The Fujifilm X-H2 High ISO Noise vs the X-T2

yomimoi wrote:

Erik Baumgartner wrote:

..

While the older X-Trans IiI sensor is cleaner at very high ISOs, the new 40MP sensor does OK with a little processing finesse and looks plenty clean if compared at the same viewing size.

Here's the X-H2 at full size with Lightroom/IXT and Topaz DeNoise (A.I. Clear mode) vs. the full-frame Sony A7III with Just Lightroom at ISO 12800. Is this better than you're getting?

X-H2 full-size (L), Sony A7III (R)

X-H2 resized (L), Sony A7III (R). Printed at the same size, the X-H2 would have the same noise, but better detail than seen here.

Thanks for these, Erik, I think I'd seen them before? This is similar to what I'm getting, although I'm looking at ISO 6400 which is realistically the highest ISO setting I'll need/want to use on either camera.

Although not so much in these crops, if I compare my results at the same size in Lr (so 100% for A7M3 and 76% for X-H2), while the noise is easy to clean, there's a lot of fine detail that is just gone (or in some cases wasn't even there in the first place?) on the X-H2 file, such as on the spools of sewing thread, the coins on the bank note, the pen drawing with the dog and the women and a few others.

Even on your crops, both at the resized and non-resized samples, A7M3 shows more detail on some the thinner concentric half circles on the coloured slab next to the bottle (if that makes sense, I'm reaching the limits of my English skills here).

They don't look too far off to me, and downsized, you aren't seeing the X-H2 at full resolution either.

Therefore, I'm not sure that printed at the same size the X-H2 would show better detail.

I'm not sure why you think it would. We're taking about a much smaller sensor at very high ISOs here. At base ISO the X-H2 should have an edge, but if you're looking for the best high resolution low light camera, a FF sensor is going to win here.

I understand this can be of no consquence to many users, but it is to me. Firstly, we're comparing with sensors (A7M3 and X-T2) that are much older and there really shoudn't be much debate as to which delivers a cleaner, more detailed image (at least for the X-T2, which is also an APS-C camera). Secondly, it seems that at ISO 6400 and higher (maybe also at 5000, have not checked) the new sensor requieres an extra processing step to basically match what the X-T2 was delivering 6 years ago. Not a win in my book.

The X-T2 is cleaner, but because mostly it's slower. They likely could have made the X-H2/X-T5 cleaner as well. but would most people want X-T2 AF and rolling shutter?

Again, some people won't be bothered as they rarely go higher than ISO 3200, that's fine. But ISO 6400 shouldn't require that much careful processing and an additional step (and financial investment) on a second app in this day and age.

I suspect using DxO with DeepPrime will produce significantly better results with the 40MP sensor at very high ISOs. That would probably be worth a try.