PL 9.2.1 is out

andras.csore · November 15, 2025, 11:09am

‘Shared GPU memory’ starts with 0.2GB used out of 15.9GB available, grows steadily, and then DopCor crash happens when it reaches 15.3GB.

Yes, i see - not the crash, what i’m sure is happen in 16GB as you describe - but is see the ‘shared memory’ rising. Update: crash at 16GB.
Example: 256 photo - Oly 20Mpix RAW. No NR, no editing. 32GB system memory, 4GB GPU VRAM. OpenCL enabled. 1 (one) export thread.

The DopCor ‘shared memory’ rise is linear.

Shared goes up like 2.2GB
In the end of the Export, its stay (not change)
Of course kill the DopCore its goes down.

I re-test (after kill DopCor) now with NR: Standard → Shared goes up to 3GB.

After i not kill the DopCor, start same export again (NR: Standard). As expected, its not ‘reset’ the shared, but continue from 3.0GB, raise is linear.
End like 6GB as expected (5.7GB, what means one export batch in this example takes like 2.7GB shared).

Note: i not start data capture from the start of export.

@BHAYT - may you also interested on that.

add-on: a made a small screenshot when export start, to better see the linearity

BHAYT · November 16, 2025, 10:51am

@andras.csore and @Wlodek I am interested but I had already started a post, which I have now saved because I encountered other “interesting” “features” (for want of a better word, and I can think of some but they are not permitted in the forum).

Yesterday I was busy on basic car maintenance and then gardening because we had some sun here in the South of England!!

This morning I located my Bulk Test images (Bulk rather than BULK, 1,000 images versus 11,000 but that 11,000 is made up of 7,000 RAWs and the rest are JPGs) on a USB3 connected SATA SSD and ran the ‘NO DOP’ group with ‘06 - No Correction’ applied on the 5900X with 5060Ti(16GB) and got the following,

But there is a little message peeking out from behind the graphs and the graphs have “flat-lined” rather than “saw-toothed”.

The message was

It had failed on image 614, the images are numbered from 1

and it did finish processing after the fault

KenA · November 16, 2025, 6:09pm

I have inatalled PL 9.2.1 on W11. So far all seems to be running OK and all types of masks working normally. Exporting also OK but still quite slow.

Stenis · November 17, 2025, 1:41am

Finally I have got the machine they have built for me the last weeks. It has 32 GB RAW, 4 TB SSD and Nvidia RTX 5070 Ti with 16 GB. It is a significant upgrade compared to my two generations older RTX 3060 Ti with 8GB.

I don´t seem to have any problems at all now and it is very responsive even when using the premade AI-models in the menues.

I just wonder if they have changed the behavior of the menu in this version because now it is instantly selecting the different areas corresponding to what I am about to select in the menubefore I even click on them. It is diffrent from how it was last time I used it.

Is there more users than I that have found that my “Fine Contrast”-slider is gone after installing this version and version 7 of Filmpack. Is that a way to get me to sign up for next version of Filmpack?? … or what?

Wlodek · November 17, 2025, 2:55am

No such problem with FP7 and fine contrast here with PL9.2.1/Win installed.
I have license for PL9 and FP7, had a FP8 trial installation. Deinstalled FP8 after trial expired but then an FP8 Ad appeared in PL9 GUI. Deleted some FP8 leftovers in “Program Files”, “ProgramData”, %LOCALAPPDATA%\DxO, including %LOCALAPPDATA%\DxO\Licenses\filmpack8.key file and the FP8 Ad didn’t appear again. No problem with Fine-contrast sliders at any stage.

Check PL–>Help–>About if it sees FP7 license – maybe you have deleted it by mistake and it needs to be reapplied using license management option.

Stenis · November 17, 2025, 3:17am

Thanks Wlodek for your input. I’ll try a reinstall tomorrow. Fine Contrast is pretty essential.

zkarj · November 17, 2025, 7:54am

Well, I should have stuck with 9.2 because 9.2.1 performance is terrible. Oh well.

Fineus · November 17, 2025, 8:01am

Based on a brief 20 minute play, my experience is now generally positive v9.2.1 (build 542) is largely positive ((edit: spoke too soon - see further updates below)).

Masking seems to work reliably (in that it functions) and looking at some of the concerns above: I can export and I do still see Fine Contrast controls (and they work) based on a FP7 license.

Bits of note:

Performance continues to be slower than I would like, especially when the number of masks increases.

