Firstly, lets clarify something about artwork and resolution. File size is whats ie, its ultimately down to pixel count of your file. An 8X10 image at 300 dpi is the same filesize and ultimately resolves the same information as the same image at 4X5 @ 600 dpi. In other words, its the same number of pixels the exact same image, the same amount of data stored identically. The only relevance to the different denotation of the image size is that it sets the size for your RIP.
When you take an 4X5 image thats 300dpi and enlarge to an 8X10 image at 300dpi, you must perform re-sampling. This is because the grid of pixels that makes up your image is now 4 times larger in area. The software must now fill in information to occupy the new larger pixel grid based on the original file data. The
computer re-samples the image adding color information to the new pixels, remember for every 1 pixel in your starting image, the
computer has to add 4 new ones based off the color of that one pixel and the pixel surrounding it. Photoshop uses bi-cubic re-sampling which produces heinously ugly results. Photozoomer and my personal favorite Genuine Fractals use more sophisticated mathematical models to add in this new information. The resized files look waaayyy friggin' better. In general, bi-cubic re-sampling results in alot of chroma noise as the color model does a relatively crude average of the surrounding RGB info, the plug ins produce enlargements that looks smoother and alot less noisy. Noise Ninja is a great plug-in to run after re-sampling to help further tame the noise added by the re-sampling process.
Boogie is also correct, if you use a factor common to the various output resolutions found on your machine (generally 12) your RIPs will go faster and your prints will look better. This is because the rip has to do generally less work interpolating the image to meet its native print resolutions. For example, if you have a machine that prints at 720 dpi, you'll get a generally better result if the image you rip is 360 dpi vs 400 dpi even though the 400 dpi image ultimately resolves more data! This is because the pixel grid is easily translated to the dots of ink fired by the print head. Simply put, its easier to make 4 pixels 8 dots than it is to make 5 pixels 8 dots. The
computer will have to compromise on the values of said pixels to best represent the input data where as when you work within the multiples of the native resolution, it simply multiply the quantity of any give pixel by (in this case) 2 on both axis.
Finally, working size in illustrator does matter, it shouldn't but it does. Working in a smaller scale for what ever reason in illustrator has benefits in terms of speed even while working with pure vector data. This is a product of crummy programming. If you use a spline based cad application, the only thing that impacts performance is the number of vertices, the complexity of the curve interpolation and the level of detail you choose to resolve to on your screen. Vectors are mathematical representations of lines and shapes (remember Y=X squared makes a smile!) and the complexity of the math should be the only determinate factor of load placed on the cpu. For some reason when illustrator is worked on with a big project and you set your art board at 1:1 scale, things slow down more than they would with the same vector data scaled down. It makes no sense, but it happens in practice (thus far in my experience with cad and graphics software this issue is pretty unique to illustrator). Also, vectors are cpu number crunchers and raster eats ram, so spec your system to your bias. Don't think that doubling your ram is going to impact vector editing beyond helping cope with the excessive bloat of the software itself. It won't make your machine crunch the vectors any faster. Same goes for RIPping, the large quantity of ram is helpful in that it lets you store the data to be ripped in ram vs spooling it off the hard disk (the FSB speed is wwaaaay fastater than the SATA, IDE SAS bus) but the real work is done by the CPU. Also, most rips aren't multi-threaded in the sense that it can rip a single file using multiple cores or chips. Most will apportion a core or cpu to each rip job. So if your a 1 job at a time guy, a really expensive quad core xeon isn't going to help much. In fact, a 3.6ghz p4 extreme might be a better route.
