Lightroom performance tips?

Having just upgraded from a 5D3 to a 5D4 I'm finding the extra resolution is really pushing Lightroom hard, and things are getting pretty laggy when editing in the Develop module.

I'm running a Windows 10 machine on an i7-5930K, which isn't exactly a cutting edge CPU, but it's no slouch (and the Task Manager isn't indicating I'm CPU bound). Plenty of RAM too. OS and LR catalog are running from an SSD.

Given that one of the big lags is switching between images I'm wondering if it's just too much for my old graphics card (Quadro 4000). I did find it helps performance with 5D3 files, but perhaps these are just going beyond some memory limit or bus transfer speed.

I'm wondering if any 5D4/5DS/D800/D810 owners are running LR and might have some tips on improving performance? In particular, what graphics cards (on Windows) are working well?

You're 5930K is perfect and I even over clocked mine to 4.6Ghz. had 16 and 32G of high speed ram. I have an NVME SSD for the OS and caches, then the catalog and previews was on SataIII (6Gb) 550MB/sec or so external SSD.

LR with my 5DSR files and NVidia was a pig with my NVidia 680 TI. So I upgraded that to 980Ti for more perf.

Nothing worked with making LR and 5DsX files..

I moved to capture one and I have damn near instant rendering as I would expect our machines to do.

Ron

sploo should overclock (why buy a K processor if you aren't going to use the K feature?) and edit his/her files on an SSD (even if it means installing another one).

Ron Hu wrote in post #18134985
You're 5930K is perfect and I even over clocked mine to 4.6Ghz. had 16 and 32G of high speed ram. I have an NVME SSD for the OS and caches, then the catalog and previews was on SataIII (6Gb) 550MB/sec or so external SSD.

LR with my 5DSR files and NVidia was a pig with my NVidia 680 TI. So I upgraded that to 980Ti for more perf.

Nothing worked with making LR and 5DsX files..

I moved to capture one and I have damn near instant rendering as I would expect our machines to do.

Ron

Did you see a big improvement from going from the 680 to the 980?

I guess I could look into Capture One - it's just that I've gotten very comfortable with LR over the years.

mclaren777 wrote in post #18135153
sploo should overclock (why buy a K processor if you aren't going to use the K feature?) and edit his/her files on an SSD (even if it means installing another one).

I have done that, but mostly it seemed to improve import times. I probably should enable the overclocking again and try.

As far as I can see, once you have an image "in" and it's being edited, the machine isn't really touching the disk on which the image resides (i.e. I can see the catalog disk being a bottleneck - though I have an SSD, not the "data" disk).

LR does not use the GPU to render previews/images, it is a totally CPU bound operation. It only makes very light and limited use of the GPU for some of the adjustment brushes, but on systems with faster CPU's the time it takes to off load the task to the GPU is often longer than it would take it just run it through the CPU.

So the biggest performance improvement you can make in LR is to turn off GPU acceleration, the second biggest improvement is to not use it. It's based on a database type that is known to be slow. Of course the list of alternatives is very short, and they have their own problems. Capture One might render previews faster than LR, but it has other issues with stability and does not play well in any kind of enterprise based network.


We run 5DS and 5DSR at work, on HP Z series work stations with 3 SSD's. This allows us to separate image storage from the scratch/cache and keep all of that separate from the operating system and programs. It is at bout as optimal a set up as you can get for running LR, and when working on 6D files it is lightening fast and very responsive. Working on the 5DS files however it is slow and clunky.
Fortunately we work in an environment where we do not do any post, everything is worked out in camera. So RAW files are only assigned a WB and key words in LR before export to archival storage.
I would not want to be editing, selecting and processing 5DS files on a regular basis.

I have found that giving LR time to build 1:1 previews before you work with the images does smooth things out, especially for load times when moving between images.
But any editing, especially working with batches is still slow.

Moppie wrote in post #18135244
LR does not use the GPU to render previews/images, it is a totally CPU bound operation. It only makes very light and limited use of the GPU for some of the adjustment brushes, but on systems with faster CPU's the time it takes to off load the task to the GPU is often longer than it would take it just run it through the CPU.

So the biggest performance improvement you can make in LR is to turn off GPU acceleration, the second biggest improvement is to not use it. It's based on a database type that is known to be slow. Of course the list of alternatives is very short, and they have their own problems. Capture One might render previews faster than LR, but it has other issues with stability and does not play well in any kind of enterprise based network.


