Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1249
Article Thumbnail

Lightroom Classic CC 2018: AMD Threadripper 2990WX & 2950X Performance

Written on October 4, 2018 by Matt Bach
Share:

Introduction

With the launch of Lightroom Classic CC, the developers at Adobe made a bit of a statement by dramatically improving performance for almost every task. With the 7.2 update in particular, we saw a big improvement in how well Lightroom was able to take advantage of higher core count CPUs. However, the new Threadripper 2950X 16 core and in particular the 2990WX 32 core CPUs have extremely high core counts, so it is time to see if Lightroom Classic is able to take advantage of this many cores or if we have reached the point of diminishing returns.

If you would like to skip over our test setup and benchmark result/analysis sections, feel free to jump right to the Conclusion section.

Test Setup & Methodology

Listed below are the test platforms we will be using in our testing:

To thoroughly benchmark each processor, we will be using two sets of images: one set of 22MP.CR2 RAW images taken on a Canon EOS 5D Mark III and a set of 45MB .NEF RAW images taken on a Nikon D850.


While our Lightroom testing is still evolving, we are currently able to accurately benchmark the following tasks with both sets of images:

Bulk Tasks

  • Import
  • Export to JPEG
  • Build Smart Previews
  • Convert to DNG

Module Tasks

  • Library Module Loupe Scroll
  • Develop Module Loupe Scroll
  • Library to Develop switch

Photomerge

  • Panorama Merge
  • HDR Merge

We are currently working on putting up an alpha version of our benchmark for public download (similar to our Photoshop Benchmark) which will have a much more in-depth description of each of these tests. Keep an eye out in the coming weeks (or months)!

Benchmark Results

Benchmark Analysis

The Scores shown in the chart above are relative to the best possible performance for each task with a Core i7 8700K CPU and a NVIDIA GTX 1080 Ti 8GB GPU. In essence, a score of "90" would mean that it gave 90% the performance of the reference system while a score of "110" would mean it was 10% faster.

In the photomerge and module tasks, there was actually surprisingly little difference between all the Intel CPUs we tested. The Core i7 8700K was at the top due to its high per-clock performance, but the other Intel CPUs were only a few percent slower for the module tasks and about 10% slower for photomerge. The AMD Threadripper CPUs, however, fared pretty well for photomerge but were ~8% slower than our reference system in the module tasks.

What was really interesting were the bulk tasks like exporting and generating smart previews. Most of the Threadripper CPUs were in line with a similarly priced Intel CPU except for the 2990WX. Unfortunately, it looks like Lightroom simply can't take advantage of the 32 cores in that CPU which means that the relatively low clock speed starts to hinder performance.

Is Threadripper 2 good for Lightroom Classic?

AMD's Threadripper 2 CPUs are a bit of a mixed bag for Lightroom Classic. The 2950X is slightly faster than Intel for photomerge, but falls a bit behind for things like scrolling through images in the Library and Develop modules. The 2990WX, on the other hand, sees pretty significant performance issues across the board.

AMD Threadripper 2990WX & 2950X Lightroom Classic CC 2018 Benchmark

The chart above is a summary of all our results, so keep in mind that if you care more about a single type of task you may want to view the chart in the previous section that is divided by task type. This chart gives a great overview of each CPU performs in Lightroom Classic, but there are three primary comparisons we should be looking at based on the rough price of each CPU:

AMD Threadripper 2990WX vs Intel Core i9 7980XE for Lightroom Classic

The Threadripper 2990WX actually performs very poorly in Lightroom Classic, so this is an easy win for the Core i9 7980XE. Across the board, 7980XE was anywhere from slightly faster to 20% faster depending on the task.

AMD Threadripper 2950X vs Intel Core i9 7900X for Lightroom Classic

Overall, the difference between the Core i9 7900X and Threadripper 2950X is small enough that most users won't notice the difference. The i9 7900X has a slight lead when scrolling through images or switching between the Library and Develop module, but the 2950X is slightly better for photomerge.

AMD Threadripper 1920X vs Intel Core i7 7820X for Lightroom Classic

Once again, these CPUs are very neck-in-neck. The i9 7820X is better when scrolling through images or switching between the Library and Develop module, but the 1920X will be slightly better for bulk tasks like exporting or generating previews.

If we had to declare a winner between the Intel X-series and AMD Threadripper CPUs for Lightroom Classic, we would really call it a tie except at the highest end since the 2990WX has clear performance issues in Lightroom. This means that which CPU you should purchase is likely going to depend on the other applications you use (like Photoshop) and which CPU ends up being better for those software packages. To help answer that question, be sure to check out some of our other recent Threadripper articles to help give you an idea of how each CPU might perform in a variety of applications.

Tags: Lightroom CLassic, Skylake-X, Threadripper, 2990WX, 1920X, 1950X, 2950X, Core i7, Core i9, 8700K, 7820X, 7900X, 7920X, 7940X, 7960X, 7980XE
Behive

Nice comparison, thanks. Look forward to seeing how the upcoming i9 9900K will compare.

