I have Windows 7, 16G RAM and 8192M set in CT's Preferences>Memory.
At the beginning everything was ok, but some time ago I realised that Windows' Task Managers reports ~2G only. (The same effect with the previous CT versions and the newest 10.0.1.).
This is real problem for me, as I translate patent documents, with looong sentences, an CT, with such small memory, slows down, or even hang on in the case of long segments.:-)
Your Windows 7 may tell that you are using an older computer with an older type of the processor. Long segments need more processing power to analyze and find matches. In this case, I would recommend performing the preliminary memory matching (see it in the menu Translate). You don't have to wait until it completes. Just give it the head start and start translating. Then, the matching will take place in advance (usually several segments ahead of your own translation pace).
Thank you, Igor - I will try to use the advice when I encounter the long segment.
And what about the memory size issue? I feel that expanding it to 8G made the problem of long segments weaker.
You can check the actual usage of Java memory in CafeTran's Help > About panel. The program uses as much as currently needed from the assigned Java memory.