My previous answer has been deleted by other. If this post just seems like spam, it's because you haven't had the same problem as me.
Please leave this article alone to help others who are having the same problem as me.

possible.

Although this post is 5 years old, I am leaving a link and sample image of the package I created for the questioner and for those who come across this post after experiencing the same problem as me.

Like the questioner, I also started with matplotlib, and it was very slow.
I thought about trying to modify mplfinance, but it was difficult to do.

I eventually had to move on to pyqt and try pyqtgraph and finplot.
In the case of plotly, Bokeh, and Seaborn, they were excluded from the options because they did not seem to be able to connect with other GUIs.
pyqt was fast. However, it was only slightly faster than maplotlib.

The same problem of slowdown occurred as data increased.

I was wondering whether to try a language other than Python, but I went back to matplotlib and tried something new.
And it was successful.

I was able to create candlestick charts at high speed.
And this doesn't use pyqt.

By making it that way, it can now be used by connecting to a GUI other than pyqt.
Even when connected to tkinter, it operates smoothly.

It seems to work comfortably up to 10,000 pieces of data.
However, when the number of data is around 40,000, it is not comfortable. I had to use a little trick to make it even smoother.
I'll solve this problem someday.

https://pypi.org/project/seolpyo-mplchart/