WindowBlinds 3.3: Twice as vast as msstyles

I used to benchmark my system almost daily but haven't been that bored for years... Whatever happened to the Landmark Speed Test? I'd love to test it if someone knows a good benchmarking utility (that doesn't take hours to run). It would be interesting to see how fast different skins are too.

see, by the title i thought this was going to be about windowblinds being like the pacific ocean to msstyles atlantic...twice as vast, get it?

ve have vays.......

DavidK: SkinBench is at http://sourceforge.net/projects/skinbench/ if you want to try it out.

thanks!

It appears there isn't anyone else out there willing to do some benchmarking.

Of course, if someone does please keep in mind that a skin by itself can make a huge impact.

Pat,

I have never understood this continuing drama over WB being so slow. My home computer is nothing to write home about, especially when compared to the new offerings out now. I have been using WB for a year and a half now and have never noticed a significant performance hit. I boot minimal, with nothing to speak of running, often to do disk clean up etc. and I can not tell the difference.

So in the benchmarking of viual perception I can say WB is a transparent task!

I benched ChosenOS WB style vs ChosenOS msstyle.

http://www.windowblinds.net/benchmark_apr02a.jpg

OK Froggy - I must be an idiot. Your chart is a bit obscure. Is the bottom value expressed in milliseconds? If so, then doesn't this contradict your statement that WB is faster?

Am I misreading this data?

i gotta agree with russ here cause this chart of yours sure makes WB looks slower than .msstyles.

I too, must agree. While wb is quicker in a few areas, the majority of tasks are quicker with msstyles.

The chart is based on the skinbench score. The higher the score, the better.

MooShoo, I'm afraid you are mistaken. I am not aware of any hardware configuration in which WindowBlinds isn't significantly faster. Low end, high end, inbetween, ATI, Matrox, Nvidia, Trident, AMD, Intel, etc. Doesn't even matter whether it's UIS2 or UIS1. If it's UIS1, the performance is up to twice as fast. If it's UIS2 then it's about 20% to 50% faster. Only on extremely complex UIS2 skins will it be slower than an msstyle.

Sorry, I spoke without being educated, I just naturally thought faster meant a shorter bar, that's what I was basing my last post on.

One thing that I noticed was that MSstyles was still loading on startup, this dialog boxes were skinned with the luna skin and then would switch over to my WB skin a few moments later. I did a little research and the "Themes service" was up and running. Turning it off did make a noticable differance. WARNING!! do not play with services unless you do your HOMEWORK first. A great place to start is Black Vipers home page http://www.blkviper.com/ a little bit of research goes a long way.

alright, i bit....

firstly, let's go ahead and get the link here, so y'all can see the results...may as well open it in a new window so you can look at it and read this at the same time.

meph's skinbench results adventure ride of fun

now, up top you'll see the source data, and below that the graphs. i chose to do individual graphs per test since, well, they're individual tests so they should be shown seperately, as well as seeing what the breakdown's are like.

now, to the right is the color key, and i tried to get the color behind the individual datas in teh table to match those of the graphs, just for consistency (i'll come back to that word later)

Now, with each skin name, you'll also see a parenthetical letter/number...M stands for an msstyle skin, 2 for a UIS2 skin and 1 for a UIS1+ skin.

interestingly enough, notice Luna (which is installed by default by windowblinds) has a 1. keep that in mind after i go over a few of the tests.

alright, on to the data!

the data goes from bottom to top, so, we'll start at the bottom.

i'm just gonna point out a few things that y'all can see but that it's good to have said aloud regardless.

now, where windowblinds shines, and this data shows it, is moving windows about, partly because they are hardware accelerated. msstyles get some acceleration, but it's not in moving windows about and it's also nothing to write home about either...now notice, moving and resizing windows, windowblinds trounces msstyles by all of 150 skinmarks for moving and 230 for resizing. now then, this is only for the uis1 skin. for the 2 uis2 skins, they are slightly better to below their corresponding msstyles variants.

from there msstyles are ahead for editbox resizes, button moves, checkbox moves, radiobutton moves, up until activating a window. here's the odd thing. msstyles is just as fast as uis1 luna, and faster than the uis2 versions...by a large amount.

then again we have moving and resizing which, of course, windowblinds takes hands down. but, interestingly enough, windowblinds chokes on resizing the editbox behind windows...sure, it doesn't happen in real use much, but still, it's interesting nonetheless.

then vertical and horizontal resizing, which windowblinds takes again, due in part to the hardware acceleration i'm betting. from there it's pretty much even for the next three, msstyles takes the tab control cycles (which cycles between selected/unselected repeatedly)...

