault which was corrected with an update the following day.

I do have issues with things like the sidebar, it does and will eat RAM so mine's always switched off, hogs screen real estate as well, it's only an updated version of Active Desktop and that was a nightmare from start to finish. I tend to get my older apps and games to switch off Aero when they go full screen, massive speed improvement for those apps and Vista's ok at switching back to Desktop and Aero a number of times.

I certainly wouldn't touch anything like '98 or millenium just to get a speed increase by trading up on the amount of crashes and BSOD's. Although I still believe 2000 to be the most "robust" of the family, I like Vista has definitely come on leaps and bounds since it's release. I think everyone has been wise to hold off until SP1 though, fingers crossed it will see more improvements.

Not had any issues with Vista networking either, seems to work better than XP's did in terms of speed and reliability (I use it with a number of different WiFi's and routers, discovery and connection are quick enough).

[Chris Cheale]

Naaaah

98 takes forever to boot and then you get the BSODs and having to reboot the frigging thing everytime you make the tiniest software change.

For actual regular use, I find XP is quicker than 98... although 2k is probably quicker than XP but I don't have a 2k install for comparison.

[Shakje]

GDI

Basic Windows code is all internal, the GDI is only really there for doing things such as drawing stuff, and by drawing stuff I mean like what you would see if you started drawing stuff in Paint. It can be used for more complex stuff, but anyone who does is not doing their users any favour, or themselves. It is significantly easier to code the same effects in OpenGL or DirectX, and it will run significantly faster. Ever since the mid 90s when 3D cards started becoming standard, GDI was doomed. Slowing down GDI will be very unlikely to slow down any apps you run at all significantly (and shouldn't slow them down to the point you can notice them slowing). There will be a small slowdown.

@alistair millington

As stated COUNTLESS times before, Vista uses RAM as it is freed. You can no longer estimate how much RAM is being used, as Vista makes as much use out of RAM as possible. It will fill it up if it can, and then decache stuff when you need the RAM for other stuff.

When '95 came out everyone screamed about it for using up too many resources, then after a few months people started working out that it was actually not bad at all.

During XP's lifespan it got slated for being insecure, needing a reboot too often, BSODs.

Vista comes out, everyone screams because it needs a better PC than XP, even though it fixes the above problems with XP.

Also, my install works fine with my home and company networks.

[Bruno Girin]

@Nano

Same problem here. I have a home wireless network to which XP, OS-X and Ubuntu machines connect with no problems. A friend of mine came round with a Vista laptop the other day and it failed to connect. A handy patch cable came to the rescue.

[John Savard]

File Copy

The article noted that the main problem was that GDI is slower, so screen-drawing, at least for non-game applications, is slower. But that shouldn't affect how much time it takes to copy a file, so apparently there's something else slow as well.

[Anonymous Coward]

Stop complaining about RAM usage...

I get so enraged when I hear people moaning that "70% of my 2GB RAM was in use, and I was only running notepad!"

Empty RAM is a waste of money, and that's why Vista will purposely use all the RAM it can get. It's not all bloat, it puts stuff it thinks you will use into RAM, so when you open it, it's much quicker!

Run "top" on Mac OS X after its been running for a few days, and see how much memory is "free" - not much - it will be marked as "inactive" - the remains of applications you've closed still there, in case you open them again.

So if you love your empty RAM so much, unplug it and keep it on your desk.

[Nexox Enigma]

@Speed

"""If you want a really fast machine, run Windows 98."""

If I remember correctly Win98 and its siblings cannot use more than 512MB of ram, plus I know that they don't do multi processor/core at all.

There was a time that I swore Win98 was the best, but that all changed with Win2000 SP2 or SP3, it's been a while. I was dual booting for a while on my (then) incredible 1.333 GHz Athlon, and Win2000 was faster for everything, including games.

I'd say that if you want a fast OS today run Win2003 - I'd say that its the best desktop OS Microsoft has come out with in years. Until SP1, which broke all of my important programs (Quake III.) I eventually had to ditch it in favor of XP because XPSP2/2003SP1 was required for applications, but I didn't have the time or will to get 2003SP1 not to suck.

Its now just a matter of time before I get some hardware with enough power to virtualize... Then it's Slackware from here to eternity.

[Andy B]

Vista easier?

I'm surprised at the people saying their users prefer Vista to XP - have you talked to them? My users generally dislike it - they find the extra visual clutter offputting and confusing. Combine that with some serious performance problems, especially with some network and ZIP file operations, and the level of DRM, and it doesn't paint such a rosy picture. And yes, these comments are from personal experience on my own and users PCs.

(As an aside - I recently had reason to run a Vista and a 98 box side by side - and the extra visual clarity on the 98 box, especially when using Explorer, was quite enlightening. Vista is 'pretty' - but they seem to have forgotten that the point of the graphics is to allow quick and easy visual comprehension of the data presented. Compare the directory tree views in Vista and 98 some time and see if you agree.)

Andy

[Sheriff Lobo]

@Jim and his Poor Statistics

Jim,

Michael is in fact correct when it comes to percentages, and I will use as simple a scenario as possible to demonstrate.

A = 50s

B = 75s

A is 33% slower then B.

B is 50% faster then A

Percentages faster and slower do not come out to be the exact same percent. 50% of 50 is 25 added up is 75. 33% of 75 is 25 subtracted equals 50. As Michael stated many people have trouble when it comes to percentages because the natural way of looking at it is to assume if B is 50% faster then A then A should be 50% slower and that just isn't correct.

[david]

Graphics moved out of the kernel

The graphics subsystem has been moved out of the kernel. No doubt that makes some graphics stuff faster, but the reason it was in the kernel before was explicitly for speed. MS reasoned that for an average user, having the graphical interface die was the same as having the kernel die, so the extra performance hit (10 or 15 years ago) wasn't worth the extra security.

That's changed now. Now they think the extra performance hit is worth it for the extra security.

[Patrick R]

@Sheriff Lobo and his confusion.

Michael is in fact wrong ... and you prove it."Vista is 70% slower than XP" does NOT mean "it takes Vista 70% more time to complete a task", it means "Vista's speed is 30% of XP" so it is 3 times slower. Point.

If my car goes 50% slower than yours (ie you go 100 and I go 50), it doesn't take me 50% more time to do the same distance, it takes me 100% more time (200% of your time).

Hence if I was 70% slower, my speed would be 30% of your, it'd take me 333% of your time and makes you more than 3 times faster.

[Daniel]

how to accelerate Vista?

drop it out of a 4th floor window ...