Processing RAW > JPG in 7 takes a really long time...
Hi there,
Prospective Capture One purchaser here. I'm testing out Capture One 7 because I was interested in its tethering abilities.
However, I did some very basic benchmarking, and have noticed something very strange.
I imported a bunch of RAW files into Capture One, and without making any edits or changes, processed 10 of them as jpgs. The whole batch took 6 minutes and 15 seconds, which is about 37.5 seconds per image.
I did the same thing with LR4, and managed to export the same 10 images to jpgs in only 1 minute 12 seconds (7.2 seconds/image)
Like I said, I'm a total Capture One noob. Am I missing something obvious? I am honestly ASTONISHED at how much slower Capture One is. If it's really going to take 5 times longer to process each image as Lightroom, there's no way I can make this work.
It can't be that bad, can it?
Prospective Capture One purchaser here. I'm testing out Capture One 7 because I was interested in its tethering abilities.
However, I did some very basic benchmarking, and have noticed something very strange.
I imported a bunch of RAW files into Capture One, and without making any edits or changes, processed 10 of them as jpgs. The whole batch took 6 minutes and 15 seconds, which is about 37.5 seconds per image.
I did the same thing with LR4, and managed to export the same 10 images to jpgs in only 1 minute 12 seconds (7.2 seconds/image)
Like I said, I'm a total Capture One noob. Am I missing something obvious? I am honestly ASTONISHED at how much slower Capture One is. If it's really going to take 5 times longer to process each image as Lightroom, there's no way I can make this work.
It can't be that bad, can it?
0
-
After browsing the forums a bit and seeing people complaining about speed/performance in 7, I tried an older version. Just tried the same experiment in 6.x... and it only took 2 minutes 25 seconds to process the same 10 images.
Considerably faster than 7, but still twice as slow as LR4. Is that just the nature of the beast?0 -
As you might have discovered while browsing around here, CO7 and its new CO7 engine performs a much more complex computation for output compared to CO6 (and LR4 I assume). This should be reflected in the end results (in better image quality).
So yes, it is slower on the same hardware. And yes, it can take advantage of newer hardware, in particular with OpenCL support. OpenCL support was extended to processing in CO7, besides display support from CO6.
My assumption is that with proper OpenCL support, processing in CO7 can be much faster then CO6 or LR4 (both without that kind of support).
So if you want to do the math, it is what you are comparing, in terms of quality and quantity.0 -
[quote="convoke" wrote:
Hi there,
Prospective Capture One purchaser here. I'm testing out Capture One 7 because I was interested in its tethering abilities.
However, I did some very basic benchmarking, and have noticed something very strange.
I imported a bunch of RAW files into Capture One, and without making any edits or changes, processed 10 of them as jpgs. The whole batch took 6 minutes and 15 seconds, which is about 37.5 seconds per image.
I did the same thing with LR4, and managed to export the same 10 images to jpgs in only 1 minute 12 seconds (7.2 seconds/image)
Like I said, I'm a total Capture One noob. Am I missing something obvious? I am honestly ASTONISHED at how much slower Capture One is. If it's really going to take 5 times longer to process each image as Lightroom, there's no way I can make this work.
It can't be that bad, can it?
That sounds very slow. What kind of files, what kind of hardware (processor, gpu and hdd)?
Please try writing a tiff instead and see if that helps for speed (just for troubleshooting).
As Paul mentions, we can use your Graphics card (if it is fast enough). This way we can process a 80 mp IQ180 file in a few seconds. No need to mention that a way smaller dslr file will be much faster.0 -
[quote="convoke" wrote:
...
I imported a bunch of RAW files into Capture One, and without making any edits or changes, processed 10 of them as jpgs. The whole batch took 6 minutes and 15 seconds, which is about 37.5 seconds per image.
I did the same thing with LR4, and managed to export the same 10 images to jpgs in only 1 minute 12 seconds (7.2 seconds/image)
....
I use both CO7 and LR 4.4. I find each to operate abot the same on my Win 7 machine.
LR and CO also operate a bit differently as I understand things. LR will look for processing instructions, and if you didn't give it any, it won't spend anytime in reprocessing the file. I believe CO will.
I'd recommend trying your comparison with edited files. Change the same parameters in both.
But I also agree - 6 minutes seems quite long.0 -
Alright. So granted, this was running on an older, secondary machine. But so was LightRoom. Here are the specs:
CPU - Intel Core 2 Duo E6400
RAM - 8 GB
HDD - SanDisk 256GB SSD
GPU - ATI Radeon HD 5700 series (I think it was a 5770)
Do I have to do something special to allow Capture One to utilize the GPU?
