Did you notice some substantial differences between XD2s and XD3 on Bayer sensors? Is the export performance similar?
Any typical differences on very high ISO portraits with contrasty light, when the difference between DP3 and XD2s starts to be seen?
I don’t have PR to try, using PL9.5 where XD3/Bayer is not available (yet?).
I’ve been testing PR6 for two days. I’ve run about 200 images through both singly and in batch and both in the standalone and from Lightroom. I shoot with an OM-1 Mark II and all these images were taken with the Olympus 150-400. My computer is an M2 Pro Mac Mini with 32 GB RAM. Here are my observations:
Despite the claims of many YouTubers (including some I very much trust) I cannot see any difference between XD2s and XD3. This is across ISOs from 200 - 12800. At 12800 in the deepest shadows there may be a tiny difference but you have to hunt at 400% to see it and even then it may not be there.
Processing time for a single image is about the same. In batches of 20 (the max in the trial) PR6 processed about 20% faster (2m40s versus 3m20s). Nice to have but not really that significant. Maybe if you process batches of 100s of images but my largest batches are 50-60.
The file size on the compressed DNG is about 1/6 of the size of the PR5 DNGs, even smaller than the RAWs. I cannot see any difference even after editing. This is the most significant change and it is a big deal but I’m not sure I will upgrade just for this.
I did not experiment with spot removal.
Bottom line: I do not think the upgrade is worth it although the compressed DNG is on the edge. If I could see some improvement in the XD3 maybe that would tip the scale but I cannot.
Thank you for valuable input.
To be honest, I didn’t expect any miracles, but good to hear about XD3 performance being similar to XD2s.
My general experience was that XD2s brought less microcontrast than XD but it restored real details better, with less artifacts, e.g. on hair, or cheeks. Probably differences between XD3 and XD2s show only in some very specific cases, and I’m speaking only about edits where the difference between XD2s and DP3 is clearly to be seen at 100% on 4k. It just doesn’t look to me that much improvement could be done in XD2s, say by 1/2 stop. Maybe XD3 is all about larger ML training input, possibly getting less artifacts, although it’s hard for me to find any with XD2s already. Any substantially better NR would probably require some kind of object recognition and reconstruction, things very prone to wrong guesses and computationally very costly. Just my 5 cents.
Do you remember this thread?
It was mentioned that “DP3” corrected chromatic aberrations (somewhat better) than “XD/XD2s” and also reduced the pinkish color cast that was visible in some backlit photos by @BHAYT. – So why not incorporate this technology into the new “XD3”?
Maybe, although I didn’t find anything substantially new about XD3 on DxO pages, except perhaps “Built using a larger neural network”, or “Compared to conventional RAW conversion technologies, DeepPRIME 3 delivers the equivalent of an extra two stops of ISO of detail; with DeepPRIME XD3, it can be three stops and sometimes more.”, as stated on DxO DeepPRIME: Still the world’s best AI denoising tool - DxO
My current experience with CA, whether used with DP3 or XD2s, is that in some cases magic wand works best in some cases but can introduce sometimes 1-pixel yellowish halos in other areas, which go away if you diminish CA ‘Size’ to 4, which looks like a safest default in my case. This applies only to some Z8 shots with Z24-120/4S at wide angles. For most of my other lenses it’s safe to turn off CA corrections. I’m getting CA or backlit fringes problems in less than 0.1% cases, so not a real problem.
I’m using PL+FP+VP+Nik, no PR, but if I find time, I’ll download the PR6 trial and test XD3 on some of my photos. It seems XD3/Bayer will get into PL9.6, like it happened with DP3 and XD3 X-Trans in PL8.5 and PL8.8.
Well, add this
The latest generation, DeepPRIME 3, incorporates a third operation as part of the RAW conversion process: chromatic aberration correction. …
to your quotation and there you have it.
.
My current experience with CA …
… some ( ) shots with Z24-120/4S at wide angles.
Hmm, I bought a Z 6 III and this exact lens in the last winter sale, my only Z-mount lens so far. All my others are F-mount lenses and some feel a bit front-heavy with the FTZ ‘extender’. I expect to have more trouble with CAs … and hope DP3 (and XD3) will help with that.
(ed)