...then we come to the oddball. that being the progress controls, and this being where windowblinds makes up it's difference. if i were doing this for a lab, i'd have to throw out the progress controls as being an out-lier(sp?) because there's no way it's statistically near any of the other tests. something's going on there, but i have no clue what it could be, especially when you've got windowblinds up 10 times what msstyles has, and such great variation within it's own data (between windowblinds skins)...

moving on you've got a listview header control cycle, which is about 'even' aside from the luna 2485 skin which is actually not skinning the headers so it can be thrown out.

then the totals. windowblinds wins in most cases, xp ocean (another default install skin) loses to both luna and preluna (it's msstyles counterpart). now, interesting turn of events, take out the progress controls score, and i'm betting windowblinds wouldn't win the total score, which is what most people would care about. i'd do it myself, but i don't know the math as to how the scores are figured up cause it's based on cycles per a given amount of time with some proportionality factor as compared to classic thrown in to get a relative score.

now, conclusions:

this is very interesting data i believe. just so y'all can see, i'll now post the contents of my wb.ini file to show that no favortism was given:

NoShowWarning=0 noProgress=0 noTabs=0 noScroll=0 nosound=0 SizeLimit=0 WindowLock=0 UseHWAcc=0 WARNINGMESS=6 Hyperpaint=1 ForceDDB=1 HyperPaintMode=4 HideMinMax=1 Gamma=660 wbd=0 Path=C:\Program Files\Object Desktop\WindowBlinds ResizeOpt=1 ;skindisplay=1

note also that this is wb 3.3g, newest build off of odnt that i could find.

now, what does all this really say? it says that windowblinds being TWICE as fast as msstyles isn't quite right. more like TWICE as fast...in some situations, and about as fast in others.

now, onto that word consistency that i said i'd come back to...

first of all, skinbench is good in theory...no offense to greenreaper, but it's not fleshed out well enough to really mean all that much. the newest version i could find was from october 19, and it had just added in menu and header tests (i think). now, in the 6 months since then, there's been no new modules added in, yet there've been new controls skinned and able to be tested. now, if i knew how to code for new modules, i'd try my hand at getting something put in there, but, i don't, so i won't. i'd probably add in the start panel being opened and closed, submenus on it being opened and closed, taskbar animations, and system tray open/close. but that's just a feature request i think.

the second is overall consistency in general. i was disappointed to not be able to find a true luna skin to be able to test with, since the included luna isn't quite the same as the msstyles luna (colors different, menus aren't skinned, fonts are way off, some images (scrollbars, toolbar backgrounds) are outdated, etc.)...

...and that's the thing. yes windowblinds is still evolving, but, consistency (and i've harped on this before, i know) is still lacking. you never know what you're going to get when you install a new skin, much less open up a new app or even new dialog box....

...it's like you go out and you buy this game that you've heard rave reviews about and you're playing and it's great, it's just what you expected. but then, you're going, and something's just not right. like, some texture doesn't line up right or there's a spot where you shouldn't be able to jump up to but you do make it up there and it causes you to get past the bad guys without firing a shot...something that just makes you think for a split second that you're not suspending belief and that you are in fact playing a game. that's how windowblinds is. benchmarks can only show so much, actual use shows so much more.

windowblinds is great, no question. but, it's not living up to some of the hype. there's just enough small things to pull you out and say 'hey, i can see what windows is really like, and it frightens me' or 'hey, isn't windowblinds supposed to take care of that? what's going on?'

it's like quake 2 multiplayer was to the old quakeworld...quake2 was a far superior game, no question about it (sonic mayhem's soundtrack couldn't hold a candle to NIN's quake1 work though) but still, quake 2's multiplayer felt so lagged compared to quakeworld...and patch after patch after patch didn't help. carmack eventually just said 'look, quake2 IS the faster networking subsystem' and had to 'break' quake 2's multiplayer so it felt as fast (and not as lagged) as quakeworld...

the moral? it's not so much raw speed as it is the illusion of speed...

hope this helps whoever needs it...oh, system specs in case it's necessary:

64mb geforce 2
winxp
512 mb ram
p3 550
detonators, version 28.32

Holy smokes! It is much much faster then the old version.

It might be that this is a slow system and now it is using hardware acceleration.

Mephisto,

We both know each other's arguments so I'll boil it down to the basics:

There are a handful of GUI elements that WindowBlinds doesn't currently skin such as the task pane in Explorer and the spinner controls and the scrollbars in drop down combos.

On the other hand:
Windows XP by default doesn't skin 99% of existing applications. It only fully skins "theme aware" apps which you can count the # of those on two hands and only in their latest/greatest version.

For me, XP without WindowBlinds looks very incosistent. It is weird loading up Word XP and having the scrollbars be the Windows 95 style. It is odd loading up Corel Draw and having it look mostly like a Windows 95 app. But when I run WindowBlinds, it skins all that.

So the question is which inconsistency is worse. For me, XP not skinning existing apps is worse. Clearly you feel differently.

I think it's really a race - will WindowBlinds skin those remaining GUI elements BEFORE a significant number of apps become theme aware. If we're having this conversation a year from now, then WB has lost that argument.

