In debug mode try a limited load of 10 or fewer rows. This indicate if your problem is due to invalid associations between tables or synthetic keys.
Otherwise check that you are not combining data from tables that are not associated - this will cause a cartesian join between the tables.
If table A has 100,000 rows and table B has 100,000 rows and the tables are not associated, using fields from table A and B in the same chart will cause QlikView to generate 100,000 x 100,000 roes of data 10,000,000,000 rows!
Hence you run out of RAM.
so if the problem is only in one of the sheets, i recommend to generate the .mem file or check the calculation time of the charts in this sheet to see what is the chart that is causing this anomaly, check if you have calculated dimensions (not recommended) and if you have complex expressions maybe it would be better to minimize the charts by default or add calculation conditions on the charts to force the users to make selections before displaying this chart this will definitely improve the performance. other thing to check is the triggers on the sheet opening.
Agree with Malek,
unfortunately QV does not have tools available to dedect, which object requires the RAM (unless opened).
If open, the MEM.file might assist.
We went through the painful way in opening each object in each sheet and to watch the RAM-consumption simultaneously on the server in order to identify the object.
Finally it turned out to a combination of several causes: a calculation referring to the total number of records, an object (pivot-table), building a cartesian product and finally a clone of this object as server-object ....
The main issue comes with the tables, if you displaying huge set of rows then you will get this issue or if you are doing any complicated expressions or many calculations.
You can check the memory occupied by various objects in
Sheet Properties -> Objects -> check CalcTime and Memory
Memory - is the size of the object
CalcTime - is the time taken for the calculation.
Based on this you can proceed further by limiting the number of rows display by forcing user to select some filters.