XP v Vista & 32 v 64

FatRakoon · 11 Jan 2008 at 06:24

I have been doing a couple of tests in a kind of Curiosity thing more than anything in any true serious capacity, mainly for myself and a few friends, and I have setup a PC with all 4 of the major windows versions.

XP Pro SP2
XP Pro 64Bit SP2
Vista 32Bit Ultimate
Vista 64Bit Ultimate

And I am in the process of running a bundle of BenchTests on the PC to find out which one really is better / faster.

Anyway, its early days so far, I am hoping that I can build up a better page than what I have done, but this is only a basic comparison.

I have updated my findings to this page

http://www.fatrakoon.co.uk/benchies.htm

( Dont moan about it looking poo - I know, but it does its job )

Now, if anyone has any specific suggestions, then I am all ears. I will be more than happy to run any specific tests required.

One thing I will ask though... Because I have done some other tests and I am finding that I am sat at the PC with a stopwatch, and this over 4 run throughs of any particular test is a killer... So, any Benchmark I run, I would really like to be able to just click on go, and walk away and a score will be displayed when I get back... That will make things much easier all round.

Any particular game test I might not be able to do, unless I either have the game, or the Demo has some benchmark option and then I will be happy to download anything necessary. I have 16MB unlimited so I will be more than happy to DL anything really.

Also, the PC I chose is nothing overly special, but certainly many people have a similar spec PC to the test machine so, if anyone could also get back on their scores too? - especially if they could run the test on at least 2 different versions of Windows to see any differences there?

Maybe we will find out which versions really are faster or slower perhaps?

Probably not!

Anyway, please let me know what you think.

1337z0r · 11 Jan 2008 at 08:07

This seems like a really good idea, if you want me to run a couple of tests I will be more than happy too

. Spec is in sig....

FatRakoon · 11 Jan 2008 at 08:59

Yes, if you have a dual booting PC, then run the exact same tests in each OS and let us know the figures.

I fel that the absolute best possible way to get a better and more acurate idea, would be, that if we had a number of different PCs or various specs, but for those PCs to all be running all 4 of the above versions of Windows and they also did the very same tests.

This way if my tests showed one version to be, say, 5% faster with Vista for example, and the other PCs also showed the same 5%... Then that would be the best way to verify things.

Burnsy2023 · 11 Jan 2008 at 09:09

Subscribed

Burnsy

v0n · 11 Jan 2008 at 10:03

I have a few dual CPU (Xeon) Dell Precisions at work, but we don't have any more spare licenses for XP, could test both Vistas though. And I only have DX9 cards.

By the way - it's worth looking into the weight and overheads of the os as well - take few components that always run - say explorer.exe, dmw.exe in vista, few other etc etc and compare binary sizes and memory usage at idle...

FatRakoon · 11 Jan 2008 at 10:28

Burnsy2023 said:
Subscribed

Burnsy

v0n said:
I have a few dual CPU (Xeon) Dell Precisions at work, but we don't have any more spare licenses for XP, could test both Vistas though. And I only have DX9 cards.

By the way - it's worth looking into the weight and overheads of the os as well - take few components that always run - say explorer.exe, dmw.exe in vista, few other etc etc and compare binary sizes and memory usage at idle...

It does not matter if its DX9. The tests only need to show the difference between the 4 and DX10 would kind of leave XP out wont it? - again, many say DX10 is slower and so a 9 v 10 bench would be a nice touch.

Overheads of the O/S ?

Possibly... I just wrote a load of lines about this, but decided that I may very well be completely wrong, so its been deleted. I myself feel however that while Vista is a great deal larger and does use up more of the system, it has been designed that way and so therefore, it may very well be a mute arguement perhaps? - similar to that its more code for the PC to wade through, however, its better code, so the PC does wade through it quicker... Erm, kind of?

bledd · 11 Jan 2008 at 11:08

fairest way to do it..

use the same hard drive for all tests, acronis the fresh installs (with drivers and patches)

have no av on any setup, and disable defender real-time scanning

niceone

also, use 7zip for the tests (there are 64bit builds too), it has an inbuilt benchmark system, i wouldn't bother testing network & usb speeds, as we all know vista still has it's problems

company of heroes
source engine
crysis

^would be cool if you're able to test those games

edit3

autogk tells you how long it's taken to do it's set task also.

