Every time your PC folds a protein, it gives them a "Simulation" of how the protein would react in a certain situation. Because proteins act diferently many times, you need to run many, many simulations over and over again to see the probability of something happening. Because there are so many proteins and each protein needs to be folded many times and in different situations, we have a extremely large amount of folding work to be done.
I'll give you an example: We know that if you rolled 2 dice, the most common number that would come up if you added them together would be the number "7". (3+4= 7...4+3=7...2+5=7...5+2=7...1+6=7...6+1=7) Does this mean that if you rolled a pair of dice, that you would get the number 7 every time? No... You could get a 10 first roll, a 3 the next, a 4 next and so on. You could probably roll 20 times and not see a number 7 at all. (I have seen this a few times luckily in Vegas...then I blew all my money later on...UGH) Anyway if you rolled a pair of dice 1,000 times, you would find that the number 7 would come up the most way more than any other number. Also, if you ran a computer simulation that had 2 dice rolling over and over again, you would find that 7 is the most common and that is pretty much what we do.
Folding is much the same way. The proteins do not fold the way we think they are supposed to every time. Like the pair of dice I spoke of, we need to run the simulation many times to find out what the most common way a a certain protein folds and also, what happens to it when it "MisFolds".