Geoffrey_Coram
|
There's some overhead involved in creating the various threads and splitting up the work; I guess Cadence has decided that one generally gets a *slower* simulation with multi-threading unless the circuit is "big enough."
Do you have access to APS? I think Cadence advertises that as having better multi-threading.
Do you need one long tran? Or are you computing the response to a long set of inputs, such that you could split it up into 10 simulations, each of which takes 1/10 of the set of input transitions?
I don't think I can give you any further suggestions without understanding why you need very accurate results for such a long transient.
|