View Full Version : Image quality comparison

5th July 2001, 23:28
No, no, wait! I'm not off-topic :)

As you may (or may not) know, all the images on seti.matroxusers.com are pre-computed on my own computer and then uploaded to the server.

This is becoming a problem in several ways:

- It's a waste of bandwidth (~10 MB is uploaded every hour just for seti@murc; the other sizes are comparable, but the uploads are less frequent). Personally, I can live with that. My employer is connected at > 1 Gb/s so I'm not really worried. What I don't know though is whether Ant will have to pay for the FTP-upload traffic.

- It's also becoming a bit of a burden on my (aging) workstation... Computing the images takes away about 10 percent of (one of) my workstation's CPU/memory. This is not only bad for my crunching statistics ;), but also makes my computer rather unusable for a considerable amount of time.

Consider the above a prelude to my real question...

I've cooked up an alternative solution... I've made a script that can run on the server and will generate the images "on-the-fly". I would just have to upload a small datafile (I have to do that anyway) and the images are created from that.

Right now, the script is only for seti@murc. What I'd like y'all to do is to compare the "static" images with the "dynamic" ones. I see two possible problems:

- The images are perhaps not quite as nice. I can't use the fonts that I'm using for the static images. On the other hand, it would now be feasible to enlarge the dynamic images (it would take too much bandwidth for the static ones). To make for a fair comparison, the dynamic image is now just as large as the static one.

- Maybe the computation of the images will have an impact on server performance. It's difficult for me to judge, but I don't think so (it only takes a fraction of a second, and just when someone requests an image). Maybe Ant can comment on this.

Boy, this post is getting long...

Short summary: please compare this (static):


with this (dynamic):


and tell me what you think.


6th July 2001, 03:37
Can't see the diff (have an old ati instaled in this comp)! Use whatever you want to! :)

6th July 2001, 06:36
Hey Martin,

you forgot one option for voting:

I see a difference, but I don't care which one to use ... ;)

Honestly, the dynamic version looks a lot sharper than the static one and if you could disable the bold text and use regular instead, I'm more than just happy ... :)


6th July 2001, 06:46
I prefer the dynamic one, looks a bit sharper to me. As for the ftp upload issue I believe that will form part of my monthly bandwidth allowance so all the more reason for the dynamics ;)

6th July 2001, 11:14
61 wievs 9 votes!!!!! This suxs big time! Come on you guys if you have an opinion on the subject please vote!

6th July 2001, 11:39
Well, it looks like Ant has decided the issue right there. If it costs you anything: dynamic it is. Seems to be liked better as well, so that's perfect...

Right now, only seti@murc gets the dynamic version. I've scaled the images up a bit, and removed the bold text. Would this be about right?

Thanks for the feedback guys.


6th July 2001, 11:52
Ah, so I voted for the dynamic ones, and won't get them yet at genome :(

Might be something for the future :)
The bold text made them better to read though, Martin.


7th July 2001, 02:13
> Might be something for the future :)

The very near future... I've just converted Seti clubs&teams to the new images. Genome will follow today or tomorrow.

> The bold text made them better to read though, Martin.

Well, I can't really decide which is better. On the one hand, it's a bit easier on the eyes without the bold fonts. On the other hand, I think I agree it was better to read. I'll see...


7th July 2001, 08:42
All systems go :)

I hope they all work. I'll probably tweak some things over the next few days. Feedback (as always) welcome...


Liquid Snake
7th July 2001, 10:30
Actually, I think the image is a little too big. The old graphs were a little too small. The text of the dynamic ones is fine, though.

8th July 2001, 04:17
All looks good to me!

However, I'm afraid the slope on my particular plot is a little to close to the horizontal, anyone wanna donate some CPU cycles? ;)

Helps my progress if I notice a machine hasn't been crunching for 2 days though...

Incidentally, I already thought the graphs were done server side - thanks again for these stats though, they still amaze me :)