If you're working with a large file- mine are "only" about 50 megabytes but I have several of them to look at at once- it would be useful if Origin didn't load all the data into memory. If you have 6 million points, there's no way they can be plotted on the same graph- most will be missed. So why can Origin not simply calculate how many datapoints to display (this can be user defined, in the same way as the speed mode currently works) and load just the required datapoints from the file, using seek and read operations? If the user then wished to zoom in, you repeat the process to get a set of datapoints from one region. This could be enhanced by giving the option to average in the missed data rather than decimating it- again with the user deciding if the extra time is worthwhile.

For small datasets this is pointless because the disk access time will slow things down, but where you're at the point where all the data is being paged in and out, it should give a speed increase.

It would also be useful if you could save links to datafiles, rather than embedding them. I got half way through some processing, then got close to my 1 gig pagefile limit, and when I tried to save the file ran out of disk space- the project was getting on for 700 megs.