Hi and thank you,
I'm using the script provided by QVSources examples exactly as they give for Google Analitycs Connector.
There aren't synthetic keys, those QVD are Qlik Sense, the core of the script looks like:
Load a,b,c,d,e,f,g from .\filename*.qvd
(after a long series of script lines in order to build those QVDs).
There aren't cartesian products, and, if I copy the script in QlikView I don't have any issue.
The mistery is the error message: "Data are loaded. Correct the error and retry to load."
I didn´t have one of those errors in Qlik Sense yet, but i remeber those "cryptic" error messages from QlikView.
In most of the times it hat to do with syntax failures which weren´t shown in the error messagen. For example, i didn´t have all necessary fields in a group by statement, problems with nested load statements (called a field by a wrong name etc.). Those failures didn´t produce a nice and readable error message, just an "error" like this.
Do you drop all tables after you don´t need them anymore ? Like tabels for transformations etc.. Did you check with Task Manager or other tools if really the RAM is the Problem ? Because if you don´t get a black screen in the end because you´re RAM is full, i don´t really think so.
Your load script may require more memory than the 1.2GB in your source data, depending on what you are doing in the load script. Joins will always require additional memory to construct the join and if your data set is large, this memory might be large. So I cant really much advice apart from reducing the data size (by eliminating unused fields,especially ones with a high cardinality) or reducing the scope (date range, or partitioning the model into one per region) or increasing the memory.