1337z0r · 11 Jan 2008 at 11:20

I plan on doing this tonight, I will install 32-bit XP and Vista on one hard drive and 64-bit XP and Vista on another hard drive. Both are the same model ect. Hopefully I will be able to submit some tests tonight

. Good idea actually to acronis the fresh installs. Maybe we should set some guide lines in the first post?

Maybe something like...

Most be a fresh install of each operating system
No AV on XP and Vista and also disable real-time scanning in vista.
Latest drivers and all updated installed
All operating systems must be on the same hard drive or same model hard drive.

Now we can really know which OS is better

.

Burnsy2023 · 11 Jan 2008 at 11:49

v0n said:
By the way - it's worth looking into the weight and overheads of the os as well - take few components that always run - say explorer.exe, dmw.exe in vista, few other etc etc and compare binary sizes and memory usage at idle...

Like FatRakoon said, this tells us nothing useful, so there's no point

Burnsy

FatRakoon · 11 Jan 2008 at 12:03

bledd. said:
fairest way to do it..

use the same hard drive for all tests, acronis the fresh installs (with drivers and patches)

have no av on any setup, and disable defender real-time scanning

niceone

also, use 7zip for the tests (there are 64bit builds too), it has an inbuilt benchmark system, i wouldn't bother testing network & usb speeds, as we all know vista still has it's problems

company of heroes
source engine
crysis

^would be cool if you're able to test those games

edit3

autogk tells you how long it's taken to do it's set task also.

Yup, the HD I am using a pair of Hitachi 80GB SATA II Drives.
They are both 99.99% Exact.

Acronis the installs - Way ahead of you there skipper.

No A/V plus no Anything thats not absolutely vital. Although in another "To be Fair" way, I was also seriously considering installing everythign as if it was a properly used PC. After all, we all use our own ideas of protection dont we? - Anyway, in the end I just could not be arsd cos I have to do everything 4 times.

7ZIP - Again, Im on it, although while I use it myself, and I have Vista64 on my main PC, and its working fine, it has NOT setup right on the test PC? - Its not in the context menu at all?

Network/USB Speeds? - No, again, I was not going to bother with those tests, although I was considering a simple copy to and from a USB HD with a large file and lots of little ones to see any difference.... Again, I might, I might not?

The Games tests.
No problems on those... I have a few other games lines up too... Doom3, Quake4, UT2004 & UT3, HL2 ( Ok, the source Engine ) ... COH? - I got that but not sure how to bench it, so I will get that done tonight for sure.

Autogk? - ok, Ill have a look.

1337z0r said:
I plan on doing this tonight, I will install 32-bit XP and Vista on one hard drive and 64-bit XP and Vista on another hard drive. Both are the same model ect. Hopefully I will be able to submit some tests tonight . Good idea actually to acronis the fresh installs. Maybe we should set some guide lines in the first post?

Maybe something like...

Most be a fresh install of each operating system

No AV on XP and Vista and also disable real-time scanning in vista.

Latest drivers and all updated installed

All operating systems must be on the same hard drive or same model hard drive.

Now we can really know which OS is better .

The 4-installs are for me, also on 2 drives... the Hitachi 80GB Drives are the exact same model, but I am also using another 120GB Drive for various tests too!

I set them up :-

HD1-PART1 = XP32
HD2-PART1 = XP64
HD1-PART2 = VISTA32
HD2-PART2 = VISTA64

In that order... The XP PArtitions are both 30GB and the Vista partitions are the rest of each disk.... Slightly more

All drivers and progs etc are all installed in the same order... Dont know if that makes any difference and I seriously doubt it, but hey... Who knows?

