In the Lightroom Classic 11.4 update, one new performance enhancement got my attention: GPU acceleration for exporting. A quick test shows that GPU acceleration can result in faster multi-file exports that use computer resources more efficiently on less power, which in turn produces less heat and fan noise, potentially improving battery life.
Exporting before GPU acceleration: It was all about CPU cores
Up through version 11.3.1, Lightroom Classic used only the CPU to process image exporting. To finish a bulk export as quickly as possible Lightroom Classic employed all cores, so the more CPU cores were available the faster a multi-image export could complete. But on large batch exports, maximum CPU usage could result in unwanted side effects such as other applications becoming less responsive, and reaching the CPU temperature limit. High temperature can increase the noise level of a fan-cooled computer, and if the computer can’t cool itself fast enough, it may slow (throttle) its CPU to stay under its safe temperature limit.
About the quick tests I ran
I exported 624 raw, adjusted time-lapse images in Lightroom Classic 11.3.1 and 11.4, running the test twice in each version, on my 14″ M1 Pro MacBook Pro with 8 CPU cores and 14 GPU cores, 32GB unified memory, and about 200GB free out of 1TB of internal storage. The images are stored on an external SATA SSD connected using 10Gbps USB 3. Export was to JPEG format, quality 85, with no resizing or other processing applied in the Export dialog box.
After GPU acceleration: Faster, more balanced exporting
The most obvious change is that Lightroom Classic 11.4 exported this job about one-third faster:
|Lightroom Classic 11.3.1 (CPU)||10 minutes 7 seconds|
|Lightroom Classic 11.4 (CPU + GPU)||6 minutes 10 seconds|
Lightroom Classic 11.4 clearly shifts much of the export processing from the CPU to the GPU. GPU usage during export went from minimal (probably system use) in Lightroom Classic 11.3.1, to heavy use in Lightroom Classic 11.4.
The CPU cores became less busy in version 11.4, with headroom visibly available at the top of each CPU core’s usage graph.
The CPU usage, GPU usage, temperatures, and fan speeds were measured four minutes into the export job. Because the GPU can reduce the CPU workload when exporting in version 11.4, the measured CPU temperature dropped by 10 degrees Celsius, from 95C in version 11.3.1 to 85C in version 11.4. (Peak temperatures were higher.)
With heat reduced, cooling fan speed also fell by about 30 percent, from about 3800 RPM in version 11.3.1 to about 2650 RPM in version 11.4, reducing noise.
The graphs are from Apple Activity Monitor, which is included with every Mac. Temperatures and fan speeds were measured using TGPro.
What we see looks familiar to me. It’s consistent with what happens after GPU acceleration is added to video editing applications. GPU usage goes up by taking much of the job away from the CPU and doing it more efficiently. Because the CPU has less to do, it no longer maxes out, so its temperature drops, letting the computer run cooler and more quietly. It’s a virtuous cycle.
The performance changes you see will probably be different based on the number of CPU and GPU cores in your computer, its operating system, and its amount of system and graphics memory (unified memory on Apple Silicon Macs).
If you regularly export large numbers of files for projects such as time lapses or events, GPU acceleration for export is a big win all around. It saves a lot of time while reducing overall load, resulting in less heat and fan noise, and longer battery life.
Export GPU acceleration depends on available graphics memory
For GPU acceleration to be most effective, there has to be enough graphics memory (also called video RAM or VRAM) for the GPU to use. The Lightroom Classic system requirements call for a minimum of 2GB graphics memory, but for full GPU-accelerated export they recommend at least 8GB of discrete graphics memory, or 16GB total system memory for integrated graphics or unified memory. For an Apple Silicon Mac with unified memory, the GPU can use memory not needed for the system and applications. For example, if your Apple Silicon Mac has 32GB unified memory, and macOS and applications are using 20GB, then a very generous 12GB might be potentially available for the GPU to use. However, this works against Apple Silicon Macs with 8GB unified memory, because macOS and applications might not leave enough unused memory for effective GPU acceleration, depending on the current workload.
If your system has enough graphics memory to support GPU acceleration for export, the Use GPU for Export option becomes available in the Performance pane of the Preferences window.
Keep in mind that most of Lightroom Classic is still not GPU accelerated. When I built 1:1 previews in version 11.4 the GPU was not used, even though the process is similar to exporting. In Lightroom Classic 11.4, GPU acceleration applies only to the Develop module, image display operations such as zooming, and now exporting. Adobe may have been motivated to prioritize GPU acceleration for exporting after Capture One added it earlier. Not every operation can be GPU-accelerated, but hopefully we can look forward to more performance enhancements as Adobe identifies and optimizes more areas for GPU acceleration.