Sunday, May 10, 2015

Pimp My QuickSilver - Part II: RAM

Welcome to Part II of my series on upgrading a G4 Quicksilver (QS) to its maximum capabilities. See Part I here. So a few of the hardware upgrades arrived including the 1.5 GB of RAM, SuperDrive, and nVidia 7800 GS graphics card. Before I could really begin this adventure, I wanted to reload the Quicksilver in its current condition with fresh OS installs with a little bit of customization done for each. That's where the battle began.

First, both 200 GB WD IDE drives failed to be recognized by any installer or currently running system of mine. When they would be recognized, it caused all sorts of modprobe errors and slowed the installer considerably, especially when the installer reached the disk partitioning step (of course). I found two matching 40 GB Quantam Fireball 3 drives that did work though and successfully swapped them out after zeroing out both disks using Disk Utility in OS X. Immediately after that, I went to start the install of Debian, but quickly learned that my install CD was not being read properly.  Turns out it wasn't the CD, but the actual CD/DVD drive itself.  I pulled the working one from my identical Quicksilver to quickly resolve that issue.

Not only that, but starting the OS X installer off of my Leopard DVD had to be done in Open Firmware (OF) off of an external DVD drive.  Here is the command I needed to run with the DVD drive plugged into the rightside on-board USB 1.1 port.  The left would be usb1.
boot usb0/disk@1:,\System\Library\CoreService\BootX

If there were multiple partitions on the install DVD, I'd include the partition number the installer was located on after the colon in the above command.

The first 40 GB Fireball HD will house OS X Leopard and the second was going to be home to Debian Wheezy running LXDE. I chose Leopard because it was my first real experience with Mac OS X and I've never been able to grow very fond of OS X 10.4 Tiger. The caveat being that the CPU in my QS only ran at 800 MHz so I had to use the OF hack to fool the installer into thinking the CPU was running at 867 MHz, which happens to be minimum requirement for Leopard.  Of course, the change does not actually increase the clock frequency of the CPU, but changes what frequency it advertises itself as running at. The change to the CPU clock speed advertisement within OF is only good until the system is rebooted.  More information on how to do this can be found here.

Eventually, I also decided to just upgrade to Jessie for the Debian install.  As expected there were video issues as the G4 currently houses an ATI Radeon card.  I had to boot to rescue mode and install the linux-firmware-nonfree package before video would work.  I have yet to add any additional sabot parameters to assist related to the Radeon card, but haven't really found the need to either.  Time will tell.

So all of that and more caused quite a bit of delay in getting this put together.  Was it worth it and perhaps a little bit of fun? You betcha. But I now bring you the results of upgrading and maxing out the system's RAM.

As mentioned in Part I, the system started with 1 GB of RAM and I've upgraded it to 1.5 GB (it is maximum memory capacity).  I ran a series of tests to help gather some hopefully more realistic day-to-day numbers you would see from increasing the amount of available RAM.  I've also ran GeekBench against the system in OS X as well just for the heck of it to provide you with those results as well.

The tests include the following:
  1. Boot Time - probably not going to be affected a whole lot, but it should be nice to have these numbers available throughout the entire series.  By boot time I mean the time it takes for the system to boot to the login screen with and have the first typed character appear.
  2. Browser Launch Time - No add-ons installed running TenFourFox and IceWeasel 31.6.0.  I thought about installing uBlock and XMarks for syncing bookmarks but decided to keep it simple this time around. I'm measuring the amount of time it takes to click the browser icon and type the first letter in the Google search bar on the default home page.
  3. Maximum # of Browser Tabs - This one should be most interesting for my use.  I decided to use Twitter as it is an incredibly memory hungry website with each tab using anywhere from roughly 20 - 40 MB of memory apiece.  I continued to open new tabs and loading Twitter until the system was starved of memory and started to swap to the hard disk.  I was NOT signed in during my testing.
  4. GeekBench Results - Ran on the OS X side only as mentioned previously. I'm using version 2.2.7, the latest version available for our PPC machines. You can download it here
  5. Other Suggestions - If you have any suggestions, I will gladly run those as well and update this post with those numbers.  I'm positive their are better tests to run than what I've outlined above.
As most of us know, some of these will also be affected by some of the other system hardware such as CPU and system bus speed (133 MHz) which could be a significant bottleneck.  To keep track of memory usage in OS X, I'll stick with Activity Monitor and for Jessie I'll use the well-known CLI tool top. I wanted to use conky, but many of the characters appear garbled.  Not sure if it's a font issue or perhaps a screen depth issue? Reloading conky after logging in helps clear up some of the garbled mess, but not all of it.

