In the ongoing fight between Opera and Safari to be the first to completely pass Acid3, Opera's released an internal build that passes in all but the performance tests:
The test log (click on the "A" in "Acid" once the test has finished) reads:
Failed 0 tests. Test 23 passed, but took 71ms (less than 30fps) Test 26 passed, but took 467ms (less than 30fps) Test 69 passed, but took 26 attempts (less than perfect). Total elapsed time: 2.09s
Ian Hickson (author of the Acid3 test) reports that there has been some debate about text antialiasing and how it affects WebKit's test performance. I have to admit that when I first saw the test I wondered if differences in antialiasing would be a factor in conformance to the reference rendering, since obviously antialiasing results depend on whether a system is doing "normal" antialiasing or one of the variations of sub-pixel rendering on an LCD screen. Hickson seems to believe that antialiasing should be consistent across all browsers on all systems, which seems silly. I'm glad they've managed to reach a compromise where they simply don't test the antialiasing behavior of browsers.