Sky Masking via AI is poor (I’ll use the Control Line workaround for now, but this does need improvement). Look at all these trees it thinks are sky!

Update: Performance really tanks with usage.

I’ve a folder of 69 unedited photos. I’ve just gotten as far as editing 4 out of that number, and exporting it. Everything is sluggish as hell, and exporting just that final image (not even the other 3) took 1 minute 11 seconds. 1 AI (people) mask, 1 graduated filter mask, 1 auto-brush mask. It was even black and white.

By contrast, loading PL9 and editing one single image felt much snappier, it had multiple AI masks (8+) plus other masks, and exporting it took 15 seconds (colour).

As I write this, I haven’t touched PL9.2.1 (it’s not exporting or anything) and it’s using 12% CPU and a whopping 70% memory.

The performance/optimisation is awful.

Restarting PL releases over half of the memory in usage and CPU use drops to minimal levels.

Update 2: Having edited just two more photos in the same batch, and exporting each individually, I’ve just hit a record high export time of 2 minutes and 38 seconds for a single image.

Update 3:

Further experimentation indicates that export time more than doubles after the first export, even after a fresh restart.

Nothing unusual is occurring with these images - in many cases there are only two masks being applied - no other big changes explain this performance shift (they’re from the same camera, shot on the same day, same shooting and editing processes being applied…)

What the hell did DxO do?!

Update 4:

Downgraded to 9.2.0 and ran more tests with the same batch of images, same editing process etc. etc… and performance is right back where it ought to be (which is still slower than PL8 but…!)

TomDX · November 17, 2025, 10:20pm

Wow… this is disheartening. First they took away half of FilmPack [for non-Fuji shooters like me], and now this broken mess? I’m also on Windows + NVidia, so I’m now wondering if I should even upgrade from PL8 to PL9, or jump ship to ACDSee [which has its own pros and cons, but it’s 40% of DxO’s bundle right now - $120 vs $310 USD].

DxO bundle:

ACDSee bundle:

andras.csore · November 17, 2025, 10:47pm

Its removed due Fuji legal request. Not because DxO want to remove it.

I’m also on Windows + NVidia, so I’m now wondering if I should even upgrade from PL8 to PL9

Why not try it (trial period)?

and now this broken mess?

I don’t think its ‘mess’. If you has like 12-16GB GPU, than all things works fine. Some issues may come up (like some case export performance and similar, but i think its can be ironed-out in later releases as seems not major bug) but otherhand works fine. If you has like 6GB GPU, ‘Pre-defined’ AI masks may not works fine (at general).

Stenis · November 17, 2025, 10:54pm

I have reinstalled and then Fine Contrast showed up again. I don´t know what I had done without it I think it is the most valuable asset in Photolab beside Deep Prime and the new AI-functions.

zkarj · November 18, 2025, 12:12am

It’s “not good” but that’s in comparison to the less capable PL8. It is not even close to the stage where I would consider switching to something else.

TomDX · November 18, 2025, 12:22am

Intentions don’t really matter - I bought something that they then took away, and I can’t use it anymore, and didn’t compensate me or anyone. They simply took our money, and removed the functionality. It was Fuji’s legal request? Fine, I can understand that… but reimburse us, then.

I don’t like installing several versions of PL, but I may end up not even doing that based on your next comment… I have 32GB of RAM and my GPU is only 4GB [RTX 3050] so sounds like I’m ~~at the bare minimum~~ not going to be able to even run it, let alone have memory to spare for the bugs/inefficiencies.

TomDX · November 18, 2025, 12:30am

After reading more comments, I am considering switching even more because of the many bugs and overhead specs needed to avoid crashes - I only have 32GB of RAM + 4GB GPU [RTX 3050]. That, coupled with the price tag, is making me at least trial ACDSee [which is far from ideal for me, I really dislike the import-photos-and-build-catalogues workflow, I much prefer DxO PL’s simplicity to work with photos directly on folders, plus learning a new program].

I’m toast lol:

John-M · November 18, 2025, 12:50am

Other than AI masking there’s actually not much in the way of “must haves” in PLv9 compared with PLv8 (which is/was a very stable, well performing variant of PL).

So, I suggest taking advantage of any upcoming pricing specials to purchase PLv9 - then “put it on the shelf” until DxO has all its GPU-related issues sorted out (as I’m confident they will be, eventually).

Alternatively, give PLv9 a miss altogether - and wait another 12 months for PLv10.

