On the left, the Sony A900 image is reduced to 12 megapixels using Bicubic Sharper - this emphasises, rather than reduces, the grainy appearance of noise and matches the file size precisely to the Nikon D3 on the right. The actual light levels involved here (1/800 at f6.3 for the Sony, 1/1600 at f4 for the Nikon) are within 1/3rd of a stop of being the same, and both images were overexposed by one effective stop in camera, then set to -1 exposure using Adobe Camera Raw. In both cases, no sharpening, no luminance NR, and 50 chroma NR have been used. The black point of the Sony shot is at 12 while the Nikon is at 2 - this may tend to emphasise the Sony noise in the darker areas by increasing its contrast (but was a better setting for the overall look of the image).
While the Nikon file shows a smoothed luminance structure, the Sony applies almost no in-camera NR to the raw file, resulting in a very crisp and clear grainy structure. However, the Nikon does not appear to lose fine detail, indication that an adaptive NR process is happening.
Taking this comparison and others, I conclude that while the A900 is not identical in 6400 ISO performance to the D3, in practical applications it will prove as robust and deliver similar levels of fine detail and noise. Each file would normally be custom processed - applying some luminance NR to the Sony file makes it closer to the Nikon 'look'.