Comparison of 9.x and XP

Alright, this was discussed here a while back and even though many of the members disagreed with me back then (and now you will be named and shamed ) I decided to go through with it. But first...

Quote:
No i wouldn't, and i don't think it would be helpful.



Quote:
It wouldn't show much difference.



Quote:
What's the point?!!



They almost convinced me not to bother with this whole thing and just buy and install XP straight away. But I would now like to thank these twelve people for giving me firm backing on this and making the vote tilt my way just that little bit :

Quote:
Yeh it'd be interesting to see / could be helpful.

Thank-you all so much .

Right then, down to business.

The purpose of all this has been to show differences in performance between the 9.x series of Windows OS's and Windows XP Home Edition (excluding 95 due to major compatibility issues with both hardware and software & 98 as it was a: replaced by se for a reason, and b: I didn't have a copy of it ). A secondary objective was to see how much variation in performance a clean installation (only the benchmarks and boxed drivers installed) had over a full installation (basically everything I had installed on my old version of ME thrown in and messed about for a day or so + the latest driver updates).

I used 3D Mark 2001se, PC Mark 2002, AquaMark 3, Halo: Time Demo, SuperPi, and ScienceMark2. The reason for choosing these benchmarks was covered in the thread mentioned above but I will make a small point about why I didn't use the the later versions of some of the benchmarks (ie. PC Mark 05 and 3D Mark 05) for this test: "They would not have run on the 9.x OS's". This is very important and is one of the main reasons why the results came out as they did, 9.x is now showing it's age in both support for software and for the latest hardware (the one that really matters to US here at Dev Hardware).

The System Specifications were not altered through out the benchmarks in any way (yes, I really did go two months without changing things about in there!) and the hardware specifications are as follows (please don't laugh):

CPU: Pentium 4 'E' Prescott 3.0GHz S478 1Mb L2-Cache (Stock frequencies and Voltages)
RAM: 256MB PC3200 Single-Channel CL: 2.5-3-3-8
Motherboard: P4P800 E-Deluxe
Graphics Card: ASUS Radeon 9200se 128Mb Memory (Stock Frequencies)
HDD 1 (Primary): Maxtor 7200rpm 40Gb
HDD 2: Quantum Fireball 5400rpm 30Gb
PSU: Sumvision Generic 400watt (cost me £12, what can I say?)

Now that that minor embarrassment is over with I'll get down to the nitty gritty details of the tests:

3D Mark 2001se,PC Mark 2002 & AquaMark 3:
Just ran each under default settings.

SuperPi:
Ran the 128k, 256k, 512k and 1M calculations.

Halo: Time Demo:
There are not options for this other than adjusting the game detail levels. I simply set the game to 800x600 with no Specular or Shadows and with Particles off and Texture Quality to Low. ( come on! like a 9200se could handle any more than that ).

Science Mark 2:
Originally I had planed to run every test on this, but it turned out that neither ME or 98se could handle the 'STREAM' or 'Memory Benchmark' tests. So instead I ran a combination of 'Molecular Dynamics', 'Primordial', 'Crypt' and 'BLAS/FLOPs' (basically the tests that WOULD run) one after the other and all with their default settings. These seemed to show good results and appeared to be mainly CPU intensive (I'll explain why i think this is important in a sec.)

I gave each program a fresh boot so they got as much of the resources as possible. However I wasn't going to re-boot after each test in Science Mark or after each calculation in SpuerPi.

For each OS I did:

1) A 'Clean' install, which consisted of installing the OS, the standard drivers that came with the motherboard and graphics card, and the benchmarks. Using DirectX 9.0b (required to run the Halo benches). Nothing else was installed.

2) A 'Full' install with all the gubbins, bells and whistles I have lying around thrown in. Updated to the latest driver versions and DirectX 9.0c. The idea of this was to mess up the registry a bit and fill it up with entries, but also to see how much performance can be lost by using the 'stock' drivers instead of updating them.

And I also included the benchmarks from the '6 Month Old' version of ME had, just to act as a comparison on how much difference a fresh installation makes.

How This All Came About:
My hunch was that 9.x couldn't handle modern day CPU's (this is why the CPU intensive tests seemed like a good idea). I basically felt that my P4 wasn't performing as well as well as it should (this was under Windows ME) so I decided to upgrade to XP. But before I did I got thinking (a dangerous thing!)... what if I tested ME and then XP to see for sure if it was worth it. So I started the poll and found that some people through it wouldn't be worth it, others didn't want to know, but the majority did (by one!!). That was good enough for me .

You've Herd Enough:
Yes yes... I do tend to ramble on a bit so now I'll just plunge right into the results and pick up my rambling later on .