You should have been clearer.
Sorry, I just didn´t really expect everybody to understand but that´s fine. Happy New Year Wolfgang.
Funny that Adobe, ON1 RAW, Lunimar, and a couple of others I use software from didn’t seem to have had problems with the macOS cockups (I can’t speak for Windows because I don’t use it). Did everybody except DxO just get lucky, or could your explanation be wrong? I wonder… (actually, no, I don’t).
4 Likes
For example, Blackmagic Design and their Resolve suffered from the same driver related issues as well.
So just because those raw photo developer apps did not experienced problems, does not mean DxO are less competent or lacks something in the development process.
Other media related applications did suffer as well.
what DxO did lack in was perhaps in the communication area - both internal and external - there are some hints for them to learn from this.
4 Likes
What driver-related issues? On macOS? From reading this forum, I know there were Nvidia driver-related issues on PhotoLab for Windows . What driver issues affected macOS?
It just seems to me that DxO has a great deal more issues than others. That’s just my impression based on using a great deal of software from many editors, and not just in the photo processing area.
Yes, users of PhotoLab have reported issues in the Mac version as well, but I’m. not aware that any of these have been caused by “cockups” by Apple. And if it’s not Apple, who could it possibly be?
1 Like
They have to release it on the day because the venture capitalists want to see the money coming in. That’s what happens when you take the king’s shilling.
1 Like
These problems are at DxO’s door. You do not release software that is not fit for purpose. There is no point turning round after and saying “Well we know it doesn’t work but that’s because Nvidia updated their drivers”. This is amateurish at best.
There are fundamental problems. Quoting a CPU, GPU, and VRAM does not go far enough in deciding whether a machine can cope with the demands of PL9. VRAM and RAM comes in different types and speeds, the motherboard can cause bottle necks depending on the infrastructure - so there are other dimensions that effect delivery.
But the truth is that DxO knew that their minimum specifications would never deliver ( they were scared to be honest about it) or they didn’t know. Either way, it lies at DxO’s door.
4 Likes
Here’s one of their “cockups”, which affected many users, including PL: 
HOT TOPIC - macOS Tahoe 26.1 update- Gen Ai Issues – ON1 Support
An interesting statement there: “We reported the issue to Apple during their beta test period”.
6 Likes
Uh. I had a 5070ti with an ULTRA 9 275HX (12gig) and had routine export failures. But that wasn’t even using Ai features. I returned the laptop and started digging into Macbooks, but those users have given up and moved on from what I can see. Honestly, I’d rather just have 8 on a new machine and fly at this point. But I requested it 3 times and they wouldn’t send me a copy. Only wanted to keep using me to help them fix 9. So I’ve been back on my old RTX2060 wityh an i7 and 6gig VRAM using XD3 with no issue. Never a crash or failure.
1 Like
Well spotted, Wlodek. It is possible that this is indeed due to an Apple cockup, but it is equally possible that such issues occur with software versions that are not entirely tested with a new OS before they are released. Also, many developers are not strictly following developer guidelines, frequently taking shortcuts or using deprecated features of an operating system that may disappear with an upgrade.
I believe problems frequently arise because many software developers, for a variety of often good reasons, release products before they are fully tested with certain OS versions. To be fair, this is extremely complicated and costly to do, because both operating systems and applications are getting more and more complex. If one absolutely must rely on a piece of software and cannot afford to not have it work properly, then one should not immediately upgrade to new OS or application software versions.
All this being said, my impression is clearly that DxO struggles more with this than other vendors, and again, I can only speak for the macOS side of things.
In my experience, developers know or suspect that their software is more or less prone to errors, even if there is an intensive testing phase. They themselves would therefore rarely release software onto the market. Sufficient pressure must be exerted by marketing or customers to do so. The company that can keep this gap as small as possible brings products with high customer satisfaction to market.
1 Like
let just say there hasn’t been a PL release with so much complaint about, sure release in a Fall always have some minor issues that get fixed before black friday but PL9…. ouffff!!! i have no intent to buy it, even though i’m still using PL7 and will loose my 2yrs upgrade, nothing good in this thread regarding the latest version. how did other software got away with latest MAC update but not DXO? even ON1 seem a better take than Lr.
2 Likes
Still using PL7, PL9 sitting on the shelf 
I don’t deny the problems that users have had with PL9 yet I do use it, including simple point & click AI masks, on a truly ancient and well below spec Win 10 PC that only has an equally ancient and below spec GPU.
It’s not slick but I’m an amateur only poking at the odd image every now and then, and I have plenty of time in which to do this.
At the end of the day, it’s what works for you that’s important. Be that PL9 or any previous version.
2 Likes
Note: 2025.01.06 - edited due spam detection / flag and small update for clarity
I think other software’s use different AI ‘stack’ than DxO (i think). DxO use ONNX Open (Neural Network Exchange - The open standard for machine learning interoperability) . What is great.
Seems the GPU driver (and may some case OS parts) what’s connect to the ONNX provider(s) not so good (buggy) sometimes (in some GPU driver version). Looks especially in NVidia (?)
May this AI area is still very flexible. Also, i think in few software ‘keyword masks’, like: ‘sky’ may not or not fully AI - but some very-very fine tuned Luma/Chroma/hue whatever algorithm.
Note: why in ‘CPU mode’ in DxO everything works? → Not just because its ‘Software only’, but due (because) ONNX ‘software only’ provider handle (implemented) all software call - and not depend on GPU/NPU drivers.
2 Likes
Very nice explanation. Makes sense. I’m heading to get a new MAC M4 PRO 48gig, 1TB, with 20 cores. We’ll see if the new update clears everything up or not. Either way, I’m going MAC and possibly another software should this not work right.
1 Like
Do you still have problems with AI masks on PL 9.3 ?
I have the Mac Mini M4 Pro (12 cores) and 48GB RAM. Performance is acceptable and Export is up to 10 concurrent images. I have seen some slight lags when editing 50mp images from my a1 with a lot of active masks but on the whole I’m happy with it.
1 Like
I haven´t seen the slightest problem with the AI-masking with Capture One for years.
It seems like the Photolab premade AI-presets use far more demanding local AI-models than many others and that is a little strange because these AI-models ought to be designed just to focus on very well defined objejects like people, vehicles e.tc. The C1 tools ought to be more stressful since some of their AI-models are targeting small eyes, pupils, scaleras e.t.c. So DXO has som optimization work to do too.
3 Likes
Happy New Year
600 Posts !!!
Let see what the final result will be.
2 Likes