problems w/ samples loaded into ram

[frugalpole]

ok ... so here is my scenerio.

1.67 15" powerbook (august 2005 model)
1 gig of ram (should be 1.5, but the lower ram slot died. the logic board needs replacing)

running live 5.0.3.
-my buffer size is 256 samples.
-triggering w/ the trigger finger
-w/ a plugin buffer of 128 (not sure this matters...)
-w/ all of my samples loaded to RAM. (total size: 4.5mb)
- w/ quantization set to NONE on EVERY clip.

reason 3.x rewired through.
----

the problem that i am encountering is a sort of lag. the audio file seems to fatigue a bit, making me think that live's 120bpm's and the click's 120 were not the same. (they were not sync'd).

but after i worked on it some more, and tried some more things it seemed that it was processor or RAM access' that caused a sort of momentary hiccup. for example if i would play my maelstrom patch it would occasionally happen. or if i would retrigger a sample rapidly it would again, on occasion happen.

i want to put together a stable solution for running loops/sequences/sampling and synths in the live setting, and i am nervous because live isn't acting as stable as i had anticipated. up to this point (2 years, since version 3)i had run everything quantized. the drummer would lock on to live's click... and everything was beautiful. but the problem has started as i have explored using live as hardware sampler of sorts. loading my samples into ram... triggering unqunatitized, etc.

so i am unsure on what to do. i wonder if hardware needs to be my solution. a device like the mpc1000 or (2500). or perhaps picking up like a dave smith evolver or poly evolver... to offload cpu strain on my computer.

anyone have any suggestions or direction on how i can solve my dilema?
anyone running live in the live setting as i am describing? is it stable for you?
any other details anyone needs?

again the problem happened occasionally during sudden aggressive strain on either the computers processor or the RAM. but only occasionally. weird.