Two-Dimensional Sediment Transport models can take a long time to run.  

MethodDescription/RecommendationComments
Optimize Hydraulic Runtime

Align cells with flow.
Consider a channel cell size that is 1/5th to 1/7th the channel width and larger cells in the floodplains.
Try to keep cell sizes relatively uniform (very small cells will drive stable time step)
Apply hydraulic best practices to make the model hydraulically stable before you add sediment.
Choose a stable time step
    -consider the automated time step selection tool

A large time step that iterates can take longer to run than a short time step that is stable.
       -Find and work on iterating cells
       -Avoid thin slivers of flow through cells when possible

Limit the Number of Grain Classes

Try to limit the model to 3-5 grain classes, at least in early, exploratory simulations.
Be careful to use the same, targeted, grain classes in the bed and boundary gradations.		Runtime increases approximately linearly with grain classes
A model with 5 grain classes will take ~5X as long as the same model with 1 grain class
2D models tend to use fewer grain classes than 1D sediment models.
In fact, this kind of gap-graded approach is discouraged in 1D sediment modeling.
Note: Skipping grain classes is NOT recommended in 1D models but best practice in 2D

Matrix Solvers 

Set Sediment Matrix Solver to FGMRES-SOR, at least for early exploratory runs and the SOR solver for the 2D hydraulic solver.  

HEC-RAS uses the Paradiso solver by default. 
Paridiso is more robust (slightly more accurate and stable), but it takes longer to run than SOR or FGMRES-SOR. 
SOR and FGMRES-SOR run faster without much loss of accuracy.

Sediment Time Step Multiplier

Water surface elevations respond to river changes faster than bed elevations or gradations.
HEC-RAS time steps are generally selected for hydraulic stability.
Larger time steps might be stable for sediment.
This feature directs RAS to compute several hydraulic time steps between each sediment computation.  
Factors of 2, 4, or 5 are common.

Because sediment computations tend to be more expensive than hydraulic computations, doubling this factor can almost halve the run time.
However, if the bed does change quickly, this factor delays the sediment feedbacks, and can decrease stability.

Morphological Acceleration

The Morphological Acceleration feature scales bed change during a sediment transport simulation.
It increases the morphological response, allowing modelers to run a shorter time series.
(e.g. A 30 day event could be compressed to a 10 day event with morphological acceleration set to 3).
Morphological Acceleration Factors <25 are recommended.

Morphological acceleration introduces more approximation to sediment model results.
This feature can also make the model less stable.
The flow time step does not update automatically.
Users must change the time step of their boundary conditions to compensate.
For example, if you have a 24 hour flow time step and use a morphological acceleration factor of 4, change your flow time step to 6 hrs.
More guidance is available here.