@victoria99 My apologies, Victoria. It was very late when I was figuring your memory situation out, and I was going off of your machine's spec sheet instead of what the system actually reported when you provided the screenshot of its memory banks. I suppose I wasn't paying attention to the fact that despite what public knowledge dictates, your machine does in fact support PC2700 memory (the system reports it as PC2600 for some reason; mine does that as well).
Taking that into account, here's the cheapest listing I could find for PC3200 memory, which although natively runs at 400 MHz instead of 266 or 333 MHz, will automatically clock down to 333 MHz, making it equivalent to PC2700 memory. I will now explain why in the following paragraph:
Find many great new & used options and get the best deals for HYS64D64320GU 1GB (2 x 512MB) DDR 400 PC3200U Memory Ram at the best online prices at eBay! Free shipping for many products!
www.ebay.com
If a memory module, even if it's just one out of four, naturally runs at a slower speed than what the system can support (in this case PC2100), then the system will just slow down the other memory modules to match the slowest one's pace (for example, it will run PC2700 modules at PC2100 speed). Now, that doesn't necessarily mean that you'll notice the OS (Operating System) get slower too, it just means that the system isn't using its memory quite as fast as it could be.
On the other hand, if you install faster memory than what the system can support, in this case PC3200 modules, the system will simply slow down the modules to match its maximum supported speed, in this case being PC2700. This is possible because PC2100, PC2700, and PC3200 modules are all part of the DDR (or DDR1) memory family. Interchanging PC2100 modules with PC3200 slots and vice versa is completely safe because the computer automatically handles the timing and clockspeed configuration. They are electrically compatible.
However, DDR2 is a big step up from DDR, and therefore uses different electrical signaling entirely. While the same speed scaling functionality exists for DDR2 modules in DDR2 slots as well, if you were to hypothetically insert a DDR2 module into a DDR1 slot, the modules would fry the board and potentially damage themselves in the process. But thankfully, this is physically impossible to do because the module designers made memory sticks only able to be inserted into slots designed to accept them by placing plastic notches into special locations in the slots, corresponding to the gaps in compatible modules, or at least for DDR1 and DDR2, anyway.
DDR2 support was first introduced with the Late 2005 Power Mac G5s and PowerBook G4s. Most all Macs before then used DDR memory, and before that in pre-2002 used non-DDR SDRAM.
Yes, the GPU is effectively a special type of CPU designed for graphical workloads. It works in tandem with the main CPU to handle pixels, polygons, vertexes, textures, and many other parts of 2D and 3D graphical rendering that the main CPU is not so good at, because the main CPU is already busy dealing with everything else under the hood, or that is to say, what the user cannot see on screen. Every time you plug in the monitor's VGA or DVI connector (the wide plug with lots of little pins on the end) into the computer, you are plugging it directly into the GPU (also known as the video or graphics card), so that it has a direct connection to the monitor, and can work more efficiently than if it was routed through the motherboard.
In a nutshell, VRAM is the GPU's own personal RAM bank. When the main system RAM runs out of space, it starts using a small scratch space on the hard disk for RAM duties, which is a lot slower than just using the dedicated RAM and is why you usually want as much RAM as possible (up to a certain point) so that it doesn't have to fall back to that. VRAM works the same way, because while the system uses regular RAM to temporarily store file data, computational data, and operation data, the GPU uses its VRAM to store textures, bitmaps, and raster data.
When the GPU runs out of VRAM during a graphically-intensive job or moment, it has to start using the regular system RAM instead, which is, while faster than the hard disk, still slower than using the dedicated VRAM. So instead of having to fallback to a slower medium, that is why you also want as much VRAM as possible up to a certain point (for most modern 2D usages, 256 MB to 512 MB is plenty, and you can usually even get by with 64 MB to 128 MB well enough too).
What EveryMac is referring to when they list "32 MB" is the VRAM capacity of the default GPU that the system shipped with, in this case being the GeForce4 MX graphics card. There is no consequence to having higher amounts of VRAM, because then you reduce the chance that the GPU will ever need to start using regular RAM and thus slow the system down. That being said, VRAM capacity alone does not determine whether one graphics card is inherently faster than another, because there are more factors at play, like clock speed, memory speed, and pipeline size. In this case however, the above linked graphics card is superior in performance to the existing GPU (GeForce4 MX) in most every way.
And, CPUs have their own dedicated "RAM" too, in the form of Level 1, 2, and sometimes Level 3 cache, abbreviated to L1, L2, and L3 respectively. But that's a separate topic.
Anyway, I hope that was helpful.