: Umm, whatever.
:
: so exactly how many dead animals do you have up your ass?
:
:
: The man asked about some consumer grade benchmarks that are good
indicators
: for troubleshooting. Sandra is. I installed an application from Creative
: Labs once not to long ago for my NOMAD Jukebox.
:
: After a few hours though the system was a bit sluggish but its hard to
tell
: sometimes on such a fast machine. I didn't pay much attention to it and
then
: some days later noticed it again when doing certain things. I ran a CPU
: bench in Sandra and noticed that the score was allot lower (25% or more
less
: then usual). I did a system restore to before the application install and
: things were right back to normal again.
:
: Sandra was very helpful in discovering this performance robbing
application.
: If I had 'just' built the system and installed the application from the
: start I would have not known that this application was robbing my system
of
: performance.
:
: For years now, when I install a fresh OS I run a set of benches and
: occasionally check them. Especially after driver installs. I also find it
: informative to compare to others and see if I am in par.
:
:
:
:
: : > SST,
: > My question still stands what are you trying to prove? CPU performance
is
: > only one criteria in the overall performance of a system; MIPS or MFOPS
: are
: > only important if you're executing a particular algorithm. The fastest
: > processor in the world doesn't make much difference if only 5% of the
: > processing time is spent multiplying or dividing.
: > Drystone/Whetstone are synthetic tests they measure certain attributes
of
: a
: > processor but they're not much good at telling you what will happen in a
: > real world situation. It would appear you are interested in numbers to
: > impress in which case they may be of value to you. If you're really
: > interested in benchmarking try the ACM or IEEE sites for info.
: >
: >
: >
: > : > : not really.
: > :
: > : CPU Dhrystone and Whetstone in terms of MIPS and MFLOPS are standard
and
: > : commonly used to accurately compare CPU's. Same for Interger and FPU
: > : calculations.
: > :
: > : This isn't like 3D Marks bench where a video card can be tricked.
Their
: is
: > : no driver responsible for running a Pi Calculation or a
: > Dhrystone/Whetstone.
: > :
: > : I'm using Sandra 2003 (2003.7.9.73)
: > : P4c/Abit IC7 @ 13x X 250.6 = 3257.5 MHz (1.60v)
: > : Sandra Mem 5472 mbs | 5474mbs (200.5MHz, cl2-2-2-6)
: > :
: > : Sandra CPU 9899 MIPS | 2812 / 6191 MFLOPS
: > :
: > : Sandra MM 15,013 it/s | 23992 it/s
: > :
: > : Sandra HD (C:\) 30846 KB/s (Western Digital 120GB SE)
: > :
: > :
: > :
: > : 3DM01 (371/344):
: > : 19,292 (dx8, cat 3.4)
http://service.futuremark.com/compare?2k1=6661809
: > : 18794 (dx9, cat 3.6)
http://service.futuremark.com/compare?2k1=6769660
: > :
: > :
: > : : > : >
: > : > > Performing what, matrix inversions, Fourier transforms, gaming?
What
: > : > > do you want to prove to your friends? Anyone worth his pencil can
: > : > > write code that will make your box look like an absolute cripple.
: > : > > Going the other way is not quite so simple. Benchmarks don't
prove
: > : > > very much except in specific cases. From the hardware you list
you
: > : > > have a fairy high end chip and a cheap old cycle stealing sound
: card.
: > : > > You don't say whether your modem is hardware or software. Not to
: rain
: > : > > on your parade but sounds like the on commission computer salesman
: go
: > : > > to you.
: > : > >
: > : > > claus
: > : >
: > : > I agree. benchmarks are only useful when performing tests on the
same
: > rig.
: > : > ie overclocking memory while raising timings. As nvidia has shown
us
: > : > hardware can be optimized to benchmarks so it's value in comparing
: > : > different systems has become an apples to oranges ordeal.
: > : >
: > :
: > :
: >
:
:
Well woop-de-do. Animals? Ass? Are you trying to be funny or insulting
instead of just plain stupid?
And after accumulating all these numbers what do you do with them add them
to a little table and do cross tabs or co-variances and tell yourself "if
I'd just installed x instead of y I'd get a higher or lower number? DO you
get right in there and modify the kernel so its better at interrupt handling
for fast devices or perhaps you've optimized the buffer size for DMA write
or something. You really should go back to dos 2.0 of you want a tight
little operating system fast as hell but not much good with anything more
that 640k.
How many drivers have you written? Not a ****ing one I'll bet, but hey you
can say rev2.3.5 is faster than rev2.3.4 but it doesn't works unlike the
previous one so I'll stay with the fast one cause jesus the number tells me
so.
You enjoy yourself.
claus
