Choosing Methods Based on the Number of XYZ Data Points
The size of your input data set should be considered when selecting a gridding method. For example, some gridding methods interpret small data sets more effectively than others do. Surfer requires a minimum of three X, Y, and Z points to perform the gridding process.

Ten or fewer points are not enough to define more than a general trend in your data. Triangulation with Linear Interpolation and Moving Average are not effective with few points. As with most data sets, Kriging and Radial Basis Function methods will produce the best representation of your data in this situation. If you want only to define the trend of the data, you can use Polynomial Regression. With 10 or fewer points, gridding is extremely fast, so you might want to try the different methods to determine the most effective method for your data.

With small data sets (<250 observations), Kriging with the default linear variogram, or Radial Basis Function with the multiquadric function produce good representations of most data sets.

With moderatesized data sets (from 250 to 1000 observations), Triangulation with Linear Interpolation is fast and creates a good representation of your data. Although Kriging or Radial Basis Function generate the grids more slowly, they also produce good data representations.

For large data sets (>1000 observations), both Minimum Curvature and Triangulation with Linear Interpolation are quite fast, and both produce good representations. As with most other data sets, Kriging or Radial Basis Function probably produce the best maps but are quite a bit slower.

Using Kriging or Radial Basis Function with large data sets does not result in significantly different gridding times. For example, if your data file contains 3,000 or 30,000 data points, the gridding time with Kriging and Radial Basis Function is not significantly different. Either data set (i.e. 3,000 or 30,000 data points) might take a considerable amount of time to grid, but the two gridding methods (i.e. Kriging and Radial Basis Function) will take approximately the same amount of time.
See Also
General Gridding Recommendation
Creating a Grid File from an XYZ Data File
Exact and Smoothing Interpolators