The "best" depends probably on the context of usage. Amongst the more interesting ones (both, in terms of potential applications as concerns calculations, as well as graphique representations) with which I have worked myself are Pajek and UCINET. Also "R" ans "NetMiner" seem to be interesting tools for SNA.

I personally like UCINET as it has quite some sophisticated functionality and the UI is decent (also it is cheap). If you are into command-based system, R (I used this briefly_ is also a good option. Netminer is more commercial with really polished UI but I think they offer the same things as UNCINET so  I am not sure if it is worth paying that much.  Pajek is good for really large scale data set, as I find the UCINET cannot really handle really large dataset, but I really find Pajek UI a lot more difficult to learn.  NodeXL seems to be an option too, which is a plugin to Ms Excel. 

I like VENSIM , it is easy for implementation from causal diagram where specify  all variable in  social network analysis.

The interpretation of results of a network analysis can be distinguished according to three

different analytical levels:

1. interpretation of the whole network;

2. interpretation of clusters and components;

3. interpretation of individual positions.

For the case of knowledge sharing,

• Size 
• Network centralization
• Density

then measure alignment

http://www.wissensnetzwerke.de/files/kb_ebook02_social_network_analysis_knowledge_sharing.pdf

NODEXL

I think there are too many ways to measure "best" to be able to answer your question :-)

My impression is that UCINET remains the most widely used package in published academic work. It's also well documented, both in the free Hanneman & Riddle e-book  http://faculty.ucr.edu/~hanneman/nettext/ and in the more recent Borgatti et al. text  http://www.sagepub.com/textbooks/Book237890.

Pajek undoubtedly can handle bigger networks, and is stronger in some specialized procedures (e.g. triads, patent and citation networks ...). It also has a very good textbook

http://www.cambridge.org/ie/academic/subjects/sociology/research-methods-sociology-and-criminology/exploratory-social-network-analysis-pajek-2nd-edition. In addition you can find a lot of slide decks and tutorial material from the authors online.

I think NodeXL appeals most to people who are most comfortable with Excel, and it also directly supports collecting and analyzing data from online social networks. It has an accompanying (again excellent) text

http://store.elsevier.com/Analyzing-Social-Media-Networks-with-NodeXL/Derek-Hansen/isbn-9780123822291/

The various packages in R - sna, igraph, statnet, RSiena etc. - are often much more powerful in statistical sophistication. If you're comfortable working in R and already have data in that form, or want to analyze it also in conventional quantitative ways, then that may be the way to go. I'm not aware of any textbooks on these, but again there's lots of material on CRAN and elsewhere online.

But there are so many more - each with its own strengths: Gephi, ORA, Siena, lynks ...

NODEXL --> GEPHI

Take NODEXL for [really] basic calculation --> Export your data from NODEXL to GEPHI

And enjoy GEPHI ; For me, it is the funniest way to analyze graphs.

http://gephi.github.io/

UCINET is really great but to begin with it is complicated. And it is not very cool for visualization.

And about comparing performance, it depends what you need, as well explained in the above answers. Examples of indicators of quality of a network analysis software : implementation of a (large) variety of metrics ; calculation speed ; maximum size of the dataset ; visualisation capabilities ; connection with other softwares...

The different programs used in ARS identified, represent, and analyze data generated graphs and matrices that will give us the information in a visual manner. These programs help visualize networks and examples, calculating a fast indicators that can be used to find and detect some trends, eg centrality, proximity, etc. The performance of the software could be based on the following aspects:

a) Availability of software: free, paid or shareware.

b) Interface: clarity of the user interface.

c) Metric: quantity and ease of use of metrics and results

d) Import / Export: the ability to import or export to other file formats, either from other analysis software, Excel or others.

e) Graphics. Ability to plot networks.

f) Documentation: availability and clarity of documentation.

g) Actual: refresh rate and recent new software versions.

NodeXL used to be free) and Gephi are the good ones but don not scale well. You can use Python/R/Matlab for data collection and can have better visualizations with D3.

I agree with previous comments and recommend UCINET.  I use it for medium sized data sets.  However, i find its accompanying visualisation tool 'ugly.' I have started pairing UCINET (for analysis) with KUMU (for visualisation).  Kumu is free for academics/students.  I generate a data set in excel, import to kumu to refine and do visualisations, from there export to ucinet for analysis.  Kumu is pretty simple, but depending on what you want to do it can look great.

https://kumu.io/community

The igraph package in R is your best bet in my opinion. Gephi and Adobe Illustrator are good for visualization.