The fact is (for me anyway) no matter how pretty you can make WinXP with an MSStyles theme, it still will never compare to the advantages of running Windowblinds. I have become addicted to being able to roll up a window with a right click on the border. I love having a clock in the window, being able to play MP-3's from controls on the window border, launch an app... etc. Plus, for my money, WB looks so much better than XP themes it's not even a fair comparison. The added benefit of skin suites letting you match the theme with ICQ and Winamp, NextStart, etc. skins is another bonus.

It's not a debate worth wasting my time on, if MSStyles can't do all the extra stuff and can't skin every program and ONLY works on XP and is limited to so few original themes (the abundance of them being straight XP looking themes or rips) then, where's the debate? My Win2K box looks as cool as my XP box and both of them let me work the way I prefer. I get frustrated trying to work on a machine that doesn't have WB installed....

As far as I can tell, the MSStyles stuff adds nothing to WindowsXP but different colors and border styles. Windowblinds (expecially coupled with ObjectBar) frees the user to customize the functionality as well as the appearance to whatever his imagination can dream up, and it does it consistently across ALL windows platforms.

No comparison at all.... (unless you know so little about computers that changing colors is still the farthest you've gotten in customizing how you work on your machine, in which case you are just beginning a wonderful journey into tweaking and will some day realize the difference.)

hehe, David, "...some day realize the difference", I must say that your last thread is a very good answer!

Frogboy, the 99% of the people don't care at all, about the WindowBlinds speed, they don't even think about it, they just like WB Skins and they use them. Any way, it is good for us to know that WB are very fast too!

Of course, it's the skinners doing the skinning that will opt for the freedom of design that Windowblinds allows....No-one wants to be shackled to limitations of both creativity and universal application.
Imagine doing an 'XP-only' LiteSTEP theme....the guy with the 2K box is going to feel miffed, and the old guy in the corner still bumbling along with 95....he's gonna mutter 'not fair'.
Solution is to create OS ver-independent themes.....another reason Windowblinds will always be favoured over styles, by both skinners and users alike...

mephisto, in your wb.ini you have
UseHWAcc=0
Does this mean you are not using hardware acceleration? How does changing this value to 1 affect the benchmark?

I have tried both Mmstyles and Windowblinds and I prefer WB - adni18 is right, speed is not important (just as long as it is not slow...umm is that a contradiction??) the extra features in WB such as smart buttons, roll ups and custom colours all make me want to stay with WB.

Which is faster? Astin Martin Vanquish V12 or Ferrari 550 Maranello? Who cares I'll take either home

(Actually, I would always go for the Aston Martin)

SkinBench does need a bit of work, as indicated in it's version number. And some of the tests are blatently incorrect. Part of this is due to me doing timing incorrectly, which will be fixed. I have had a few bugs reported which I unfortunately can't reproduce. Yes, the edit sizing thing is strange, and no, I have no idea why it happens. I tried adding menu stuff a while back but wasn't able to get them to open etc under program control (if anyone knows how to do this, I'd be glad to hear). I agree that the Activate test result is surprising in relation to the other tests.

The total score is meaningless, as it says (I took it out, but people asked me to put it back in). Who decides how much each thing matters? This feature will *not* be in the next version of SkinBench, and anyone who tries to total up the tests is not really using the results data correctly.

The tests I consider to be core are moving and resizing ones (particularly the Back ones, since they involve both resizing and repainting of revealed windows) because they are generally the actions that take the longest time, happen the most, and they're also the most reliably benchmarked. Any test that gives wildly different results like the progrees bar test is highly suspect. Incidentally, that's why you should calibrate to classic - that way if anything gets significantly more - like, 1000's of marks more - you can be fairly sure SkinBench has messed up somewhere.

I'd also point out that SkinBench can't measure everything. WB 3.2 introduced global bitmaps that cut the load time of windows to a third of what it was previously. Yet this would not have been indicated in any benchmarks (possibly the CreateWnd one, but I doubt it). That's just one example of factors that may have a huge effect on the user perception and "real" speed, but which may never be benchmarkable.

Hum... I don't know about benchmarks and all, but I am have been working without Windowblinds at home for a week or two now, and I can feel the difference. Things seems smoother for some reason.
But this is my home PC mind you, it's old and crappy, so maybe WB is faster but on faster machines? My home PC is a 333 Celeron with 256 Megs of RAM, and an 8 Megs 3DFX Voodoo3.

I love Windowblinds, but for the moment I actually like the plain look of Windows XP without fancy running. Don't worry, I'll grow out of that phase too.

The main thing we're trying to do is dispel the whole "WindowBlinds is slow and bloated" stuff that you occasionally still see posted.

It's really quite discouraging seeing some guy say how much faster msstyles when he hasn't tried WindowBlinds in years and when he did it was on like Windows 95 or something.

I want to make sure that from 3.3 on out, that the difference comes down to skinning and features, not performance and memory.

Maybe you should just go with saying "faster" rather than "twice as fast", then. More reliably true.