Why is PL2 so slow?

(Alec Kinnear) #21

FRV is very ugly in its default implementation and chooses to work with RAW files. FRV is actually faster and better if you choose to shoot companion full size or near full size jpegs and use those for the previews. Very snappy. To really know how a RAW will react to processing one has to either 1. have experience or 2. actually open it. The jpeg preview gives ample info.

This is double true for Fuji X files (my carry camera is an X-T20 which is a dream to shoot with manual lenses like my old Leica R and Nikon AI and Takumar lenses), not that Fuji X handling will matter much to PhotoLab users (still hoping for PhotoLab support for Iridient Transformer’s demosaiced files). In the meantime, Iridient Developer does a decent job with the Trans X files.

Anyway with FRV, the trick is to get rid of everything you don’t need from the interface, rearrange the panes, strip it bare. Cleaned up, it’s not much worse looking than the very beautiful PhotoLab. You can run FRV with just a single info line at the bottom of full screen true 4K previews (I usually keep a single pane open with histogram though but not when working through events where exposure levels don’t change much).

#22

The purpose FRV is to view RAW files, especially from the camera card.
Viewing embedded jpgs defeats the purpose of FRV.

(Alec Kinnear) #23

The purpose of FRV is to evaluate large quantities of originals quickly and efficiently, without paying $150 for PhotoMechanic (even clunkier cross-platform interface than FRV but with more powerful keywording abilities and even faster) or $295 for an iMatch 5 pack (Windows only so out of the question). Fast Image Viewer is also Windows only.

I know my cameras so if I took the trouble to shoot RAW + JPEG (I used to shoot RAW only), it’s much faster to use the full size JPEGs in FRV and more efficient as it’s easier to judge focus and sharpness from the JPEG preview than the RAW amazingly enough.

To fully judge a marginal RAW file’s potential you need more than a RAW viewer. It requires opening it up and pushing it a bit in your RAW development tool. And who wants to spend his or her time on marginal files anyway? JPEG previews are far more than enough to judge winners.

FRV’s winning feature is that it doesn’t depend on a database but stores its data in XMP sidecar files which can be moved from disk to disk with the images and which DxO PhotoLab among other RAW developers reads with no issues.

What’s a bit sad about PhotoLab is that it does not write back any changes to evaluation marks to the XMP file. If you make changes to your evaluation in PhotoLab, it’s a one way street only for now.

(Martin Williams) #24

I’ve a 2012 iMac, Core i5, 2.9GHz, 24GB RAM
See changes pretty much instantly; files around 15MB

Wonder about things like spare capacity on main drive; and if you’ve used something like Onyx lately - perhaps all too obvious, and affecting speed for other software too.

(Alec Kinnear) #25

Unlikely. I’m a software architect and IT expert. My only slow program is DxO PhotoLab. Looks to me like PhotoLab chokes with 4K monitors and large files (as I use a 4K monitor as primary monitor on both systems). No reason for PhotoLab to choke on files like this or 4K monitors in the face of adequate hardware (see above). It looks like PhotoLab just isn’t optimised for those two conditions, at least concurrently:

  1. 4K monitors
  2. large files

Could easily be solved by any combination of hardware acceleration, restricted processing (just the visible area) or proxy rendering (slightly less accurate but much faster). The ability to work with real time sliders or more accurate preview would be a great improvement for me in workflow. I’d do most processing in real time, switching over to accurate preview just at the end.

This kind of important issue with core processing - speed - is exactly why DxO should consider carefully whether they should be looking for new fronts on which to start wars (DAM, soft proofing) but should really focus on their core value proposition: fast, efficient and amazing RAW processing.

#26

Sounds like a reasonable assumption. After all, 4K monitors require a lot of data to drive them properly. Add to that the data in a large file and it may be “a bridge too far” for some machines.

(Platypus) #27

This test was done to clarify what bugs me most: GUI delivery times in CUSTOMISE view

Just did the following: flipped through the film strip as quickly as possible using the arrow key.
With DPL, it takes 130 seconds from the start to the end of the set of 60 images.
With DOP11, the same procedure takes just 25 seconds and I can click any random image and DOP will immediately select the one I clicked

With DPL, I have to wait until all images have been selected one after the other, random clicks are just ignored until all keystrokes have been worked.

Compared to CUSTOMISE, Catalog view is fairly snappy

I’ve now removed DPL 2.2.0.22 because I get too many beach balls and will test again on the next release. (DPL 2.1.2.25 is slow too, but fairly stable)

1 Like
(Christian) #28

I see similar behavior in DPL 1.

