In "Design of Analog CMOS Integrated Circuits" by Behzad Razavi, the thermal noise of MOS is I2=4kTγgm
while in "The Design of CMOS Radio-Frequency Integrated Circuits" by Thomas H. Lee, the thermal noise of MOS is I2=4kTγgd0.
If the MOS work in saturate area, we can find that gm=gd0.  So these two functions is the same.  But if the MOS work in non-saturate area, the result is different.  What is the thermal noise of MOS in non-saturate area?

Thanks to didac.  I didn't see the footnote by carelessness.

There is another question.  As you said, in non-saturate area noise is equal to I2=4kTγgd0.  So the noise should be independent to Vds, which is not agree with the simulation result. (I use spectre to do noise simulation.  The process is TSMC 0.18um CMOS)