Samples come in 2 forms - multisamples and phrases.
Phrase is what most peopl are familiar with. A phrase is recorded (could be an electronic 2 bar phrase, a drum fill or a short singing phrase by a diva). They don't usually go for very long - perhaps a few seconds or a few bars. Reason is that it takes up memory space. This recorded phrase can then be played on a keyboard. Each key can either me mapped to one phrase but at different keys +/- different tempo, or each key of the keyboard can be mapped to a different phrase. The latter of course will take up more memory. There are hardware that does these, where phrases are triggered by pads and buttons rather than keys of a keyboard. This can be either looped (of which the phrase keeps repeating itself), or un-looped.
Multisamples are when an instrument is sampled note by note. For hardware samplers, because of memory limitations, some instruments are not sampled chromatically (ie note by note). However, with softsamplers, limitations of memory had been removed. We now use samples that are: 1. chromatically sampled (note by note), 2. not looped (ie one piano note is sampled until it decays naturally, a flute player holds the note until he runs out of breath - so it is more realistic), 3. multiple velocities per note (from ppp to fff - this is a difference between simply adjusting the volume and actually recording at different volume as the timbre actually changes with different loudness), 4. different articulations per note (eg for strings - you have different playing styles like spiccato, staccato, detache, marcato, legato, pizzicato, tremelo, thrills etc).
For example, my piano itself is a 2GB, 36 samples per note piano. No hardware can take that size of piano. It is 12 samples with pedal up, 12 samples with pedal down, and 12 release samples (the samples when you actually release a piano key). Plus each not is sampled un-looped.
The idea of softsamplers is realism.