It´s annoying when your computer is not very fast. The app could just abort the current preview rendering if the user moves to another picture or changes sliders.

(John Barrett) #29

Dosdn’t that leave the information in the PL data base still?

(Alec Kinnear) #30

I don’t think this is an issue of the speed of the computer. These are poorly optimised core routines which will affect just about everybody, regardless of GPU or number of processors. I’m astonished DxO is adding features which slow the program further (slowness is what everybody hated about Lightroom 3 and 4 and 5, versions 1 and 2 are pretty spritely) like DAM before fixing core routines, improving hardware optimisation (particularly on the GPU) and adding newly released cameras.

If PhotoLab becomes this impossibly slow eight armed monstrosity, photographers will not be able to afford to use it. Slow post-production is both psychologically frustrating and a great thief of time.

If PhotoLab were the fastest post-production tool (C1 or Lightroom would win here now, both are pretty quick, sliders are mostly in real time) that would be an enormous win for the existing userbase and very attractive to new users. “Powerful but slow” is not attractive marketing for photographers. Most of us feel too much of our lives is lost in the digital darkroom already.

8 Likes
Adjusting sky colours
(Platypus) #31

Completely agree, Alec :+1::+1::+1:

1 Like
#32

Thanks for your reply Alec, i hope it will be fixt

1 Like
(Garret) #33

Alec, great post! I totally agreed.

I use DxO PL2 Elite, DxO FP 5 Elite, and DxO VP 3, and they develop images remarkably. However, my user experience has been very frustrating. I just bought a new MacBook Pro with max specs last month including the new Vega 20 GPU. Also I am now using a brand new Blackmagic eGPU Pro that I got yesterday from Apple (yes “Prefer eGPU” is on for all three apps and everything is fully updated). I performed a Geekbench on this new set up (Open CL on GPUs):

Single-Core CPU 5639

Multi-Core 24997

Intel UHD Graphics 23058

Vega 20 80453

eGPU Pro 156685

Blackmagic Disk speed test >2700MB/s read and >2600MB/s write!

In short, a mobile Mac photography powerhouse it would seem.

The folder in question contains a lone 9 year old Canon 5Dm2 image that I am testing on this fresh and updated configuration. Alas it is slow, slow, slow! The sliders are as sticky as LR or Aperture were on day one with 12+ year older hardware on first their releases. Even an export of a single image (of course using Prime NR and the full gamut of adjustments) took 33 seconds. What is really fascinating is that my top spec 2015 5k iMac is much faster with slider adjustments and 34s on the export (of course same image in a folder by itself for comparison’s sake). Both setup’s have 32GB of RAM, but I shut down all other programs for this test making far excess free RAM available. The iMac’s benchmarks are:

Single-Core CPU 4958

Multi-Core 16679

Radeon R9 M295X 4GB 85413

Black Magic Disk speed test ~640MB/s read and ~610MB/s write

From practically any point of view the new setup should mop up on the old one— but in fact the old one performs better while developing and requires about the same time to export. My speculation at this point is that DxO uses the Intel GPU on the MBP despite the two other far better GPU’s being available. Anybody have any idea what is going on?

Great photography software unfortunately just does not exist these days. I loved the UI, library organizing tools, places, faces, print, share, plug-ins, and —most of all—the fluid speed and rock solid nature of Aperture, but sadly Apple killed it. Also Aperture never did deliver the goods image-wise like DxO (nobody else has either). Unfortunately, DxO has always been clunky and awkward to use (e.g. does not even support multiple monitors) and lacks even a basic library management feature set much less facial detection, mapping, 3rd party tie-ins, etc. It is a great shame that Steve Jobs died so young; it’s tough for me to believe that he would have allowed Apple’s Pro Line to be ruined. I would gladly pay more for software that would fulfill on all fronts and really take advantage of the latest hardware!

Cheers,

Garret

2 Likes
(Platypus) #34

Indeed, programming can burn up system resources easily when done badly…

#35

I have a 2018 15" MacBook Pro, although perhaps not as highly spec-ed as yours. I keep my images on an external ssd so I do not have the IO speed that you do using your internal ssd.

Still, I decided to test PL2 with one of my old 5D3 raw images to see how it compared to your results. As part of that I made all of the adjustments I could, keeping the image reasonable, and set Prime NR even though the image did not need that. The output took just under 40 seconds, but that was almost all Prime NR as changing to HQ dropped that time to below 8 seconds. Since I almost never need Prime I do not have an issue with how long images take to export.

