As already commented on the SO question this is hard to reproduce. So anything you can add concerning the configuration in use may help here.

I guess I'm confused why anything plots at all w/o a plt.show() command. Is this some kind of IDE you are running in? Can you reproduce w/o an IDE, and can you perhaps temporarily remove your ~/.matplotlib/matplotlibrc just to make sure you don't have some funky setting.

I usually use the PyCharm IDE, but I can replicate this from the terminal (which requires using the plt.ion() and the plt.show() commands), in either a normal python session or an IPython session. I have tried removing the matplotlibrc file but then I get the same behaviour (it upsizes the default 6x4 figure size to 10x9, but not the 15x9 requested). I have also tried editing the rcParams file but to no benefit.

To address my configuration, I have noticed this problem since I upgraded to a larger monitor. I now use a 3 monitor setup (laptop + 2 external displays approx 15", 23", and 43" respectively). I am running on a lenovo thinkpad T460 and using Ubuntu 16.04 LTS. Let me know what other details you might need.

I suspect the 15" monitor is the primary one in this setup. Hence the figure is resized to match this screen resolution.

That's correct, changing the larger screen to the primary display did resolve this. One interesting remaining question though is why does this affect not still occur when I set the size to be 30x20? If I set the size to be this large then surprisingly it then works fine.

I have updated my answer on Stack Overflow, but to summarise:

Root of the problem

As explained in the accepted answer, the issue is that I am using more than one display, and Matplotlib cannot produce a window that is larger than the primary display. Unfortunately changing what Ubuntu considers to be the primary display is currently an unresolved bug. Hence the problem does not lie with Matplotlib, but with Ubuntu. I was able to resolve this issue by setting the display to the top left monitor in my setup.

@beangoben Your issue is completely different. You need to load the inline backend before setting the rcParams.

I'm going to close this as not a MPL bug....

@jklymak Well, it's not a bug, indeed. But the point is maybe that it should be possible to create a figure which is larger than the screen somehow. Ideally I could imagine something like plt.show(force_figsize=True).

As an extension to the suggestion by @ImportanceOfBeingErnest , perhaps a setting in the rcParams to state that the user is aware that they are asking for a larger plotting window than be displayed on screen.

The limit on window size is imposed by the window manager, so there's really nothing that Matplotlib can do to make it bigger. There is the possibility of adding scrollbars as suggested in #7338.

@QuLogic Not sure if I understand this restriction. As I see it, it is imposed by matplotlib. As an example, the following creates a Qt Application with a figure size much larger than (my) screen.

import matplotlib.pyplot as plt
from PyQt4 import QtGui
from matplotlib.backends.backend_qt4agg import FigureCanvasQTAgg as FigureCanvas
from matplotlib.backends.backend_qt4agg import NavigationToolbar2QT as NavigationToolbar

class HugeWindow(QtGui.QMainWindow):
    def __init__(self, fig, size=None):
        self.qapp = QtGui.QApplication([])

        QtGui.QMainWindow.__init__(self)
        self.widget = QtGui.QWidget()
        self.setCentralWidget(self.widget)
        self.widget.setLayout(QtGui.QVBoxLayout())
        self.widget.layout().setContentsMargins(0,0,0,0)
        self.widget.layout().setSpacing(0)

        self.fig = fig
        self.canvas = FigureCanvas(self.fig)
        self.canvas.draw()

        self.nav = NavigationToolbar(self.canvas, self.widget)
        self.widget.layout().addWidget(self.nav)
        self.widget.layout().addWidget(self.canvas)
        if not size:
            w,h=self.fig.get_size_inches()
        else:
            w,h = size
        self.canvas.setFixedSize(w*self.fig.dpi,h*self.fig.dpi)
        self.show()
        exit(self.qapp.exec_()) 


# create a figure and some subplots
fig, axes = plt.subplots(ncols=4, nrows=5, figsize=(24,16))
for ax in axes.flatten():
    ax.plot([2,3,5,1])

# pass the figure to the custom window
a = HugeWindow(fig)

Of course a Scrollable window would be equally nice, as seen in this SO question.