Comparing RAW conversion speed; RT/DT/ViewNXi/CameraRAW

Software RawTherapee
1

I’m still figuring out what is the most capable, handy, pleasant and fast(est) RAW converter. Not all qualities are equally important. Although I like the RAW conversion quality of Nikon’s ViewNX-i; it’s speeds, or better is slowness is becoming more and more annoying.
Originating from the Nikon ViewNX-i RAW converter, I’m still searching. The last few weeks I’ve compared RawTherapee 5.3 ‘Alberto version’, Darktable 2.4.0, ViewNX-i 1.2.10, Adobe CameraRAW 10.1 (with OpenCL GPU support on) conversion speeds, with some interesting results.

System I use: a Windows Pro 64bit 1709, 24GB RAM, i7 980X CPU, AMD R9 270x GPU with 2GB RAM. SSD system disk, HDD data disks (SATA 3)

I’ve setup a test directory with 200x Nikon D800E NEF (RAW) files. Total 10.5GB of RAW data.
Converted to 8 bit TIFF files, sRGB without any further special adjustments.
The resulting TIFF files are about 106MB each, total output data approx. 20.2GB

ViewNX-i inputdir on HDD1 → outputdir on HDD2 – 24 minutes
ViewNX-i inputdir on SSD1 → outputdir on SSD2 – canceled after >35minutes (! very strange)

RT 5.3 inputdir on HDD1 → outputdir on HDD2 – 12 minutes
RT 5.3 inputdir on SSD1 → outputdir on SSD2 – 12 minutes (no difference with HDD)

DT 2.4.0 crashes after a few minutes, tried several times. :confused:
darktable.exe caused an Access Violation at location 000000006C68FA9F in module libexiv2.dll Reading from location 000000001A8260C0.
Fault bucket , type 0
Event Name: RADAR_PRE_LEAK_64

CameraRaw 10.1 inputdir on HDD1 → outputdir on HDD2 – 7 minutes
CameraRaw 10.1 inputdir on SSD1 → outputdir on SSD2 – 7 minutes

So Nikon’s ViewNX-i is definitely the slowest converter. RawTherapee is about twice as fast, and Adobe’s CameraRaw is even faster. I think the OpenCL GPU acceleration does a good job for CameraRaw. Just a pitty that RawTherapee doesn’t use GPU acceleration yet… (hint, hint :wink:slight_smile: )

Remarkable is also that the conversion speed is not depending on the type of storage device! I can’t see any difference between SSD or HDD conversion speeds. Although when I do a straight copy of the files, the transferspeed difference is very obvious! (of course) and SSD is lot faster.

I can’t judge the TIFF output results from each converter yet. That will take some time for me to figure out what I like most/best/better…
As regarding to each GUI:
I still like the rather ‘simplicity’ of the ViewNX-i GUI, just some basic settings, nothing overcrowded, just a few clicks and off you go with the TIFF conversion.
The RT GUI is loaded with features; that’s a good, but also a confusing thing. You don’t know where to start/look/click.
The same applies for darktable.
And as for CameraRaw; I don’t especially like that ui either. Although I’m familiar with Adobe since PS2.

2

It’d seem that you’d want to turn off all image processing options in RT to truely test its speed, I’d expect that even applying the auto exposure, as it does by default when you open an image, would slow down the raw conversion as it needs to compute those things to export.

Also we don’t really know what CameraRaw and the Nikon software are doing behind the scenes either.

It isn’t clear that your comparisons are apples-to-apples.

1 Like
3

You’re absolutely right! Just comparing windows-to-windows…:smiley:

I’m just comparing the basic setup, installed the programs, filled a testdir with lots of RAW data and off I went. Just like an average daily user would do. Daily practice. I think it does give a reasonably well impression of what those RAW converters can do.

4

Turning it off would be unfair, as ACR applies an exposure curve and you can’t turn it off. It does other things you can’t turn off, making this test produce a figure in a very large ballpark.

5

More like comparing person-to-person. You can never know a group of people well enough to truly know what they are capable of – virtues and vices. That is what makes people interesting :wink:. Each Raw processor has its pros and cons, though on this forum we prefer the FOSS options.

6

You might consider adding dcraw to your comparisons. A version compiled with OpenMP was briefly distributed with ImageMagick, and it was fast but not reliable.

7

I typically export to 8-bit JPG or 16-bit TIF, rarely ever 8-bit TIF. You might want to try different settings (file type, bit depth and compression) as they would impact your export time.

8

Indeed.
And I just compared the speed factor for a number of RAW converters. Although it might be a non-scientific executed test, however it does give a good idea of the differences in conversion speed between the mentioned programs. Coming from a slow one (ViewNX-i), every other one I’ve tested is at least faster to do NEF → TIFF conversion.
I’ve also done some tests with an older version of CaptureOne (v 9.x) and C1Pro was even faster. But that is also, imho, a reasonable expensive product.

BTW: I don’t prefer or detest FOSS :grin: (or any other nicknamed s/w); I only prefer that piece of software which does the job for me. The ‘best’ and ‘free’ RAW converter would give me excellent results, like ViewNX-i, being user friendly and uncomplicated to use, runs stable and fast. Oh… and it should be self explanatory, without struggling through hundreds of scattered wiki pages, having to watch numerous tutorial videos just to figure out what they mean by a specific term or word, etc.
Yes, I know… I keep on dreaming. :thinking::smiley:

9

I assume that you would see a difference here if the cache and similar paths point to a SSD. However, your measurement seems to show, that 90% or even more of the overall time is not I/O related …

1 Like