I will try the test a few more times, modifying the files slightly, and also processing as TIFFs to see if that makes a difference.
Thanks for all the support, guys.
Edit: Oh, also the source files were from my Canon T2i.0 -
Hi,
I'm not sure if there may be a relation to the commented supposed slow performance but I noticed on my Win notebook more recently aquired for running C1V7 (i7, 16GBRAM, Nvidia Geforce 670MX) that sometimes the processing of selected (edited) images doesn't start immediately. No more importing, proxy generation or something else happening which could interfere with the image processing.0 -
[quote="convoke" wrote:
Alright. So granted, this was running on an older, secondary machine. But so was LightRoom. Here are the specs:
CPU - Intel Core 2 Duo E6400
RAM - 8 GB
HDD - SanDisk 256GB SSD
GPU - ATI Radeon HD 5700 series (I think it was a 5770)
Do I have to do something special to allow Capture One to utilize the GPU?
I will try the test a few more times, modifying the files slightly, and also processing as TIFFs to see if that makes a difference.
Thanks for all the support, guys.
Edit: Oh, also the source files were from my Canon T2i.
The Core 2 Duo is actually the minimum requirement for CO7. That means it'll run, but it won't be fast at all.
It's the same with the gfx adapter.
If you upgrade to some more current hardware, you'll see a major improvement. With a newer gfx card (such as those from the ATI HD 79xx series or Nvidia Gtx 6xx series, you would also see a huge improvement (when it comes to processing and adjustment stuff)0 -
Hi Christian
following up on your comment about improved speeds of CP1 using later (newer) graphics cards ..i've a I5 760 CPU with 16 gig ram..a 250 GTS card...up until now Ive been unable to use open CL for processing , but as of the latest release i'm happy to say i'm now able to use open cl for both processing and display . ive the latest nividia driver (3.20) installed...
my question is if I upgrade to a newer 6xx nividia card , what sort of speed improvement would I see in adjustments and processing...I also run a 120 GD SSD for OS and programs...but the image files are on a larger standard HD0 -
I saw your extended times and ran the same test processing jpeg images. I copied the same RAW file (25.7MB) 9 times and used Capture One 7 to reset the images and process them to the same directory (2TB HD) as jpeg images (12.2MB). Total time was 23 seconds. I reset the RAW images again and processed them as TIFF images (120MB) and the total time was 24 seconds. As with your test, these images did not contain any adjustments which would have extended the processing time.
Phase One's new processing engine puts a heavy demand on your computer system.
I'm running an AND FX8350 processor on an Asus 990FX Re Motherboard with 16GB RAM and an AMD 7770 Graphic card (OpenCL).0 -
[quote="Tommo1965" wrote:
Hi Christian
my question is if I upgrade to a newer 6xx nividia card , what sort of speed improvement would I see in adjustments and processing...I also run a 120 GD SSD for OS and programs...but the image files are on a larger standard HD
That really depends on what card in the 6xx series you buy 😉0 -
[quote="Glenn101" wrote:
I saw your extended times and ran the same test processing jpeg images. I copied the same RAW file (25.7MB) 9 times and used Capture One 7 to reset the images and process them to the same directory (2TB HD) as jpeg images (12.2MB). Total time was 23 seconds. I reset the RAW images again and processed them as TIFF images (120MB) and the total time was 24 seconds. As with your test, these images did not contain any adjustments which would have extended the processing time.
.......
I have been curious about all this, and decided to make my own comparison; except with files that had some processing. I also compared CO 7 with LR 4 (latest versions of each).
Summary - Light room is very consistent between jpegs and tiffs. CO 7 is inconsistent, seemingly doing some processing in the background.
Exp - I also started with a Canon crw raw file of 25.7M, and I made 9 copies, for a total of 10 files.
Processing : exposure, levels, sharpening, noise reduction, clarity, wb, and lens correction, incl ca. There was no hdr or curves or cropping or local adjustments. I duplicated the CO and LR as closely as possible. OpenCL was in "auto".
I repeated this with another 25.7M raw file.
In each case, I did a system restart, and only opened the one application to be tested, giving full resources to it.
I then made 10 full size tiffs and 10 full size jpegs at 100%.
Lightroom: 10 tiffs in 1 min 13 sec to 1 min 15 sec.; 10 jpegs in 1 min 19 to 1 min 23 secs.
If I exported the tiffs again, I got the same result. This isn't surprising, as each pass is exactly the same; read an instruction, then execute it.
Capture One was strange, and I surmised that it was doing a lot of processing in the background; esp if I waited a bit.
