Optimization - Why do skin designers ignore it?
Tuesday, March 23, 2004 by chris lee | Discussion: WinCustomize Talk
Is it just me or do many skin designers just not take into account of performance when designing skins? 
I have two machines, a fast one and a slow Windows 2000 machine.
Some skins just don't run fast on the Windows 2000 machine, until I optimize them that is!
The simple process of reducing the number of colors on the bitmap images to 256 or 16 has a dramatic effect on how fast the machine will run. You can reduce transparency in a lot of images where the background can be part of the image.
99% of skins seem to have all 16 millon colour images, this is just crazy as the reduced color images look just as good as long as you use an optimization tool such as the one in Paintshop Pro.
Sure you could say get rid of the old machine, but I can still tell the speed difference even on the new machine.
So why have lower performance however fast your PC is when the results are just as good?
I have two machines, a fast one and a slow Windows 2000 machine.
Some skins just don't run fast on the Windows 2000 machine, until I optimize them that is!
The simple process of reducing the number of colors on the bitmap images to 256 or 16 has a dramatic effect on how fast the machine will run. You can reduce transparency in a lot of images where the background can be part of the image.
99% of skins seem to have all 16 millon colour images, this is just crazy as the reduced color images look just as good as long as you use an optimization tool such as the one in Paintshop Pro.
Sure you could say get rid of the old machine, but I can still tell the speed difference even on the new machine.
So why have lower performance however fast your PC is when the results are just as good?
 |
|
Page 2 of 2 | |
|
|---|
chris lee
Reply #22 Wednesday, March 24, 2004 4:16 AM
Reply #22 Wednesday, March 24, 2004 4:16 AM
Shameless Designs said:
By the time I have everything optimized for appearance and functionality, there is no way I'm going to go back and spend hours and hours doing color conversions and then rechecking the skins in order for someone to save a couple hundred kilobytes of RAM or hard drive space.
When talking about performance it is not about the small amount of disk space saved its all about rendering of the graphics on the screen.
When skinning with such as Windowblinds the borders of windows has to be rendered again and again for every single window displayed.
You don't have to spend hours optimizing every little graphic, targeting specific areas such as borders will do the trick. Reducing transparency in a few locations can work wonders.
I know people say 'get a better PC' nice for those of us lucky enough to afford it, but for millions a PC is a major investment.
I've yet to find a machine, no matter how powerful that does not benefit from skin optimisation especially when you've got the system fully loaded with most of the object desktop suite.
if you try to reduce the color depth, the carbon fiber effect is completely lost.
Ok so I don't want to get in the way of artistic flare, I don't expect you to reduce colors when the effect is a poor looking skin.
Not every skin is suitable for such optimization. If it's just too much hastle in the post process then ok don't do it, I know its free so as designers its your call.
If you want more people to enjoy your work then do a little tweaking in a few places.
Automatic optimization would be nice, but I don't think its a realistic prospect since skins are such a complex subject.
By the time I have everything optimized for appearance and functionality, there is no way I'm going to go back and spend hours and hours doing color conversions and then rechecking the skins in order for someone to save a couple hundred kilobytes of RAM or hard drive space.
When talking about performance it is not about the small amount of disk space saved its all about rendering of the graphics on the screen.
When skinning with such as Windowblinds the borders of windows has to be rendered again and again for every single window displayed.
You don't have to spend hours optimizing every little graphic, targeting specific areas such as borders will do the trick. Reducing transparency in a few locations can work wonders.
I know people say 'get a better PC' nice for those of us lucky enough to afford it, but for millions a PC is a major investment.
I've yet to find a machine, no matter how powerful that does not benefit from skin optimisation especially when you've got the system fully loaded with most of the object desktop suite.
if you try to reduce the color depth, the carbon fiber effect is completely lost.
Ok so I don't want to get in the way of artistic flare, I don't expect you to reduce colors when the effect is a poor looking skin.
Not every skin is suitable for such optimization. If it's just too much hastle in the post process then ok don't do it, I know its free so as designers its your call.
If you want more people to enjoy your work then do a little tweaking in a few places.
Automatic optimization would be nice, but I don't think its a realistic prospect since skins are such a complex subject.
Shameless Designs
Reply #23 Wednesday, March 24, 2004 4:51 AM
Reply #23 Wednesday, March 24, 2004 4:51 AM
OK, I'm getting real curious now...
Hippy and I maintain that the performance gain will be marginal. You, Chris, along with paxx and Digital Fusion insist that it is significant. So give me an example with some numbers behind it. Apparently you have tested this theory, so let's have some details. Which WB did you use? What method did you use to reduce the color depth of the graphics? How did you benchmark the before and after performance? What are the results?