We run 5DS and 5DSR at work, on HP Z series work stations with 3 SSD's. This allows us to separate image storage from the scratch/cache and keep all of that separate from the operating system and programs. It is at bout as optimal a set up as you can get for running LR, and when working on 6D files it is lightening fast and very responsive. Working on the 5DS files however it is slow and clunky.
Fortunately we work in an environment where we do not do any post, everything is worked out in camera. So RAW files are only assigned a WB and key words in LR before export to archival storage.
I would not want to be editing, selecting and processing 5DS files on a regular basis.

I have found that giving LR time to build 1:1 previews before you work with the images does smooth things out, especially for load times when moving between images.
But any editing, especially working with batches is still slow.

Yea - as far as I understand it's only used in the Develop module. What surprised me was that, with a processor rated something like 350% of the speed of the old one (I know different loads would get different results) I was seeing something like that speed improvement with video rendering, but the improvement in importing photos in LR was relatively small. I can't remember the figures, but it was something way way less (like a 30% improvement).

I definitely found the editing operations I do in the Develop module were quicker with the GPU on than off - though I haven't compared it with 5D4 files.

I do build 1:1 previews for all images in import BTW. Definitely worth the upfront time.

Double check your preferences and try toggling GPU acceleration. They seem to have a bug in how they're implementing the feature, which makes for a very hit and miss performance gain. Processing the .cr2 files from my 7D seems to like GPU acceleration on in my setup, but processing .tiff files from my scanner makes for a not so happy system when it is on. Kind of weird, and seems to be something they've been exploring and trying to improve going off Adobe blog/forum comments here and there.

I noticed an improvement in computer speed when I moved originals to an external drive, disconnected it, and edited Smart Previews rather than the originals. Originals are updated when you reconnect the external drive. Using a 2008 MacBook Pro.

sploo wrote in post #18135220
Did you see a big improvement from going from the 680 to the 980?

...

Unlikely to have any impact at all.

Larry Johnson wrote in post #18135448
I noticed an improvement in computer speed when I moved originals to an external drive, disconnected it, and edited Smart Previews rather than the originals. Originals are updated when you reconnect the external drive. Using a 2008 MacBook Pro.

I assume that's a bit like editing proxies. My worry is that I like to do some work at 1:1, so by definition, a lower res proxy wouldn't give me that.

Luckless wrote in post #18135351
Double check your preferences and try toggling GPU acceleration. ....

I am sure you mean to say "toggling to off" ?

This often helps speed things up.


The issue with LR is almost always tied to it's insistence on allowing itself to be bottle-necked by the slowest part of any computer, ie: the file storage. The only repeatable improvements to speed we have seen on this forum have to do with SSD, RAM, moving libraries for faster access, etc.

CyberDyneSystems wrote in post #18135452
Unlikely to have any impact at all.

CyberDyneSystems wrote in post #18135457
I am sure you mean to say "toggling to off" ?

This often helps speed things up.

The issue with LR is almost always tied to it's insistence on allowing itself to be bottle-necked by the slowest part of any computer, ie: the file storage. The only repeatable improvements to speed we have seen on this forum have to do with SSD, RAM, moving libraries for faster access, etc.

I do see (even with the 5930K) a significant improvement in responsiveness and speed when enabling the use of an old Quadro 4000 card (for cropping, zooming, gradient filters etc). There's definitely some lag when switching from one image to another though. That was with an earlier version of LR 6 and 5D3 files - I haven't yet had chance to play with 6.7 and 5DIV files (i.e. experiment with turning the GPU off).

I'd assume that a more modern card would be better - and in particular one thing noted on an Adobe forum by one of the developers was that newer cards will be better in terms of data being moved to the card. Problem is, getting a new card would be a relatively expensive experiment (without real evidence to confirm it'd make a difference).

CyberDyneSystems wrote in post #18135457
I am sure you mean to say "toggling to off" ?

Actually I meant "Flip it to whatever setting it currently isn't at". If it is off, try turning it on, if it is already on, then turn it off. As I said in my post it can actually even depend on the specific kind of file you're working with, so it is worth trying either state any time Lightroom appears to be lagging oddly.

It is a truly weird system they have with it, and the only way to be sure which works better on a given system is to try the setting in both modes and see whether or not having it on actually helps or hinders the processing.

sploo wrote in post #18135469
I do see (even with the 5930K) a significant improvement in responsiveness and speed when enabling the use of an old Quadro 4000 ..

Ah, a Quadro is a different beast all together.

CyberDyneSystems wrote in post #18135504
Ah, a Quadro is a different beast all together.

It is, but it's rated down around ~1900 by the CPU Benchmark site. A modern GTX 1060 is over 9000.

Granted the Quadro cards have features lacking in the gaming cards, but the gist I get is that the GTX range are good for LR, and I'd be surprised if they haven't surpassed a ~6 year old Quadro card.