The paper presents an interesting and comprehensive study on the load and performance of three floating off-shore wind turbines, considering a very wide range of operating conditions, constructed by resorting to a well documented data base of ocean conditions (and already used in previous studies). Three different HAWTs are considered, each one featuring a different floating foundation system. Three different simulations packages, covering two main fidelity levels (BEM and LLFVW), are used to simulate the movement and the load acting on the turbines. 

The paper highlights both the effects of the floating operation on the turbine behavior and load as well as the capabilities of the codes to predict such behaviors. In the end, the paper documents a study which is fully relevant for the WES journal and certainly of archival value.

Despite the general positive evaluation of the paper, in my opinion the document could be improved. In the following I provide some remarks that the authors may consider in their review.

 1. In general, the paper is difficult to read; it is very long and there is an exaggerated use of acronyms, which complicates the comprehension of the text in several instances. Some sections may also be divided in subsections or, at least, to be divided in more paragraphs. I invite the authors to find a way to make the discussion more synthetic and more effective or schematic, eliminating what is not strictly necessary for the paper (maybe referring to the reports of FLOATECH project). In the results section of the paper, for example, most of the comments are descriptions of the figures and of the differences between the codes, and less space is dedicated to the explanation of the observed differences. 

2. The conclusions are, in some aspects, weak. For example, the self-excitation phenomenon (one of the most interesting findings of the paper) is predicted in different way by the codes, but the explanation of the over-estimate of Q-Blade code is demanded to another paper (Behrens de Luna et al.), still unpublished; it could have been a relevant discussion and conclusion also of the present paper. As a second example, the results often show differences between the predictions of OpenFAST and DeepLines Wind, despite the two codes are of the same fidelity level; the explanation of these results are demanded to future studies, while one would have expected such discussion in this paper, for example by resorting to the different structural model of OpenFAST or to the compiling/import issues of DeepLines Wind mentioned in the paper (and see below for the remarks on these aspects).

Detailed comments:

 - In section 3.2, the reason for using a simplified structural model in OpenFAST is not convincing; if one wants to see the effect of using a simplified structural model, a comparative analysis must be done changing this particular model within the same simulation framework, and not introducing a further variability in a context of code-to-code comparison. The implications of this choice should be better highlighted in  the results section.

 - At the beginning of Section 4, it is mentioned that not all the simulation runs reached convergence in OpenFAST or DeepLine Wind. While this might be expected, it would be interesting to briefly discuss what are the reasons for these failed convergences.

 - In section 4.1, line 307, a compilation issue is mentioned in DeepLines Wind, which has a relevant impact on the results. It is not clear, for the reader, what could be the practical consequences of this issue: if the compilation issue could be easily solved by the Authors, the technical relevance of the results obtained with DeepLines Wind is questionable; if, instead, the issue is an inherent feature of the code, a general improvement to the code is needed. Please explain.

- In section 4.2, line 374, an issue of DeepLines Wind in importing the wind field is mentioned. Again, is it an inherent limitation of the code with respect the other ones, or was it a simple issue in the set-up that could have been solved by the Authors?

 - Figure 10 (and following similar ones): personally I do not fully appreciate the use of 'cumulative PSD' in the plots, but I recognize its effectiveness with respect to the standard PSD diagram; for clarify, I recommend to specify in the paper, when commenting Figure 1 for the first time, how to read a cumulative PSD diagram.

 - Figure 20: a significant difference exists between the Q-blade and DeepLines Wind in the frequency range 0.4-0.6 Hz, the authors are encouraged to comment this aspect in the paper.

The manuscript discusses the performance of floating offshore wind turbines. For that, the differences in modelling these facilities are studied using three different simulation codes and three substructure designs and considering realistic environmental conditions. Furthermore, the study is made within the EU project floatech. The authors find that while all codes provide similar results for the system dynamics, differences are observed in the estimation of fatigue loads.

I find the paper well written and the analysis sound for the offshore wind turbine community. I therefore consider it is suitable for publication in Wind Energy Science journal, provided the authors address the following minor comments:

- The environmental conditions used for the present work correspond to the Scottish island of Barra. The authors argument this choice because, among other arguments, this location presents harsh conditions. The question arises therefore about how representative is this study of more standard offshore environmental conditions. I would appreciate if the authors can address this point in the manuscript.

- Table 1 caption should describe the variables in more detail. Overall, while in another work from the authors this dataset is discussed, more detail about it would help the reader of the present manuscript.

- Why DLC 1.4 simulations are only 10 minutes long? What is different with respect to the other ones?

- In figure 2, the turbine models from left to right do not agree with the order in their description in the sections below. This is a very minor issue but the authors may want to correct the inconsistency.

- I agree with the other reviewer that the lack of convergence in some simulations in OpenFAST and DeepLine deserves a deeper discussion.

- While the manuscript reads well, it presents several typos that should be corrected.