The seriousness of the graphics performance issue cannot be overstated.
I have tested ImageGraphTest, which draws to an offscreen Image and then
draws that image to a window on both jdk1.1.4 and jdk1.2beta. The
results (in seconds, as timed by Java and displayed by the program) are
as follows on my machine, which is a 200 MHz Pentium Pro with two 4byte
Matrox Millenium graphics cards installed:

                             256 colors true color
                         jdk1.1.4 jdk1.2beta2 jdk1.1.4 jdk1.2beta2
Clear image bitmap time: 0.04 0.19 0.14 0.18
Draw vectors to image time: 4.27 49.7 4.24 49.8
Blit image to window time: 0.64 2.05 2.17 2.56

This same 10x slowdown also manifests itself in very sluggish
performance of Swing on JDK1.2, which I am certain is because Swing
draws to offscreen images as well. Also, when running in 256 color mode,
the quality of the output is unacceptable, most likely because of the
translation from the true color offscreen buffer to the color palette of
the 256 color mode.

I don't think I'm exaggerating when I say that this poor performance
threatens the viability of Java as an alternative to building pure
Windows applications. Any vendor who uses Java for a commercial product
will be simply blown out of the water by vendors who instead build
Windows applications.

Test Code too large to save within bug report, please see test case located
in dirtectory:
  /net/galapago.east/files/test/bentley/graphTest_critical/

Run the example by compiling it:

javac ImageGraphTest.java

and then running it, giving it the name of the graphics log file as
input:

java ImageGraphTest map.log

It takes a while to read in the log file, and you don't see any activity
except for disk activity during that period.

The performance of the example in JDK 1.2 needs to get within a few
percentage points of the performance of JDK 1.1.4 for it to be
acceptable.