Forum Discussion
PowerBI.Dataflows vs PowerPlatform.Dataflows
Bump.
We've been through a similar exercise and have found that the Legacy PowerBI.Dataflows is significantly faster than the new PowerPlatform.Dataflow connector, we're at a point on reverting back to PowerBI.Dataflow connectors, but trying to understand if there are any real advantages on leveraging PowerPlatform.Dataflow connector and what is the future state for the 'Legacy' PowerBI.Dataflow Connector overall.
nxlu OP here.
After much testing I can confirm the following.
Generally speaking for tables that return more than 2,000,000 records PowerBI.Dataflow is faster, less than this there are negligible differences.
Where PowerPlatform.Dataflow comes into their own is when query folding is used, this can be enabled by ensuring the Dataflow itself has Enhanced compute engine settings set as On.When this is set, your Power Query steps will actually write SQL code back to the source (Query Folding) and this can drastically improve performance as the data is requested with these conditions in place, i.e. a filter that reduces the number of rows or a remove column step that reduces the number of columns.
A PowerBI.Dataflow does not query fold and instead downloads the data first and then applies the subsequent steps. So if your source table is 30,000,000 records but you filter and agregate it down to 10,000 rows then the process still has to download the 30,000,000 records first. However, PowerPlatform.Dataflow would return 10,000 without having to download the 30,000,000.
I hope this makes sense, in summary, it depends...Where do you see that option? in PowerPlatform.Dataflows? Enhanced compute engine settings?
Click the three dots on the dataflow that you would like to turn this on with.
Then select Enhanced compute engine settings and click On.
https://learn.microsoft.com/en-us/power-bi/transform-model/dataflows/dataflows-premium-features