Give me some solid evidence, something we can verify ourselves. I'm not trying to be a smart alec here. I really am curious now. I want to check this out. In fact, I'll try to find a couple popular skins here that have fairly uncomplicated graphics and were saved in 24 bit color depth and in the next few days I'll convert and test them myself. But I really would like to hear the evidence you have already that led you to your current stand on this subject. Who knows? You might convince a few skinners to change their habits...
| When talking about performance it is not about the small amount of disk space saved its all about rendering of the graphics on the screen. When skinning with such as Windowblinds the borders of windows has to be rendered again and again for every single window displayed.You don't have to spend hours optimizing every little graphic, targeting specific areas such as borders will do the trick. Reducing transparency in a few locations can work wonders. |
Hippy and I maintain that the performance gain will be marginal. You, Chris, along with paxx and Digital Fusion insist that it is significant. So give me an example with some numbers behind it. Apparently you have tested this theory, so let's have some details. Which WB did you use? What method did you use to reduce the color depth of the graphics? How did you benchmark the before and after performance? What are the results?
Give me some solid evidence, something we can verify ourselves. I'm not trying to be a smart alec here. I really am curious now. I want to check this out. In fact, I'll try to find a couple popular skins here that have fairly uncomplicated graphics and were saved in 24 bit color depth and in the next few days I'll convert and test them myself. But I really would like to hear the evidence you have already that led you to your current stand on this subject. Who knows? You might convince a few skinners to change their habits...
juni
Reply #24 Wednesday, March 24, 2004 6:23 AM
Reply #24 Wednesday, March 24, 2004 6:23 AM
I haven't noticed much performance slowness when making hi-res skins, but maybe the pocket pc is different in that respect.
chould change my title to "small screen skinner"
chould change my title to "small screen skinner"
kongit
Reply #25 Wednesday, March 24, 2004 7:01 AM
Reply #25 Wednesday, March 24, 2004 7:01 AM
I for one have noticed a difference between a uis2 without some optimization and a really optimized uis1. Many themes could be optimized without any hurt to the graphics by a simple option called tile. now if people think that it takes to much time to change the parts of a skin that were stretched and could be tiled to tiled I think they are wrong. I remember finding one skin a while ago that all I did was change strech to tile where applicable I could notice a large difference in speed. However many of the skinners already know this and since is isn't exactly difficult to employ it.
just my $.o5
just my $.o5
Shameless Designs
Reply #26 Wednesday, March 24, 2004 8:01 PM
Reply #26 Wednesday, March 24, 2004 8:01 PM
re: #25
On the issue of using tile vs. stretch, I agree with you absolutely. In fact I have seen skins where appearance improved as well as performance.
I am not saying that we shouldn't do those things we can do routinely to improve the performance of a skin. The point I take issue with is reducing color depth. I honestly don't believe that the performance gain for today's typical desktop user is significant enough to warrant the time, effort, and potential risk of an adverse affect on appearance.
On the issue of using tile vs. stretch, I agree with you absolutely. In fact I have seen skins where appearance improved as well as performance.
I am not saying that we shouldn't do those things we can do routinely to improve the performance of a skin. The point I take issue with is reducing color depth. I honestly don't believe that the performance gain for today's typical desktop user is significant enough to warrant the time, effort, and potential risk of an adverse affect on appearance.
Sput
Reply #27 Wednesday, March 24, 2004 8:33 PM
Reply #27 Wednesday, March 24, 2004 8:33 PM
SD - for the most part I agree...there are some - the only thing I'd add is that there are some bits of skins which tend to be so small and simple that there's almost no risk of loss of quality, and worth while aiming to create from the outset as 8bit - such as Menu Item Extras...and most control elements actually.
Window frames etc set to tile I'm in favour of as it will tend to look better, though it can take some creative thinking if you want something more complicated than a plain background...particularly challenging to make a seemless tile if there's a distinct pattern and finding a way to "hide the join" for variable join points in the pattern.
That said, I'd still make no commitment to get everything optimised as sometimes it's hard to remember when the challenge you're concentrating on is making it look right.
Window frames etc set to tile I'm in favour of as it will tend to look better, though it can take some creative thinking if you want something more complicated than a plain background...particularly challenging to make a seemless tile if there's a distinct pattern and finding a way to "hide the join" for variable join points in the pattern.
That said, I'd still make no commitment to get everything optimised as sometimes it's hard to remember when the challenge you're concentrating on is making it look right.
GreenReaper
Reply #28 Wednesday, March 24, 2004 8:47 PM
Reply #28 Wednesday, March 24, 2004 8:47 PM
When optimising it often pays off to consider where it's worth spending time on it and where it's not. Look at what the largest files are, or the simplest to convert to lower bit levels, or where it's easy to tile rather than stretch or convert an image to have no transparency/translucency. You can get big wins for small amounts of effort.
Be aware that images will be converted to the target display's colour depth when loaded, so there is not likely to be much change in the memory usage - to do that, you need to shrink the images, have smaller areas for tiling, etc.
Be aware that images will be converted to the target display's colour depth when loaded, so there is not likely to be much change in the memory usage - to do that, you need to shrink the images, have smaller areas for tiling, etc.
chadlachlanross
Reply #29 Wednesday, March 24, 2004 10:27 PM
Reply #29 Wednesday, March 24, 2004 10:27 PM
I used to compress all of my windowblinds skins, until the update to wb 3.5 I think it was, or 3.2... point being, that in one of the updates it just didn't seem to matter anymore if they were compressed graphics or not. and it seemed to me that it actually made things slower having them compressed.
