done #421646
Lot of hang of PhotoLab since update to 6.7, most of the time when resizing a picture.
Also more than 50% of error when processing pictures in batch.
AMD Ryzen 3800 and Radeon RX570, 32GB RAM
Doesn’t the following explain what’s happening ? (I don’t follow AMD graphic cards technology that much, only NVidia so I don’t know what are real differences between the cards we’re talking about) :
Photolab release Notes Version 6.7
Recommended system configuration :
NVIDIA RTX™ 2060, AMD Radeon™ RX 6600 or better with latest drivers
I’ve just seen it Yesterday, so I’ll buy a new graphic card RX7600. Still AMD.
It should be speedier. Two to three times the old one.
I’ve seen memory leak on the RX570 with DXO Photolab around 50MB for each processed pictures.
PhotoLab seems to hang only when DeepPrime XD is selected on my computer.
With just DeepPrime there is also less error during the batch processing.
I’ll will send result after GPU change.
I have been processing a batch of jpgs from a party had twice had PL first lock up then crash using smart lighting spot selecter. Current studeo driver.
For example, NVidia serie 10xx does not have tensor cores (AI dedicated), tensor cores appears since serie 20xx, so different architectures can be important for denoising and maybe more for the last algorythm (deepprime XD) to work properly.
Does your station meet DxO hardware specifications ?
More than
AMD Ryzen 9 3900X 12-Core Processor, 32Gb , GetForce GTX 1650 Super all SSD drives
1 Like
They specify 20xx serie for last Photolab version not 10xx serie.
GeForce GTX 1650 does not have tensor cores. Only cuda cores.
Right
but its never had any problems befor
1 Like
Before what ? Before V6.7 ?
Untill today, I was thinking of changing it to version 3 card so might give me the excus to do it
Hoping that’s really the problem.
But it probably is. Very few people have reported this.
I stil have 6.3.1 installed.
Now got a RTX4070 fitted. Export is a bit faster but not greatly so. Probably more marked with bigger RAW fills than the Sony a64000. Few years since I changed a video card, back then was a circuit board often powered direct from mother board or a extra twin cable plugged in to the big ones. So hopefully will keep on top of hardware needed for a few years, thats if it’s worth upgrading PL in future with lack of phone support.
For those having problems - NVidia released new studio drivers earlier today (29 June). Not sure if it will help or not.
1 Like
Yes it could maybe help. DXO mentions to update to latest drivers when you see splash screen.
Note for studio drivers:
536.40 WHQL Known Issue This driver implements a fix for application stability issues seen during heavy memory usage. We’ve observed some situations where this fix has resulted in performance degradation when running Stable Diffusion and DaVinci Resolve. This will be addressed in an upcoming NVIDIA Studio Driver.
Maybe but not sure.
Unless DxO colaborate with nvidia they probably did not use those drivers when (for) developing tools done before those drivers are released.
1 Like
Its possible yes. Although its also possible that neither DXO nor nVidia had enough of testing and if for most users and most situations it works, no one noticed. But for some smaller number of users it could be a problem. Eventually these issues will probably be ironed out.
As the note release of the latest driver says: " This driver implements a fix for application stability issues seen during heavy memory usage. " but I could not find the list of applications that were reported to have stability issues. It may or may not include PhotoLab. Best to test it out.
1 Like
I’ve just install the Radeon RX7600 in place of the old RX570.
No more hang nor crash of the interface.
No more error during batch processing in using Prime denoise, nor DeepPirme, nor DeepPrime XD !
No more memory leak at the GPU.
An average around 8 sec per 45Mpix pictures and 3 formats export for a batch of 160.
157 Pictures processed in 20 min 47 sec without any error !
1 Like
Graphics cards have become what cpus were. They evolve the fastest and give developers the opportunity, year after year, to produce algorithms that were considered unusable just a few years ago. But this means that users have to keep up with the evolution of hardware in order to make full use of the evolution of their software.