Forum Discussion
How do I improve render speed?
Maybe I'm using some of the wrong terminology, and I apologise.
I originally created the calculations as measures using DAX within the Reports View of Power BI Desktop.
I recreated them as summary tables in the Data View. This is Power Query, right? This is what I've been refering to as "columns" since they are both precomputed.
I don't know what the definition of large is, but I'm still concerned about the speed difference between presenting the summary table as a grids (near instant) vs displaying it with a tachomoters (22sec).
Ahh got you.
Edit Queries section is definately the Power Query area. So yes, anything you did in here is pre-calculated.
Measures are calculated at run-time, whenever the visual that references is loaded or has its context changed. Your measures should have a little calculator icon next to their name in the field listing
Calculated columns are your inbetween, they are done after Power Query, when your data is loaded. They are the slowest but are still an excellent tool when used correctly. Calculated columns will have an icon that looks like a little window with a 'fx' symbol.
To help you start off, go to the screen that deals with your table relationships and take a screenshot of it. That might give me an idea of how complex your data is.
Next, give me an idea of the number of rows each of your tables have. That should give me an idea of how large your data is.
Lastly, have a look at your measures, and see how often they refer to the same data in one expression. Use my example in the earlier post to see what i mean. See if there are any options to convert some of the data references to variables so that you minimise the number of times DAX needs to go and get a result.
The table relationships and number of rows are shown in the following image. There are no measures or computed columns. There are lots of summary and aggregate tables - some of them could probably be merged, but that can wait.
Given that everything is pre calculated during the import, I don't yet believe that the issue is related to the schema.
Its quite a small dataset, so i'm really quite surprised you are having performance issues. I was expecting something much much larger. Most of mine load within 5 seconds and are much larger. They probably have half the number of visual elements but are making much more complex run-time calculations. Here is an example of some of mine (a lot of the tables have 500,000 to 1,500,000 records):
What happens if you build a page with the same information you display now, but with simplier visuals? Try to get the same number of visuals, displaying the same information, but using a much simplier visual. Don't be afraid to let them overlap for testing. The purpose here will be to force Power BI to make the same calculations, but test whether the visuals themselves (mainly tachometer and infographic designer) are just slow to draw.
If I can refer you to my original post,
With all the visuals turned on it takes 22 seconds.When I replace each visual with a simple table or card it's done almost instantly.
I love the schema you've got there - it really puts my task into perspective. It also emphasises just how powerful the Power BI engine really is.
Thanks for clarifying, i'd read your original post as if you had replaced all visuals with 1 or 2 tables.
Try again, this time don't replace the Tacho's. Do they add much to the render speed?My gut feeling is that one of the visuals is slow to draw in the browser. I've used Tachos and they seem ok, so i'm wondering if its your other custom visual.
Under 2 seconds with Tachos. It definitely looks like an issue with the Infographic Designer.
I have now checked the report with both Edge and Chrome with the same results.
Great! So now you know where the problem is. You might be able to gain back some performance if you can convert and some of those Infographic designer to other visuals. I'd like to imagine its a compounding problem. Let us know how you go! I'm super curious.