Ok, mate... Absolutely ANYTHING that you can test, will be great, and especially if you can run the test on all 4

I know that many tests wont mean a thing, but the more we can do, the merrier all round.

Ok, Im going to se if I can doa bit more while Im back home.

v0n · 11 Jan 2008 at 12:14

Burnsy2023 said:
Like FatRakoon said, this tells us nothing useful, so there's no point

Burnsy

Well, here's what's behind my thinking - say you have 4Gb in 32bit Vista and 4Gb in 64bit vista - obvious gain is that under 64bit your hardware is not mapped to the same 4Gb space, so you have at least half a gig of physical memory more than guy with 32bit - however, how much of that w half a gig that you theoretically gained is already eaten away by the fact any 64bit component or binary loaded is just larger - and de facto - is it larger...

v0n · 11 Jan 2008 at 12:25

FatRakoon said:
I am finding that I am sat at the PC with a stopwatch, and this over 4 run throughs of any particular test is a killer...

In Windows Resource Kit there is a tool called timeit.exe, which is supposed to be equivalent to "time" command under unix and produce statistics on how long a given program took to run or launch. I have no idea how friendly or useful it is in windows, but maybe worth a try if you can launch benches or programs from batch files and time it that way...

valerian · 11 Jan 2008 at 12:28

does this mean its worth dusting off my oem xp 64?

bledd · 11 Jan 2008 at 12:32

looks like you're on top of things

as for 7zip context menu.. if you've installed it, called it from run

start, run, 7zfm

FatRakoon · 11 Jan 2008 at 13:04

v0n said:
how much of that w half a gig that you theoretically gained is already eaten away by the fact any 64bit component or binary loaded is just larger - and de facto - is it larger...

64Bit Components are larger?

7ZIP = 873K(32) v 1238K(64)
O&O Defrag 8.6 = 10,059(32) v 10,611(64)

I seriously doubt that a couple of hundred K will make the slightest bit of difference these days...

v0n said:
WRK = timeit.exe

Ok, I will have a look, although I need to be able to time a procedure in the app rather than time how long the app has been running for most tests.

valerian said:
does this mean its worth dusting off my oem xp 64?

Maybe... Maybe...

bledd. said:
looks like you're on top of things

as for 7zip context menu.. if you've installed it, called it from run

start, run, 7zfm

Ok, cheers for that, I will give it a shot in a bit... Im just copying my Games ISOs up to the test machine now... I was going to simply DAEMON TOOLS then directly from this PC here, but then I relaised that I will be mounting ISOS from a USB Drive on a LAN... OOFF!!! - Taking yonks, so I am copying them instead and I will mount them directly.

I have not needed to do the start, run, 7zfm before?

Kainz · 11 Jan 2008 at 13:10

Burnsy2023 said:
Subscribed Burnsy

Rofl, "I have to sub to this thread" <-My thoughts as I saw the title

v0n · 11 Jan 2008 at 13:37

FatRakoon said:
64Bit Components are larger?

7ZIP = 873K(32) v 1238K(64)
O&O Defrag 8.6 = 10,059(32) v 10,611(64)

Well - it's a valid question - it's always just the few kilo, and look - we all pretty much need 2 to 4Gb of memory to feel comfy. If all your progs on x64 were on average 20 or 25% bigger, surely it's worth weighting the impact, no?

Skeeter · 11 Jan 2008 at 13:46

3D Mark and PC Mark seem to be used for willy waving, run them on all 4.

Skeeter · 11 Jan 2008 at 13:47

Also, you could score each OS on its performance. i.e. 4 points for best, 1 for worst, and then have an easy to read total of the OS that has the highest score. better than trying to understand and compair each individually.

Burnsy2023 · 11 Jan 2008 at 14:37

v0n said:
Well - it's a valid question - it's always just the few kilo, and look - we all pretty much need 2 to 4Gb of memory to feel comfy. If all your progs on x64 were on average 20 or 25% bigger, surely it's worth weighting the impact, no?

I suppose you have a point. I can't say it's particularly useful compare to other benchmarks though.

Burnsy