As for “great photography software”, I guess I might point to CaptureOne. I almost always use Photo Lab for my processing but I do have last year’s version of CaptureOne and I think it is probably better than Photo Lab in almost all respects except for several PL exclusive adjustments - mostly Prime NR and ClearView. CaptureOne has the best Color Editor I have seen in any processing software and I think it comes as close to “great” as any workflow tool I have ever seen. It is pricey, too pricey for me as the upgrade is $150, but I think it is much better than Aperture.

As they say, YMMV.

(Alec Kinnear) #36

That’s funny. I own CaptureOne and find the workflow awful (trying to take over your filing system with all its own folders, storing almost all the settings in a database). PhotoLab after one has customised the panels is a single track path to great photos. A photographer need only put his or her panels in the order which s/he likes to apply the effects and work through systematically. Noise Reduction should come last as it slows PhotoLab quite a bit. I leave the Histogram and Level Curves up on deck on the left as histogram should always be visible and curves may need multiple adjustments.

Really fast to work through an even folder, especially if you make a preset for the main looks. Even more important, the quality of output is much better at mid or high ISO. Certainly, the Capture One colours editor is special but it requires a fair amount of training and background to use effectively. And for most photographers perfectly superfluous to developing great photographs, while good quality core processing and noise handling shows up in almost every photo.

The big issue I have with PhotoLab is that DxO is not keeping up with performance for large format RAW (Nikon D850, Z7, Sony A7R I II III, Canon 5DS, 5DS R, Fuji GFX) and/or 4K monitors. It’s not 2015 any more. Decent responsiveness at 1920 x 1080 with 22 MP Canon 5DIII files isn’t good enough in 2019.

BTW, when I speak about performance, I’m always talking about responsiveness. Export time for full quality Prime NR finished files is not an issue for me. I can do my exports in the background and do something else while they process. I can’t get back the five to eight seconds delay dozens of time per image waiting for the preview to catch up with sliders.

1 Like
#37

I too have Capture One but although it’s powerful and have perhaps the best colour correction bar none it’s tideous and time consuming to work with. Its performance is very good.

PL on the other hand produces amazing quality per invested time and have pretty much all tools you need - give it take a few - but it is not the missing tools that is holding my workflow back.

It’s the lack of sheer speed and performance that is crippling itself.

This saddens me. Because no matter how good the tool is - if you constantly need to wait for it even on the highest performaning computers the investment in time and/or money is not paying off.

DxO need to hold back on feature additions and put more manpower on code optimisations as well as sort the backward behaviour when raw render processes are prioritised above almost everything else.
With 14 bit 24 MP photos it’s slow. Can’t imagine how it will be on 36 less alone on 45MP 14 raws.

Please - give us more speed and GPU acceleration.

4 Likes
#38

Agree to uncoy, especially to the point of customizing the panel for a straight forwarded workflow. And yes, there is a lot more of editing on a photo you can do. But not everybody has the need or the time to fiddle an hour or more on a photo just to get a certain result.
And also agree not only for the colour editor of C1, but also generally, that deeper editing, which in my opinion is also possible in PL, „requires a fair amount of training and background to use effectively“. But than again time gets a matter of fact.
And last agree to Required, “Please - give us more speed and GPU acceleration.”

2 Likes
#39

I think we are probably referring to different things here. My praise of CaptureOne has to do with its ability to properly edit a raw image and the ability to adjust settings to produce pretty much whatever a user wants, not its ease of use. It has a superb Color Editor, the best I have seen in photo software, and many of its controls product accurate results with little effort. I find its Levels adjustment to be wonderful.

On the other hand its controls are scattered through different tabs and it is hard to get a single tab organized in a way that is really useful. It exports images quickly, but the configuration of the output is difficult for me to understand. The output folder you specify is not always the output folder you get, its sharpening and NR functionality leave something to be desired and I dislike having to import photos.

In my view the biggest advantage of CaptureOne is its Color Editor, and that is not only because of the quality built into it, but because Photo Lab has no corresponding color adjustment. The Mac version used to have the Multi-Point Color Editor, but it was never very good and I don’t miss its removal, but do wish they would put a decent color editor in it.

As I mentioned, I find that it can produce wonderful results, but there is a big downside in the difficulty of using it, and that is why I have a licensed current version of Photo Lab, but only a licensed previous version of CaptureOne.

2 Likes
(Alec Kinnear) #40

Thanks for the reminder, Mike. I thought I’d seen a full colour editor in PhotoLab and was surprised when I couldn’t find it for the tutorial. I must have seen and used it in a version of OpticsPro I installed or PhotoLab v1. Adding back a stronger colour editor would be a wonderful improvement which would contribute to the PhotoLab/OpticsPro historic core mission (a very fast and easy to use RAW developer of the highest quality).