Luxx wrote in post #18192004
If you had to say how many stops better the 5DIV is vs 5DIII at high ISO what would you say? 1/2 stop? 2/3?
That is actually two distinct questions which could be confused as one. One could be "what ISO on the 5D4 is like the 5D3 at ISO 6400", and the other, "how many stops less noise (read noise) does the 5D4 have at high ISOs?".
These are two different things, but they are both facing the same reality that both cameras capture pretty much the same amount of light, so virtually all of the noise difference is read noise, both quantity and quality, and has little to do with photon noise. That makes the most difference in the shadow areas, especially at low ISOs, where it only affects the deepest shadows. In a high-key or high-contrast ISO 6400 image, you would probably be hard-pressed to find much difference. In an ISO 6400 image with visible gradients in shadow ranges, you would more likely see a difference. Both set to their highest ISO, and pushed to ISO 400k, and median-filtered and reduced to a 600*400 web image, the 5D4, I think, would have significantly less noise.
Same for situations where heavy cropping is needed and more sharpening is used; sharpening increases noise, so the more you have to sharpen, the more the noise is emphasized. Since the 5D4 has smaller pixels, and both have AA filters, then the AA filter of the 5D3 dulls things more, with a larger radius, so the 5D4 should cause you not to have to sharpen as much for heavy crops, extending its advantage. We've already heard lots of people praising the greater crop-ability of the 5D4 at higher ISOs.
Empirically, as measured by DxOMark, the 5D4 has about 0.4 stops less read noise than the 5D3 at medium ISOs like 6400, and this increases to about 0.6 stops at the highest ISOs. DxO does not, however, measure the character of the noise, and the better character of the 5D4 means more practical read noise difference than the numbers 0.4 stops or 0.6 stops would suggest.
So, I don't think there is a simple one-number answer to your question; it depends on the tonal levels concerned, so you can see anywhere from close to zero stops to a few stops subjective difference, depending on what you are shooting and what scale you are displaying the result at.