Few days ago very nice set of shots appeared on Digitalcamera.impress.co.jp site where the same alabaster figure was shot with 17 different DSLRs. ISO sensitivity was changed from ISO400 to max available (1600 or 3200). As the shots are, naturally at different resolutions and in general very huge, it is virtually impossible to compare them just by switching from one to another. So this is what I did: corrected white balance and exposure to give them equal "flavour", upresed and downresed to D200 size, then cropped the part of the pic as shown. I have compared ISO 800 and ISO1600 only as I believe those are sensitivities of mayor interest. Unfortunately, I dunno anything about the way guys at Digitalcamera.impress.co.jp adjusted their cameras when making those shots. I believe I have done my part of the work correctly. I have discovered an interesting fact: when printed, noise differences are even bigger then on screen. For instance, D200 noise looks very acceptable at ISO1600 when printed, while Konica-Minolta 7D looks terrible. Obviously, croma part of the noise is very important for noise perception. I have scaled original pics in a way they fit A3 size at 200 dpi. So my crop comparison pic fits A4 paper (8"x12").