I was surprised to see that with 6.10 it takes 3 to times longer to export with DeepPrime.
With my PC config:
- exporting with 6.9 and DeepPrime takes about 45 seconds.
- exporting with 6.10 and DeepPrime takes about 2 min 35 seconds !!!
On my laptop I have an Nvidea GTX 950M, I understand that this is an older GPU and according to the release notes the minimum recommended GPU is RTX2060.
With the Windows task manager I noticed that with:
6.9 my CPU (i7) is almost not used and my GPU is at 70 to 80 %
6.10 my CPU (i7) is at 95% and my GPU is at 0%.
This means that my GTX 950M is not recognized and not supported anymore with 6.10.
I understand that with major release upgrades the support of GPU and other system requirements may change but with point updates I would assume most of the system requirement will be the same.
I am using DXO Photolab since release 3 and I am very satisfied with this software.
I also use DXO Filmpack and Nik software.
I have the same problem with V6.8. It fails to use my NVIDIA GeForce GTX 1070 GPU. DeepPRIME and XD use the CPU only for processing which, like your experience, increases the processing time very significantly.
V6.9 and v6.10 fail to export DeepPRIME and DeepPRIME XD to disk which means they are completely unusable.
I have reported these issues to DxO Support and they have passed on the issue to the development team. I suggest you do the same as John_M suggested.
This seems to be an issue with a growing number of users. DxO Support have confirmed that as my GPU is seen by the program it should work.
Until it is hopefully sorted out I have reverted to v6.7.
DxO is clear about minimum system configuration.
Here it is for v6.xx and v7.xx :
• For DeepPRIME and DeepPRIME XD: NVIDIA RTX™ 2060, AMD Radeon™ RX 6600 with latest drivers
Don’t ask any update for older cards.
NVidia older cards does not have tensor cores. Those are specialised for IA computation. Any software (software, not toy) I know which use intensively GPU computing does not support anymore NVidia v10xx serie. Some of them does not even work at all anymore with them.
There obviously are good reasons for that.
My card has worked absolutely fine with all versions up to and including v6.7 . Versions v6.8, v6.9 and v6.10 all show the card as fully compatible in the setup dialog.
DxO support have confirmed that although my GPU now falls slightly below the recommended minimum spec, there is absolutely no reason why it should stop working with V6.8 to v6.10.
DxO Support have confirmed they believe it is a bug which is why they have passed the issue to the Development team, together with the dump files I have supplied.
While it is acceptable for DxO to remove card compatibility from a new commercial release of the program (e.g. v7), it is not in my view OK to do so in a routine update of a version people have already bought.
Unless people go to the trouble of saving every update installation program, there is no way of rolling back to an earlier version. That can result in a completely unusable program which in my case is V6.9 and v6.10 which fail to save anything selected with DeepPRIME or XD.
Task Manager shows V6.7 is making full use of the GPU.
Those times are perfectly acceptable to me and I use DeepPRIME as my default because it is generally superior to HQ and only has a 9 second time penalty.
V6.8 on the other hand fails to use the GPU at all, even when it is selected as the default in setup. The same R5 RAW takes over 9.5 minutes to save DeepPRIME XD because it is only using the CPU. That time is totally unacceptable to me.
My growing concern is that DxO have deliberately modified the DeepPRIME and XD coding in version V.8 onwards which means I (and many others) will not have access to any future maintenance updates (particularly since, in my case, v6.9 and v6.10 fail to save any images processed with DeepPRIME or XD).
Well, I’ve been wondering … as when comparing the specs, your card should be fully acknowledged by PL, but I never took notes of the process’ speed and so I cannot say anything about how it was in v6.7 or else.
The comparision with DeepPrime export showed, that PL6 and 7 work faster than PL5
And it indeed is. My GPU is shown as fully compatible in the setup menu.
I was an Early Access Member during the development of V5 and V6. My tests of DeepPRIME and subsequently DeepPRIME XD during those times showed that my performance results were amongst the best being achieved.
I recognise the card is now beginning to show its age, but my Topaz Sharpen AI and Megapixel AI programs use it with no problem and achieve reasonable processing speeds.
I just hope it is a bug and not deliberate policy by DxO.
I use an Nvidia GTX 1050ti on my aging Windows 10 machine. It is a much less powerful card than your GTX 1070. In the Tech Power Up database your card has 215% of the processing power of mine.
I have had no difficulties processing DeepPRIME or DeepPRIME XD with my card. The performance is identical in PhotoLab 6.7, through 6.10, as well as in PhotoLab 7. Since my card is older and less powerful, my process timings are certainly significantly slower than any state-of-the-art card, but given the age of the card they are respectable and acceptable to me.
I regularly process DeepPRIME in around 13 seconds per single image, and DeepPRIME XD in around 34 seconds, regardless of which version of PhotoLab I’m using. When I batch process 50 images, the average timing is around 2 seconds per image faster because batch processing bypasses the initial startup overhead after the first image.
My files are from my 21 megapixel Nikon Z fc, and thus are much smaller than yours, however my point is that I see no change in performance on any recent version of PhotoLab. As a result I think I can safely say that there is no deliberate policy from DxO.
As to whether there is a bug somewhere, or your problem, as well as other users with a similar slowdown, is a result of some other hardware or software interaction, I can’t say. As for the latest versions of both PL6 and PL7, I can say I am having absolutely no performance slowdowns of any kind.
I am another one with the same problem. I’ve reverted back to PL6.7 for my day to day processing which always worked OK(ish). PL6.8, 6.9 and 6.10 always caused problems when exporting. My current system is a laptop with 20GB RAM and an onboard NVIDIA GeForce MX350 and that has been perfectly adequate for my needs to date. I note thought that the same DeepPrime XD image processed in PL6.7 using the Nvidia GPU to the full takes 26s and in PL7 which ignores my Nvidia GPU but uses an onboard Intel Iris GPU it takes 32s. Not such a big performance penalty considering. Sadly, to gain the full benefits going forward though I will eventually be forced into replacing my current set up with a whole new system. After all it is well past it at 3 years old.
FWIW, this thread spurred me to update to 6.10 (from 6.4). I made a full system backup first just in case things went south. I’m happy to report no issues thus far (Win 10, NV 1660S); my processing times for DP XD did not change. At ~39.5 seconds per Nikon D810 image I really did not want to see a worse result.
Well, all I can then say – make sure to use the latest studio driver.
My machine just got “5” in spring and it definitely has to serve at least til the end of Win 10 support, which is another 2 years from now. I’ve no plans to upgrade my graphic card – just for the sake of export …
To give you another idea … @mwsilvers mentioned 13 and 34 sec with 21 MPx from Nikon Z fc,
while from my Nikon D750 with 24 MPx it takes 10 and 23 sec – with DeepPrime and DeepPrimeXP.
my 10y+ Lenovo W520 & NVidia Quadro 2000M takes 5 min to export with DeepPrime (PL 5)
Do I upgrade? No, not necessary for me.