I could output 10 jpegs in 1 min 58 sec to as short a time as 19 sec. The 19 sec came after waiting for about 1 min and doing nothing. If I adjusted some parameters like exposure, and quickly did a export, I saw the 1 min 50 secs again.
CO and tiffs are a similar but different story. I had exports ranging from 3 min 35 sec to 23 secs.
I started to follow the processing with Win task manager.
For the 3 min 30 times, CO seemed to stop working after 8 or 9 exports, waited, and then finished the last couple.
When it didn't pause, I saw conversion times of about 1 min 30 sec the first time through.
I I repeated the export, I saw a 25 sec time for 10 files.
If I made some processing mods, quickly did an export, the time would be about 1 min 20 secs or so.
My concern was with when the exports seemed to pause. I tended to see this on "quick exports"; eg, change a set of parameters and quickly export. I didn't experience this if I waited a bit, such as looking at photos and rating them; I am guessing that a lot of processing is occuring in the background as I can see in the task manager.
Bottom line to the OP - CO and LR are about the same for jpegs, and CO can be faster or slower for tiffs, depending on the conditions.
Edit - for what it's worth, all the software is on a ssd; the LR catalog and the photo files, including the processed ones, are on an internal 7200 hd.
Edit 2 - again for what's worth - CO was run in a session mode. I brought one set of files in via Media Pro.0 -
[quote="Christian Gr" wrote:
[quote="Tommo1965" wrote:
Hi Christian
my question is if I upgrade to a newer 6xx nividia card , what sort of speed improvement would I see in adjustments and processing...I also run a 120 GD SSD for OS and programs...but the image files are on a larger standard HD
That really depends on what card in the 6xx series you buy 😉
either a new 660 or a used 670..perhaps a 6800 -
Interesting observations.
I have been running some large batch processes recently. Each process running 2 recipes, one to output 750px max width jpgs and the other 1024px, both with compression quality set to about 60%.
In each case it is safe to assume that all files had significant edits (Exposure adjustments, crops, rotation, sharpening, HDR slider tweaks, a few with local adjustments, etc. but no colour manipulation).
The first batch a few days ago was, iirc, 1385 files and took 44:44.
Last night the batch was 1410 and took about 45mins befoe the machine went into hibernation with 44 images to go. On awakening the machine the batch finished in about 3 mins.
At the start of a large batch there does seem to be some pre-processing going on to prepare the batch. For both of these sessions there was a delay of several minutes before the batch list appeared and processing started. I assume this is just time required to create and submit the batch and decide how best to structure the task when there are 2 recipes in play.
I have run other batches of a similar nature and obtained about the same times (I have not checked for accurate details) so the seems to be consistent.
Once or twice in the past few days I have experimented with deep crops and significant edits of single files with output to TIFF and upscaling. The results were multi-magabyte files produced almost instantly. Taking the first output file and repeating the process (with minor additional edits) saw a 200Mb Tiff produced in a couple of seconds. Frankly I was amazed. My initial reaction was to assume the process had failed.
However the system does have a 512Gb SSD (a single drive for everything) and the sustained R/W speeds are likely to be at or near the top of the current performance envelope compared to smaller drives. Most of the 512Gb drive specs seem to suggest that they are top end performers whereas the 240/256 Mb sizes can be a very mixed bag and the smaller drives often (but not always) top out at around half the claimed maximum sustained write speed of the largest. mSATA drives seem to be constrained to SATA2 speeds rather than SATA3.
If using external drives there may be similar issues - external USB3 drives, for example, seem to come in 2 specs with "Pro" versions (usually 7200 rpm drives) claiming about twice the speed of the non-pro versions (usually 5400rpm drives). Last night I backed up a 36Gb session to 2 external drives - a 2Tb USB2 full size unit running (iirc) 7200rpm and a 1Tb 2.5in drive USB3 unit running 5400rpm. Both devices were running on USB3 ports on the system. The backup time estimate calculation was the same in each case, the actual time taken somewhat less than the estimates and although I have not done detailed time comparisons for these activities (it was not a controlled experiment) the overall time in both cases was about the same. To put that another way, neither was so fast or so slow as to be a surprise.
Reading of the results that people obtain from different but similarly specified machines I suspect that having the "right" components is only a part of the solution to excellent performance and that the bits that blend them together may be equally or more important. (Rather like tuned gaming systems I suppose ...). That means that obtaining meaningful (ie. comparable) effective performance information that might be of use to the software developers for application code tuning or revising database settings may be a lot more difficult than it might seem.
A few more observations, for what they are worth.
Grant Perkins0
投稿コメントは受け付けていません。
コメント
13件のコメント