Email conversation
| From | Anonymous |
|---|
| To | Me |
|---|
| Subject | comment browser speed test |
|---|
| Date | 11 February 2005 22:41 |
|---|
hello,
I too love scripting CSS and javascript, knowing that everything is
perfect on a site gives me a kick :)
anyway, i read your browsertest, and i'm convinced it isnt
representative for the best performing browser...
first of all, the influence a cold or warm start should have on the
overall test results should be minimal, as this is not what counts,
unless the startup time is abnormally slow or fast i would not even
consider handing out extra points to a browser for that, besides it's
all based on how your hard drive is divided. if you installed a
certain program first it most likely will be more near the centre of
your hard disk so itll boot faster coz the disk has to make less
cycles to load it, and you also haven't taken into account that your
hard drive might be fragmented, all sorts of things you should take
into thought
and the css benchmark test is completly unreliable, my figures have
all sorts of highs and lows, so that test wasnt really fair either
i'm sorry for all this negativity, but i do have a positive thing to
say though :); i discovered a site called [name erased], it loads 250
random icons from all over the web, each 30 secs a new set of 250
icons is available to download on the page. because of the way this
works, you would have to load each browser at the same time to see how
fast they load the 250 icons, because if you were to do the test
separately, every browser would be rendering 250 different icons, and
that wouldnt be fair as some icons might be from the uk but others
might be coming from japan or australia or something and take ages to
load... ([address erased])
even this test is in some cases not fair, as the default max amount of
http connections for internet explorer is like 2 or 4 or something,
and firefox has a default of 8 or 16, i'm not too good with numbers.
and ofcourse you can also tweak all these settings via various tools
or registry tweaks which make the whole thing even more complicated
i also knew about 24fun for testing javascript, and this *is* a good
site for measuring speed, as my results hardly deviate from the
previous results if i retake the test.
in short i just think its either very hard to near impossible to
measure the real speed of a browser. the only conclusion i can draw
from any tests is that the default internet explorer 6 is reaaally
slow and the default Opera 7 is superfast at rendering javascript
anyway, sorry for making this so long, i was just typing my essay for
school and was doing some research and i came across your site. i love
the scripts, ive made many of them myself already, i especially liked
the capslock-recognizer, very original :), maybe only a gimmick but
still nice to have :)
cyaa
Anonymous
| From | Me |
|---|
| To | Anonymous |
|---|
| Subject | Re: comment browser speed test |
|---|
| Date | 14 February 2005 10:10 |
|---|
> anyway, i read your browsertest, and i'm convinced it isnt
> representative for the best performing browser...
The tests were done for the specific reasons you mentioned. I did a series
of different tests, and then summarised based on the relevance (to my mind)
of each test. So you will note that I do not include much weighting to the
startup speed, as typically that is only done once per session.
However, many 'normal' people who I see using browsers (typically Internet
Explorer) have a tendancy to open a window, load two or three pages, then
close the window, and then a few minutes later, they open a new one. This
seems very wasteful to me, but that is how they work. By giving a full set
of results, I allow those people to look for themselves and use their own
weightings, so they can decide for themselves which is the fastest for the
way they will use it.
> it's all based on how your hard drive is divided
Indeed, that is true, to some extent, although all the computers I used
have got fairly new installs without much fragmentation. As for position on
the disk, well, it's not like I have much control over that, so I have just
done the tests as best I can with what I have available.
> the css benchmark test is completly unreliable
no it is not. although I will agree that it may or may not be the most
accurate representation of a CSS driven layout, it does manage to show how
fast a browser can process CSS. As for the different results each time,
there is a good reason for that. The first time a browser loads the page,
it is usually slower, then with each subsequent reload, it gets faster, and
occasionally jumps to being slower again. this is not a bad test. This
accurately shows the way that browsers will allot memory/priority to parts
of the engine that it decides need the priority. I have discussed this with
other developers, and even with the programmers of the browsers themselves,
and they agree that this variance in results is expected behaviour.
Anyway, like I say in the article, you can feel free to prove me wrong;
write your own tests, explaining why each test is a fair test (or not as
the case may be) and what each test represents, test all the browsers,
publish your results.
Mark 'Tarquin' Wilton-Jones - author of http://www.howtocreate.co.uk/
| From | Anonymous |
|---|
| To | Me |
|---|
| Subject | Re: comment browser speed test |
|---|
| Date | 14 February 2005 22:28 |
|---|
> However, many 'normal' people who I see using browsers ...
agreed, though it should have limited influence on the actual outcome
of which browser is best
> > the css benchmark test is completly unreliable
> no it is not. [...] it does manage to show how fast a browser can process
> CSS.
also true, but I think this test should be taken with a grain of salt,
providing the results up to the millisecond is confusing, since the
test results in serious deviations you should consider rounding down
some values so only really big differences between browsers are
visible, the way i see it now, its like browser x has a 500ms
advantage over browser y, which *no* consumer will notice at all, i
mean, 500ms is really almost nothing
> Anyway, like I say in the article, you can feel free to prove me wrong;
> [...] test all the browsers, publish your results
I'm not planning to do this, but I thought your test was an
interesting read, and again, please check out [address erased],
hopefully you'll add that test to your already very extensive browser
test :)
Anonymous