That aside, let us move on to the memory install.  I don't think there really is a need to cover much here as the process to upgrade the RAM is about as straightforward as you could make it.  In fact it is so easy that it's not even a documented procedure for the QS over at iFixit's website. The QS has 3 slots for RAM, meaning each slot can house a 3.3 volt 168-pin 512 MB stick of PC 133 SDRAM.  The 133 indicating the speed of the RAM in Megahertz.  Not knowing enough about SDRAM in particular I took to the ever faithful Wikipedia to learn more about it.  Feel free to do the same. I happened to purchase 3 sticks myself from OWC.  I wanted to have a complete set of brand new RAM to run off of.

This particular RAM doesn't include heat spreaders, but I'll live with that for now at least.  I should be able to purchase those at a later time.  My question is does it really add that much benefit for this type of RAM? Or would it be more for form rather than function?  I'd be interested in hearing your thoughts.

I swapped out the RAM and filled all 3 slots, closed 'er up and watched it boot.  Here are my resulting numbers from each of the previously documented tests above.

Measured Task - Jessie (8.0) 1.0 GB RAM 1.5 GB RAM
Boot Time (in seconds) (88.22,83.57,83.64) AVG=85.14 (87.16,83.89,83.92) AVG=84.99
Browser Launch (in seconds) (7.82,8.31,8.08) AVG=8.07 (7.78,9.23,8.76) AVG=8.59
Number of Open Tabs 21 43

Measured Task - OS X Leopard 1.0 GB RAM 1.5 GB RAM
Boot Time (in seconds) (48.99,53.85,48.43) AVG=50.42 (53.67,51.70,56.07) AVG=53.81
Browser Launch (in seconds) (8.51,8.63,8.98) AVG=8.71 (9.28,8.80,9.06) AVG=9.05
Number of Open Tabs 24 43
GeekBench Score (OS X) 473 (Memory: 323) 473 (Memory: 323)

Most of the numbers you see above should be pretty self-explanatory. For the boot times, I noticed that booting to OS X after using Jessie or vice versa took about 15-20 seconds longer, so for each of the tests above, I made sure to start the boot up after having last used the OS I was currently testing.  My guess is that the last used startup disk is cached by OF somehow. The other significant difference for the boot times between Leopard and Jessie are that Jessie has to go through the necessary yaboot parameters to address Radeon driver issues which adds a good chuck more of time when booting up.

I want to quick note that it took forever to come up with the results of the maximum number of open Twitter tabs as the CPU became overworked after opening the second tab in both operating systems. :)  However, once Twitter loaded on each tab, the browser's responsiveness and stability would return.   Here is a screenshot of Activity Monitor when I hit 43 Twitter tabs in OS X with 1.5 GB of RAM.


The only other thing worth noting at this point is that the Geekbench score did not change.  In fact, the numbers were exactly the same through and through.  Makes me a little suspicious, but I"m not familiar enough with the application to really argue anything at this point.

That's it for this segment.  Not super exciting given that the system only maxes out at 1.5 GB of RAM, the upgrade is super simple, and the rest of the system is still quite the bottleneck at this point.  Part III will tackle the CD/DVD drive upgrade.  Stay tuned.

No comments:

Post a Comment