Here, let me revisit this point. The test machine is a 1Ghz Pentium III Dell Inspiron 8000, with a 60Gb 7200rpm hard drive as the system drive and a 64Mb Nvidia Geforce4 440 Go running Arch Linux. (All calculations done via Timex method.)
512Mb PC133:
Boot time from Grub to desktop: 27 seconds
Kazehakase start time: 2.54 seconds
Time to compile VICE 1.22 (make only, not configure): ~7 minutes 15 seconds
glxgears framerates: ~2175 fps
256Mb PC133:
Boot time from Grub to desktop: 0:27
Kazehakase start time: 0:2.47
Time to compile VICE 1.22: ~7:08
glxgears framerates: ~2180 fps
Start times are identical. System loads are identical. Compile times are identical. Responsiveness, shutdowns and disk access are identical. Network access is identical. Even the color of the screen is identical.
Therefore, more memory will not always magically speed up your computer. QED.
Hey K,
This is an interesting experiment. From the initial look, I’d say the speed of the memory was the most common factor. That, and the items tested might not have needed to use all the memory. Then again, I’m certainly no expert on memory.
My only argument with this specific experiment is the items tested, other than bootup. You’re much better at this stuff than me, at least for now, so I wonder if there’s a way you could do tests on similar hardware/memory configuration with:
-Time for Internet Browser to get to default page.
-How long to load various pages, especially news sites.
-How long to load a YouTube page.
-How long to startup OOoWriter (I know this can be adjusted, but the average person wouldn’t know how to do that)
-How long to do various things in GnuCash.
-How long to do various things in RhythmBox
What I’m considering here is, “What would the average person do, and does the amount of memory significantly affect those things?”
My guess is; There’s probably not a significant difference between 256 MB, and 512 MB, but I’d be willing to bet there would be a difference with 1028 MB of memory.
Then again, there probably is an minimum amout, where below that things do get bogged down, but any more than the needed amount doesn’t really help. Information like this would really help me in decision about how much memory to put in my FreekBox’s. I’ve been using hardinfo to give me some ideas, but it provides for more confusion that help.
I’d love to hear more on this practical experiment.
I don’t know if there’s a magic number for how much memory a system “needs.” I think a lot of it depends on what you do with the system.
I rarely, and I mean rarely, show a memory profile over 128Mb, and that’s after hours of surfing, working, typing, writing, watching movies and so forth.
So I kind of rebel against the knee-jerk reaction that tells everyone to put 2Gb into their system. I think that’s a byproduct of years under Windows, with incessant bloat and poor program design pushing everyone to buy more memory.
The purpose of the experiment was just to show that, for all common purposes, more memory doesn’t always make a machine “run faster.” Even compiling, which is a processor-hungry task, doesn’t go any faster with more memory.
If it was me and I had the choice of how much memory to put in a machine, I’d probably go with 256Mb. I really believe that’s enough for a lightweight desktop on an older machine with swap space for emergency. More than that is really overkill, in my book.
I don’t know if that helps at all, but for what I know about your project, I wouldn’t waste the second 256Mb on the same machine if I could give away another whole computer with that 256Mb in it.
I think I’ll stick with the 256MB installations, especially with Xubuntu. Today, I got in a Sony Vaio, PII, 64MB, 6GB hd. Ran puppy a little sluggish, but DSL just fine. I prefer puppy because of the ability to do the ‘pupfile’ thing. After I finish this post, I’m going to look through your posts for the lean install instructions.
Sounds good. Are you going to put it on your blog? I’d be interested to hear about it.
Don’t forget that DSL can be installed with the Debian framework, which makes it behave something like Ubuntu. I haven’t done that in a long time, but it might be more familiar than DSL’s in-house package management (which I always found somewhat clumsy).