General consensus seemed to agree. It was about that same time that all of the other "top" skinners stopped using compressed graphics as well.
General consensus seemed to agree. It was about that same time that all of the other "top" skinners stopped using compressed graphics as well.
chris lee
Reply #30 Thursday, March 25, 2004 7:50 AM
Reply #30 Thursday, March 25, 2004 7:50 AM
I'm pleased that the GreenReaper has got involved (the author of SkinBench)
I have just done some tests on my newer XP machine and my older 2000 Machine using skin bench, both are using WindowBlinds 4.2
Both machines are set to 1280 X 1024 resolution with True Color.
I am using two versons of XP Corona one with reduced colors and transparency and one without.
Here are the results in the areas GreenReaper says in the documentation are most likely to indicate performance:
(The higher the value the faster the skin)
XP Machine
MoveAbout
Normal: 747 SkinMarks
Optimized: 130 SkinMarks (SLOWER!)
Resize
Normal: 534 SkinMarks
Optimized: 130 SkinMarks (SLOWER!)
Windows 2000 Machine
MoveAbout
Normal: 213 SkinMarks
Optimized: 297 SkinMarks (FASTER!)
Resize
Normal: 162 SkinMarks
Optimized: 213 SkinMarks (FASTER!)
Ok so now I'm really confused, it would appear that when using Windows XP the Optimized skin is actually slower, but when using Windows 2000 the Optimized skin is faster. Does anyone have an explination for this one?
If i'm honest the XP machine is fast enough so that it is quite hard to tell which skin is actually faster but the optimized one still seems to 'feel' quicker, then again that does not fit in with the results.
On the 2000 machine I can tell when using the computer opening new windows and moving them around that the optimised skin is much faster, I'm not in any doubt here, and the statistics back me up.
So what's going on, is it a O/S related issue?
Have stopped optimizing because optimized skins are acutally slower on XP?
I have just done some tests on my newer XP machine and my older 2000 Machine using skin bench, both are using WindowBlinds 4.2
Both machines are set to 1280 X 1024 resolution with True Color.
I am using two versons of XP Corona one with reduced colors and transparency and one without.
Here are the results in the areas GreenReaper says in the documentation are most likely to indicate performance:
(The higher the value the faster the skin)
XP Machine
MoveAbout
Normal: 747 SkinMarks
Optimized: 130 SkinMarks (SLOWER!)
Resize
Normal: 534 SkinMarks
Optimized: 130 SkinMarks (SLOWER!)
Windows 2000 Machine
MoveAbout
Normal: 213 SkinMarks
Optimized: 297 SkinMarks (FASTER!)
Resize
Normal: 162 SkinMarks
Optimized: 213 SkinMarks (FASTER!)
Ok so now I'm really confused, it would appear that when using Windows XP the Optimized skin is actually slower, but when using Windows 2000 the Optimized skin is faster. Does anyone have an explination for this one?
If i'm honest the XP machine is fast enough so that it is quite hard to tell which skin is actually faster but the optimized one still seems to 'feel' quicker, then again that does not fit in with the results.
On the 2000 machine I can tell when using the computer opening new windows and moving them around that the optimised skin is much faster, I'm not in any doubt here, and the statistics back me up.
So what's going on, is it a O/S related issue?
Have stopped optimizing because optimized skins are acutally slower on XP?
Sput
Reply #31 Thursday, March 25, 2004 7:39 PM
Reply #31 Thursday, March 25, 2004 7:39 PM
Chris Lee - can I suggest that it's probably not even as simple a things as 2000/XP as the difference...it's probably down to video drivers, and WB being able to make use of Hyperpaint would be my guess. I don't know how this technology works, but I would assume that it is using texture space on the video card ram to re-paint the windows...it could be then that this only works when the graphics are at 24 (or 32) bit...so "optimised" graphics resort to non-hyperpaint re-draws?
Just a theory...now, if you had Win2K and XP on identical hardware and the graphics drivers could be somehow ruled out, then it could be shown to be down to the OS directly.
In a way, this result is good news for Skin designers though as thinking about optimisation is a distraction from thinking about the look of the skin - and as Frogboy has been encouraging more creativity, then that has to be good
Just a theory...now, if you had Win2K and XP on identical hardware and the graphics drivers could be somehow ruled out, then it could be shown to be down to the OS directly.
In a way, this result is good news for Skin designers though as thinking about optimisation is a distraction from thinking about the look of the skin - and as Frogboy has been encouraging more creativity, then that has to be good
Please login to comment and/or vote for this skin.
Welcome Guest! Please take the time to register with us.
There are many great features available to you once you register, including:
- Richer content, access to many features that are disabled for guests like commenting on the forums and downloading skins.
- Access to a great community, with a massive database of many, many areas of interest.
- Access to contests & subscription offers like exclusive emails.
- It's simple, and FREE!
|
|
Page 2 of 2 | |
|
|---|
Reply #21 Wednesday, March 24, 2004 3:13 AM