I installed the new version 5 of PureRaw. After some tests with a few of my nature/wildlife/bird pictures, my conclusion clearly is, that version 4 was MUCH better.
Even when selecting the same algorithm (XD2s) with the same settings (luminance 50, details 75), the results of version 4 are a lot better. The noise is less, and there are more details in the pictures.
I am really disappointed.
3 Likes
I thought it was just me. This is pretty much word-for-word what I was about to post. PR5 just doesn’t seem to do anything compared to PR4 on the same files with the same settings.
1 Like
@mtoeltsch , @sg78 , are you saying that XD2 noise reduction is giving you different results with DPR 4 and DPR 5 and that DPR 4’s exported files have more detail and/or less noise than DPR 5’s?
I’m asking because I can’t test, I have no license for DPR 4 - and I’m on macOS.
You’re on Windows, I suppose…
Another poster found that he could achieve similar or slightly better results by changing the settings. The defaults are way different than those for PR4 which is ridiculous. See this thread: PureRaw 5 Surprise - #12 by Rand47
… and further down PureRaw 5 Surprise - #16 by Theycallmedim
(no idea otherwise – just remembered to have read about …)
@platypus
Yes.
I’m on Windows 11, I tested PR4 (DeepPrime XD2s) vs. PR5 (DeepPrime-3, DeepPrime-XD2s).
(-) The results of PR4 clearly outperformed those of PR5.
(-) PR4 (DeepPrime-XD2s) and PR5 (DeepPrime-XD2s) are very different, even with identical settings for “luminance” and “detail”. The PR4 result has less noise and more details.
(-) PR5 (DeepPrime-3) and PR5 (DeepPrime-XD2S) are almost identical.
3 Likes
For the past day or two I’ve been running simultaneous PureRAW4 and PureRAW5 sessions from the same Windows 10 computer with a pair of BenQ monitors.
I’ve noticed a little difference between the two default DeepPRIME XD2s/XD versions, and a significantly degraded PureRAW5 DeepPRIME 3 compared to the PureRAW4 DeepPRIME. (Unless you have superb visual memory, simultaneous monitors are the only way to go.)
You can see the settings in the attached screen grabs, but these are early days and I ask you to cut me a little slack (and I probably need to recalibrate the monitors). These were shot as part of my evaluation of a recently purchased Leica D-Lux 8. Shot at ISO 12500, hand-held, f/4, 1/80th, 34mm equivalent.
This round was intended to see what it took to get DeepPRIME 3 close to PureRAW 4’s DeepPrime XD2s.
It looks like you can get the same or better performance in PureRAW5 as long as you have the patience adjust the parameters and make a bunch of presets…And the ability for simultaneous comparisons. That’s not a brilliant marketing or functional strategy on DXO’s part, but hardly unexpected.
Also note the deletion of the High Quality and Prime choices. For people who have developed work flows with more gentle noise reduction, this forces them to build their own replacement presets or batch process files without noise reduction in PureRAW 5, and do the noise reduction in PhotoLab.
1 Like
Working just fine here. Win 10 under Lightroom. Standard sharpening is more refined, noise-reduction is fine… Good stuff.
- Canon R5 II RAW
- ISO 800 - 6400
1 Like
What ISOs are you shooting with, what sensor format, and in which kinds of lighting environments?
I d/l Trial PR5 and ran 4 Nikon FF RAW through XD2S. I then compared the same DNG files to PR4. I basically couldn’t see a difference. The files ISO were 2500, 7500, 22800, and 25600. Settings were the same. Local masking not important since very limited without AI. Interface of PR5 is better, but not worth the upgrade–now.
It’s easy to see. Denoising seems to work as expected, but detail/sharpness is lost in DPR5. I checked DeepRRIME of DPR3 vs. DeepPRIME3 of DPR5:
Left: 16 bit TIFF from DPR3 compared to the 16 bit TFF from DPR5 as seen in Lightroom’s compare view.
TIFFs from PhotoLab 7 and 8:
I’ve lifted the shadows a little bit.
1 Like
This whole situation is a bit baffling for me. It does seem like DXO has pushed this out without enough comparison and analysis.
This evening I’m going to re-calibrate my monitors – color renderings do influence perception.
Meanwhile, folks could consider downloading the sample files at the bottom of the PureRAW 5 promotional page. Starting from the same source material would put people on a common footing. You should be prepared to run 4 and 5 simultaneously, even if you don’t have dual monitors. And compare both the DeepPRIME and DeepPRIME XD2s processes.
I have been doing several comparisons with RAW files from my OM-1 (.ORF). If I leave DPR4 on the settings I’ve always used (DPX2) and leave DPR5 at its default settings, DPR4 wins hands down. Only if I whack all three DPX2s settings up to maximum can I see any appreciable improvement with DPR5. Something’s not quite right here - seems to have been pushed out before it’s ready for prime time. Hopefully there’ll be a patch down the line to sort it out - it CANNOT be as intended, surely?
I’ve been fiddling with the settings since V5 was released and I made the experience that you just have to crank the sliders a lot more with V5 and the new Deep Prime 3 model. At default settings I got noticably soft images as well. I feel like this could use a little more tweaking on their side but so far the results I get are fine with the settings I set up for now (luminance 60, details 70).
I read that with the new model they went for a more natural look and I think that’s what we are seeing here. To me the files look less clinical.
Also, I saw that they react differently to the sharpening settings in Lightroom now. You can tweak the files further there. Together with the PureRAW settings I mentioned above, in Lightroom a sharpness of 20, radius of 0.8, a touch more details and some sharpness masking works really well.
The files do look different. I wouldn’t call it better but they now look more like a photo taken at a lower ISO instead of simply being denoised.
Howerver, and that is actually what I like the most about V5, it’s just WAY faster. Somewhere around 30 to 50% faster (depending on the file) compared to V4. Since I like to process large batches of images at once that is a significat reduction in processing time. Processing a batch of 250 images is now done about 20 minutes faster on my end.
2 Likes
@Mathias_A I would agree with your point about speed - similar improvements here.
I also just noticed that I see a difference between the lenses I used.
The photos I took with my Tamron 70-200/2.8 G2 look softer after they got processed. But with the same settings the shots I took with my Nikkor 180-600mm look sharper after they got processed.
Edit: Processed them again. Pushed back luminance a touch and added more forced details. Now it looks better.
Guess I will have to process them separate now depending on the lens I used.
Edit2: Naa, we good. These settings work well with my 180-600, too. Crisis averted 
1 Like
+1
And one still can crank up the settings to get a less natural look. 
Like everyone else here, I’m quite disappointed with PureRAW 5 as it gives much less desirable results than Version 4. I will say that DeepPrime 3 is fast… but sadly that’s of no value because the results are poor. I have sent information and requesting help or a refund but I’m assuming based on all the issues they are busy working on a solution. Also found that running it as a plugin from Lightroom it often did nothing to the photo at all.