Yes, in general it is possible to use initial load for the first time data load. Regarding the performance it's always difficult to decide whether it is fast enough as it depends on many factors. Generally speaking the replication process is single threaded process that applies one statement at a time right after the other using normal SQL statements. So you see the 6GB replication will take a while. There's an option that allows you to process the whole inserts faster by grouping them for large tables and using more then just one replicat process.
Another possibility would be to use the initial load extract, put the data into ASCII fomratted text file and then load it with multiload Teradato in parallel into the Terdata database. Unformtunately there are no scripts ready to use for the MLOAD tool - you need to write them on your own based on the formatted output.
Sorry, but I don't have that sort of information. It would be better to get in touch with Oracle Sales. They might have that sort of information you're looking at or they will request it from the Product Managers. As an alternative you could post it at the Teradata blog: https://blogs.oracle.com/dataintegration/entry/what_s_new_for_oracle1
Oracle GoldenGate for initial load is usually not the fastest method but it is 'easy'.
If you use that method I would strongly suggest that you have multiple parallel paths
Extract A ---> Replicat A Tables A, B, C, D
Extract B----> Replicat B Tables E, F,G, H
Extract C ----> Replicat C Table I (big table)
I don't think anyone (including the Oracle PMs) will give you any timings as it will depend on the network, how busy the Teradata Database is, type of data etc.
Strongly suggest you use the parameter BATCHSQL parameter and you may have to make some adjustments to the subparameters within it for Teradata.