Either way, you’re not missing out on much by sticking with PLv8 in the meantime.

TomDX · November 18, 2025, 12:57am

I think you’re right. And really, at this point based on the minimum system requirements, I’m definitely going to stick with PL8

Maybe next year I’ll have a new machine with a better/newer GPU, so yeah, skipping PL9.

Stenis · November 18, 2025, 1:45am

The competition works fine on 8GB VRAM on the GPU,even 6GB might do. DXO will have a lot of reasons optimizing their code to meet the level of the competition. They know they have to unless they are prepared to loose a part of their installed base. That is a pretty strong incentive.

mwsilvers · November 18, 2025, 3:50am

It is unfortunate, but the Fuji simulations were originally intended for Fuji shooters only. The fact that DxO made them available for non-Fuji shooters was an afterthought. Since the rights to these simulations was owned by Fuji, DxO had no choice in the matter. Just how would you expect to be reimbursed? A certain amount for each simulation removed? Or a certain amount for all the simulations removed as a group? And, if you then buy a Fuji camera, I assume you would be OK with DxO automatically recharging for the Fuji simulations. FilmPack has over 100 simulations with Fuji making up a small percentage of them. Beyond that there are many other features in FilmPack, among them the very popular Fine contrast sliders and the Luminosity masks. Do you expect a full refund?

Mark

mwsilvers · November 18, 2025, 4:26am

What processor and hard drive do you have on your machine? I have a old Windows 10 machine with a few upgrades and my performance is fine. My processor is an old Intel i7-6700 @3.40GHZ with 24GB of ram. I added a 2TB WD Black SDD drive and an RTX 4060 with 8gb of ram and I run PhotoLab 9.2.1 with almost no performance issues except for the AI Mask keywords problem which is a known issue in the process of being resolved.

As to the lack of new or updated features in PL 9 to which John alludes, I strongly disagree. PL 9 is the most feature rich update that DxO has ever released. While I probably have missed a few in this list, PhotoLab 9 now includes the following:

Removal of sharpening restrictions below 75% zoom for both Lens Sharpness Optimization and the Unsharp mask.
A new Lens Sharpness Optimization algorithm.
Ability to apply DeepPRIME in local adjustments.
Ability to apply Lens Sharpness Optimization in local adjustments.
A new option to view the effects of global and local DeepPRIME denoising in the editing window full screen and at all zoom levels while editing.
New AI based object selection masking in local adjustments
New local adjustment masking and sub-masking layers.
The ability to combine various local masks.
The ability to erase and shape portions of control line and control point masks. and to do the same for AI masks.
The ability to set the edge diffusion of control point mask edges.
Removal of PRIME noise reduction. Perhaps a problem for some people but it helps save space in the interface.
New filmstrip stacking features, including the stacking of virtual copies as well as separate image files.
New advanced batch renaming functionality.
New exported file renaming in the export dialog.
New project creation features
Pro RAW support
HEIF/HEIC support
The retention of Advanced History entries across editing sessions in the Windows version which has been a feature of the Mac version for quite some time.

TomDX · November 18, 2025, 4:30am

Again, intentions don’t matter. The capability was there, a capability that was even capitalised by DxO when reviews online showcased this known capability - that’s how I got to it and bought it, as did many others.

How do I expect to be reinbursed? It’s not about the money, it’s about the unethical and sleazy way they acted and dealt with this situation: they had no issues taking my [our] money, and then removing those simulations, making what was fully functional, partially non-functional. That’s bait-and-switch.

In other words, DxO was happy to receive my money, and now I can’t use what I paid for, and you’re saying I’m supposed to say “aw shucks, that’s unfortunate”? That’s not unfortunate, that’s unfair.

Moreover, they did not warn users, they silently slipped it in an update, and only then, after the fact, they announced/explained the changes.

So yes, DxO could have said “hey, in the next update we’re cutting these feautres because Fujifilm is threatening to sue. Do not update if you want to keep what you paid for.”

But they didn’t do that. So ok, failing to do that, DxO could have said, “hey, you paid for a full pie, and you only get to eat 90% of it. Here’s 10% of it back.” Or 5%. Or whatever, again, it’s not the money, it’s their unfair treatment towards their customer base.

And frankly, an option for a full refund should also be on the table. Full refund, and they revoke my license, simple as that. Why? Because THE reason I bought FP was taken away. If I ever get a Fuji, I have the option to buy it again [and I doubt it].