Comparing the speed of the computer is a bit like statistics. In otherwords it's all lies.
it's possible to get the results you want by using different standards.
There's the apparent speed to the user which can be influenced by just about anything immagineable.
The efficency of the software (including the operating system), whether or not it was compiled with CPU optimisations or not, and if so which ones. The CPU in the computer, it's clock rate, bus speed, hard drive access time, the speed of the hard drive interface (accidentially running the hard drive in PIO mode is a common cause of a PC that runs really slowly), memory clock rate, dual channel or single channel memory access and so on. For computers running windows, hard drive fragmentation can be a factor on computers that have had their OS installed for a long time. Even NTFS suffers from fragmentation, especially on a drive that is almost full. Once there is less than about 1GB to a few 100Mb free it is actually far far worse than FAT32 and almost gurantees that any files written will be fragemented because of the way NTFS allocates it's storage bands for files.
As for the question, yes the design of the operating system has a big impact on the performance from a users point of view, but there are so many factors involved it can be hard to say exactly what is going to run faster.
For example i have an old 800mhz celeron laptop with 128 mb of ram the display is smashed so i can't use it as a laptop. But that actually ran Windows XP home edition a lot faster than my current desktop even though my desktop is 1.3ghz and has 256mb of ram. The reason is this case is because my desktop has a badly designed motherboard, despite being 2 years more recent than the laptop.
So the short answer is it depends on 100s of factors and 1000s of factors if you want to consider the difference between different CPU architectures, for example clock rates between different CPU types can't be compared. To do that you need to know about both CPUs instruction sets, the pipelining of the CPU, instruction timings, cache design/performance and so on.
For example some CPUs have instructions called SIMD which basically means a single instruction can perform a calculation on a whole group of data. On some CPUs (basically everything other than X86 infact) this can be done in 1 clock cycle.
As an example this is why the sony playstation 2 (basically a glorified MIPS R5900) is actually far more powerful than the xbox (a 700mhz PIII) at performing vector maths despite being only 333mhz. The research departments of a few universities over here use them to run things like global warming simulations which need lots of vector maths. The R5900 CPU can do something like 10 64 bit calculations or 5 128 bit ones in one clock cycle as long as the data is vectors. So despite only doing that 333 times a second because the PIII needs 7 cycles to do the same thing the R5900 does in 1 it's slower at doing these kinds of tasks despite being over twice as fast in terms of clock speed.
This is also why it's completely meaningless to compare the clock speeds of PC and (non x86) Macs, unless you know the details of the software in use, and how efficent the compiler was, which by definition in closed source software, you don't. In the case of the XBOX microsoft almost cirtainly advertised the clock speed lots because they know that most people who don't know anything about microprocessor design (which is almost everyone) think that a high clock speed is always better. So they'd by the xbox. This is an example of how specs can be made to lie. Additionally a rubbishy compiler on one CPU can make the reverse of what should be true according to what you learn in CS actually the case. I speculate that this is why mac OS X seems to run so slowly on the macs that were around when it first came out. The OS most likely doesn't touch the G4s altivec units at all, wasting a lot of it's capability.
If you were to do a comparsom between the R5900 (playstation 2) and P III using another type of calculation, you'd get completely different results (anything that's not vectors would go a lot faster on the PIII, games are almost all vector processing which is why game consoles tend to use CPUs with good vector units but are bad at everything else), basically what i'm getting at is that most supposed measures of computer performance are absolute rubbish, were designed by people that know nothing about microprocessor design and can only be used as a guide at the very best.
~EG (maria)
25 Aug 2005, 4:02 pm
