Tell us what is your "nonspecific adsorption"?

If you are running crude samples, like cell lysate or hybridomas or sera samples or have general non-specific challenges, I recommend that you use the LNB-chip and the A200 biosensor from Attana. Attana has developed the qcm-technology in order to meat this challanges.

You can try to wash fast (high flow rate)

More details on your specific system of interest that you are studying will provide better answers. But without knowing the details, I can say that for proteins, BSA (up to 3%) in the assay/running buffer is often used in systems from different companies to lower NSB. Small amounts of non-ionic detergent (Tween, Triton, etc.) also can be of help if they are compatible with your system. Anything that has been used to reduce NSB in other bi-molecular assays like ELISA could work in your system. If you are having issues with NSB binding with small hydrophobic molecules then increasing or including cosolvent (DMSO/MeOH) can be of help, so long as you can correct for the signal response from these cosolvents. There are other things to explore with particular chips and immobilization strategies but I would need more details about the system.

To generally avoid nonspecific binding on Au surfaces try to flush the surface with BSA before performing your final (interaction) measurement step.

In case you build up your own sensor surface architecture starting from pure Au surfaces, try to flush the surface with a short chained thiol compound or cystein, resp.. This will close open Au spots and help to avoid unwanted contacts between analyte and sensor surface.

It really depends on the source of you NSB. Frequently, the NSB is because your analyte has high pI, and therefore at the pH of the running buffer it will be positively charged, and therefore electrostatically attracted to the negative surface. Simple solution will be to elevate the pH of the running buffer (as long as your interaction still works at this pH). Also, adding more salt to the running buffer might help. We found that MgCl2 (up to 100mM) does better job in reducing non-specific binding than NaCl (again - should be checked that not interfering with the interaction).

simply by using appropriate blocking agent like BSA or gelatine or may be by using some non-ionic detergent in the buffer solution

BSA, gelatin and casein are common blocking reagents used to prevent non-specific binding. However when using a complex biological sample such as serum, which contains a lot of protein these blocking solution might not be enough. Thiol terminated PEG have become quite popular for reducing non-specific binding especially because they form SAM on the metal coated sensor chip which also provides functional groups for surface immobilization. (Bolduc OR, Clouthier CM, Pelletier JN, Masson JF*, Peptide self-assembled monolayers for label-free and unamplified SPR biosensing in crude cell lysate, Anal Chem, 2009, 81, 6779-6788 for example)

I wrote a white paper that examines non-specific binding in SPR systems. It explores the many of the ideas that have been listed.

https://www.nicoyalife.com/wp-content/uploads/2015/11/Reducing-Non-Specific-Binding-in-Surface-Plasmon-Resonance-Experiments.pdf

Addind a bit of tween in the buffer may help, or try to optimize the buffer in function of the binding propoerties of the molecules you are studying. Choosing a proper reference can also greatly help; if you are sure you properly mimic your substrate except that it is not functionalized, you can subtract with more confidence your curves from the reference.

Anyway the problem (to what I think I know) is that non specific binding is due to the need for studying low affinity binding, which leads to the use of high concentrations of analyte. It is kind of a limitation of the technique. 

Incubation of analyte with carboxy methyl dextran solution ( i.e chip matrix) shown significant result in reducing non specific interaction in reference channel or with sensor chip