There's been a lot of talk about who does and does not want / need 36MP and just how much it's going to slow down post-processing. As someone who doesn't like to settle for armchair commentator theories without some data to back it up I decided to do my own tests using Lightroom 4.
I chose 3 different cameras, because I had these files readily available. The D700 is (as we all know) 12MP and the files are all 14 bit. The 5D2 is 21MP but I think it's safe to assume that the 5D3 is probably going to be within 5%-7% of these timings (above). The D800 is 36MP (14 bit).
I'm using a reasonably powerful computer but by no means the best. It's specs are along the lines of:
Intel i7 Quad @ 3.4Ghz
16 GB DDR3 1600 RAM
OS on Drive C
Data to a 6TB Raid made up of 3x 2TB SATA II drives
Video card is a GTX570
The Card reader was a Lexar USB 3.
The card was a Transcend 400x 32GB.
I chose to make timings using 100 random images from each camera so that I could get a reasonable average that could be used to extrapolate other quantities (e.g. 1,000 or more images from a wedding).
So, here goes:
Time to read images from the card to the computer's hard disk
D700: 29 seconds
5D2 : 48 seconds
D800 : 65 seconds
Time to import the files in to an empty LR4 database using the 'Add' function (leave images where they are) :
Now, with these numbers we can reasonably calculate for 'X' images by simply multiplying. There will be variations based on the actual images.
Timing for actual changes to images depends on the sequence of things done and how LR4 is caching images.
To take things to reasonable limits, 1000 images from a wedding would take the following to simply import and export without any real changes being made in LR4:
Thanks for testing and posting this information. Your computer setup is a capable imaging workstation, so it does provide a good indication of comparative throughputs for various file sizes.
Have you tried overclocking your i7? If you have a "k" series processor, it's an almost trivial process to obtain about 25-30% more raw processor speed and it would be interesting to see how the D800 files respond to a faster processor.
Yes, I have overclocked it to 4.3Ghz prior (I have a water cooled system) but I put it back to 3.4Ghz when I was experiencing some lockups and wondered if O/C was the cause (it wasn't). I just never got around to putting it back up.
So, basically the time is close to linear with MPs. D700 (12MPs / 14 bit) takes 2.7 hours and D800 (36 MPs / 14 bit) takes 7.68 hours. So if you triple the MPs, the time gets increased by a factor of 2.84.
After processing my D800 images for the past week I processed a few 5DM2 images today and the speed difference was like lifting a paperweight after handling 50lb dumbbells
gdsf2 wrote:
So, basically the time is close to linear with MPs. D700 (12MPs / 14 bit) takes 2.7 hours and D800 (36 MPs / 14 bit) takes 7.68 hours. So if you triple the MPs, the time gets increased by a factor of 2.84.
Yep, sounds about right given that the tasks described by Dave are all bound by two resources that are strongly tied to data set size: CPU and disk I/O. However, having shot a few jobs with the D800 in the past couple of weeks, my impression is that I am not really that much less efficient when compared to the D700. A few things contribute to this, but the major reason is that tasks like generating previews, exporting JPEGs, prepping for Facebook, etc. are all done on "computer time" and not "human time". That is, I'm not sitting at the computer waiting for that stuff to finish.
It still takes me about an hour to sift through 2000 images and pick out the keepers. It still takes me an average of 4 seconds to edit each image to my liking. It still takes me about 8 minutes to layout an album spread in InDesign. That's "human time", and there hasn't been much impact going from 12 megapixels to 36 megapixels.
I'm surprised to see that the change is so linear, even on a fairly robust system. I'm running a two-year-old i7 with 6GB of RAM, and I've noticed a pretty significant lag in LR4 performance between D700 and NEX-7 (24mp) files. I'm planning to add memory this month, though given this data, I'm not sure how much of an improvement I'll see for basic file handling processes.
Smiert Spionam wrote:
I'm planning to add memory this month, though given this data, I'm not sure how much of an improvement I'll see for basic file handling processes.
If your system is starved for RAM (i.e., you're not seeing much available physical memory while working away with the raw files), then adding more will help. But if the bottleneck is elsewhere (e.g., CPU or disk I/O), then obviously increasing your RAM would be a waste of money. My old system with 8 GB of RAM always had 2 or 3 GB free, which is good for caching recently-accessed files on disk. When I wanted to speed things up, I replaced a hard drive with an SSD and that made a big difference. Adding more RAM would have been expensive (I would have to replace all the existing sticks) and not resulted in any improvement.
I have been suffering serious envy since the release of the D800 knowing that my D7k is it for me, for at least the next 3-4 years and this study gives me another reason (apart from poverty ) of why not. Of course my D7k continues to amaze me with what it produces, especially when I get all the basic parameters correct.
There would be many serious pros (Some who dont like being wet and cold) who's reputations and business's depend on providing product that is always 110% of what the client expects/requires and then there are those like myself who are delighted when they oh so occasionally reach what they perceive to be photographic nirvana.
This study is also a bit of a wake up are also those who have been saying on various forums that you can never have enough Mp because storage is cheap, because there comes a point for all of us to be too time challenged to justify the hours spent before the computer
thank you Dave for what is a most interesting study.
Doesn't the D800 have a USB3 connector? I would make the D800 upload much faster to the computer and actually make them the average overall fastest files to process...
Also, you should have converted the files to DNG for a more "complete" workflow.
mantat wrote:
Doesn't the D800 have a USB3 connector? I would make the D800 upload much faster to the computer and actually make them the average overall fastest files to process...
...yes, but only if you're uploading to the computer directly from the camera.
Dave is using a USB3 card reader for his testing of the files from all 4 cameras to equalize the data transfer process for the comparisons - this lets the actual file size of each of the 4 cameras determine the transfer rate per image, and should be more representative of most people's workflows.
I'd also be interested to see if the DNG conversion process has similar timing ratios as the other tasks that Dave has benchmarked, as this is a standard part of my workflow too.
Gregg B. wrote:
<quote>
Export @ full resolution from LR4 to JPEG
D700 : 235 seconds
5D2 : 319 seconds
D800 : 622 seconds
</quote>
What do you mean D800 took 622 seconds?
That's more than 10 minute. How many images were you exporting?
From the OP statements in the original post: "I chose to make timings using 100 random images from each camera so that I could get a reasonable average that could be used to extrapolate other quantities (e.g. 1,000 or more images from a wedding)."
For a low volume shooter like my self I don’t feel this is a problem.
How ever with high volume professionals this certainly becomes a real issue. I guess I now can see why some people don’t want the D800.
I am not a computer expert by any means, but I would wonder if with more computer power you could get the 36 megapixel files to run as fast as a 12 megapixel file on the lower grade computer.
Dennis
Dennis H. wrote:
I am not a computer expert by any means, but I would wonder if with more computer power you could get the 36 megapixel files to run as fast as a 12 megapixel file on the lower grade computer.
Of course. I'm sure I could find a slower computer that would take just as long with a D700 file. In fact I've got a few in the garage
The current export from LR4 has been going for 42mins and it's at about 3/4 done. Oh, and NONE of these are from the D800 either! All D700!
) of why not. Of course my D7k continues to amaze me with what it produces, especially when I get all the basic parameters correct.
