From what I see working at an Apple Reseller store here in Vancouver, MacBooks are currently the hottest selling computers by Apple. Despite the infamous random shutdown, despite the occasional overheating, and yellowing of the case, the sales are quite high.
Now, a lot of people don’t understand why you have to always have two identical RAM sticks in the MacBook. Especially those whiny PC users – they always want to buy one 1gb stick of RAM, and keep the other slot empty in case they want to upgrade later (pff). Well, I am a MacBook owner myself now, so I decided to find out what really happens if you just use one stick of RAM. (Spoiler: There are lots of explosions, naked girls, and Bill Gates jokes involved, so read on).
Me and a co-worker decided to conduct a benchmark after another customer started complaining about having to buy 2 x 512mb RAM sticks instead of a single gigabyte. Like all normal salesmen, I’ve read the pitch about the decreased performance and unusability of the MacBook with one stick, and that the customer should just get two or else… Or else…
After the guy left somewhat dissatisfied, me and the other dude from the staff decided that it’s time to seek facts to backup the statements we were brainwashed with. For those who don’t know or care to know, the MacBooks have an onboard Intel GMA950 graphics chip which has no video memory of its own. Instead, it uses a portion of the main RAM for its magic. From that, you can quite easily conclude (duh) that the faster the RAM is, the better. Having two sticks in at the same time is equivalent to having a mouth twice as wide – you could stuff it with two Big Macs at the same time, or roll a pizza up, and eat it like a burrito. Two identical sticks maximize the speed the RAM can be accessed at (Apple calls that a “contiguous array” of RAM, but that sounds like a good recipe for an instant nap).
Back at the store, I started the MacBook up, and loaded up Cinebench and xBench. I also have Quake 3, at all times, on all of my systems, but we decided that the 3D rendering and raytracing tests in Cinebench should be more than enough for a rough benchmark like this.
The system the tests were run on was a white 2.0 Ghz Macbook, with 60 gb of the stock 5400 RPM HDD, and with the 2 GB of Kingston KTA MB667 Mhz RAM. This is the RAM Apple recommends, and supposedly it has been tested more thoroughly compared to the crappy Value RAM everyone in the PC world has been successfully using for years.
The laptop was running off its battery at half of the brightness. I also have Windows installed for all my audio apps and miscellaneous stuff Mac OS X doesn’t have worthy equivalents for, yet. I haven’t tried comparing the performance of the MacBook under Windows, though I probably should – there are far more sophisticated hardware and software tests available, with the essential stuff boiling down to Winston, a few versions of 3D Mark, and PC Mark. Yeah, I actually defend Windows now, after finally making a switch to Mac OS X. I’ve always been kind of strange though, you know, in the head. I think my mom’s not telling me everything about my early childhood.
Now, back on the MacBook, we’ve run the two tests with both sticks in, and recorded the results. After all was done, I turned Krystal off (that’s her name), removed the battery and the clip inside, and took one of the sticks out. After putting the clip and the battery back in, and turning the computer on, it started just fine.
Immediately after OS X booted up, I noticed that the interface seemed slower and sluggish, reminiscent of running PearPC on Windows (yeah, been there, done that). I checked the Activity Monitor for unusual CPU load peaks and RAM usage, and everything seemed normal, with the CPU peaking at about 4-5%, and most of the RAM being available to the OS. Clearly, something was wrong with rendering the GUI. Or I was drunk. I don’t remember now..
Next, we ran both of the tests, and here is where the somewhat interesting part begins. Here are results side by side.
As you can see, in Cinebench there is almost no difference between the results, except for the hardware OpenGL tests. When the time came for the Intel graphics chip to really kick in, and load stuff into the shared space in the system RAM, the performance dropped down. While it might seem like an insignificant drop in speed for some, it was enough to impact Quartz quite noticeably (I made that a link to make you click on it – it’s super interesting geek stuff). The lack of difference in the other benchmarks can be explained by the fact that the Intel chip really puts the memory controller to a test, while the rest of the applications don’t take advantage of increased throughput as much.
xBench displays similar results.
While it may not seem as drastic at a first glance, the results very noticeably when it comes to the Quartz, OpenGL and memory tests. The other stuff is not affected much.
From all this, I, the benchmarker, can conclude that while the difference in performance between a single stick and a contiguous array of 2 sticks doesn’t seem as dramatic on paper, it is, in fact, large enough to affect the GUI in a way that drives you up the wall. It seriously feels like running OS X on a first generation iMac G3. With 16 Mb of RAM. Off floppies. Serious-lah. Oh, and no WoW for you. Hence, use two sticks, and don’t whine.