Posted on 2018-10-05 02:54:44
Ondrej

I wait for i7 9700K or i9 9900K. Both have big potential for Lightroom work (i make upgrade from i5 4670 non K, that is too slow for 42Mpx RAWs)

Posted on 2018-10-05 19:45:07
HAS-Tecno

Hi, I think that in the conclusion this section is wrong:

AMD Threadripper 1920X vs Intel Core i7 7820X for Lightroom Classic
Once again, these CPUs are very neck-in-neck. The i9 7900X is better when scrolling through images or switching between the Library and Develop module, but the 2950X will be slightly better for bulk tasks like exporting or generating previews.

Posted on 2018-10-05 15:18:37

Looks like I just got the model names switched around in my head while writing. Got that fixed, thanks!

Posted on 2018-10-05 15:43:30
Hwgeek

Good news from AMD:
Dynamic Local Mode for the AMD Ryzen™ Threadripper WX Series Processors
https://community.amd.com/c...

Posted on 2018-10-05 20:20:53

AMD tried to push this with the original Threadripper CPUs https://community.amd.com/c... , but it is pretty cool that they made it automatic and not require a reboot. It's been a while, but back then we did test it in Premiere Pro https://www.pugetsystems.co... and After Effects https://www.pugetsystems.co... and were not very impressed. Even in After Effects which is very lightly threaded we saw a bit lower performance in "Game Mode". It will be interesting to see how much of a difference it makes the with 2990WX, however, since it seems to have pretty big issues with lightly threaded tasks.

Posted on 2018-10-05 20:30:07
Jakub Badełek

Another great article! you don't have an easy task with keeping up with changes in the industry. Intel just officialy announced Coffe Lake Refresh with 8/16 top processor... can't wait to see how it performs in LR ;)

Posted on 2018-10-08 21:11:32

Yea, there is a lot going on right now. Threadripper 2, GeForce RTX, Quadro RTX, and now the Intel 9th Gen and X-series CPUs are announced. Posted this on twitter early expressing how it feels right now: https://twitter.com/i/statu...

Posted on 2018-10-08 21:15:46
Martijn Westerink

In general, it seems that Lightroom has a reduced benefit from using more then 10 cores. Past that point, performance takes a hit because of the lower clock-speed. So ideally, you'd pick an unlocked 10-core (20T) CPU and push the frequency.

Posted on 2018-10-09 11:06:33
computerhersteller

Thanks for the tests! Just what I wanted to test myself.

If you need alfa/beta testers for the Lightroom benchmark I would be happy to test! I'm currently testing video cards to find out the benefit of gpu acceleration.

My tests are based on file time stamps (previews, imported and exported photo's). I test on importing, creating small and 1:1 previews, applying a bunch of development settings followed by recreating 1:1 previews, export without and export with watermark. But I can't measure loupe scroll, zooming,... If there's some caching involved, my measurements are completely inaccurate.
According to my tests importing with GPU accellerator turned off would be 10% faster compared to importing with GPU turned on... (530 photo's, 24mp, raw, GTX1050ti)

Posted on 2018-10-09 12:59:45

I haven't done much GPU testing in Lightroom yet, but I have heard people report higher performance with the GPU turned off. The hard part is that the parts of Lightroom that I believe really benefit from GPU acceleration is things like applying corrections, brushes, etc. in the Develop module. Unfortunately, those are all things we haven't figured out a reliable way to benchmark quite yet, so if we were to do specific GPU on/off testing I'm afraid it won't tell the whole story.

Posted on 2018-10-09 16:48:09
AleksaMK

Lightroom is really poorly whiten software, Adobe programmers stayed in distant past regarding advance of CPU's and GPU's.
Try Capture One Pro and see how a properly, let me say modern, written software, that doesn't favor one vendor, deals with your hardware, you'll be very much surprised. Not to mention that C1 raw conversion is light years ahead from Adobe raw conversion.

Posted on 2018-10-12 10:30:07
Julien B. O.

Very interesting! One thing I would love to see: The AMD machines with faster RAM. Benchmarks all around the internet have shown that higher memory speed have a greater impact on the Zen architecture than on Intel. DDR4-3200 with CAS 15 at a maximum, or better yet CAS 14, would show what can really be achieved with AMD. The difference is significant in most other types of benchmarks, so it should also be interesting in Lightroom.

Posted on 2018-11-09 17:49:35

Faster RAM will make a difference with AMD CPUs, but my guess is it would only be by 5-10%. Something to keep in mind is that AMD is pretty bad about telling you exactly what RAM speed their CPUs support since it depends on how many sticks you use and whether they are dual or single rank. They just list "Max RAM speed" on their specs which is likely what you should use with a single stick of RAM (which no one ever does). So using something like DDR4-3200 is technically overclocking and definitely does increase the risk of system instability.

I mention that because we typically don't do overclocking or anything else that increases the chances of system stability issues since our customers overwhelmingly want a system that is rock solid rather than one that is a bit faster, but Lightroom/Photoshop/Premiere/whatever crashes even a little bit more often. Since our articles are first and foremost to make sure we are selling the right hardware to our customers, testing with higher speed RAM than we are comfortable using in our workstations is not really high on our priority list. We might at some point get in some faster RAM just to answer the question of how much faster Ryzen/Threadripper would be if you are willing to take the stability risk, but with all the hardware launches and software updates going on I'm not sure when that would be.

Posted on 2018-11-09 18:29:10
Julien B. O.

Yes I fully understand that. Then perhaps using the processors at the top speed they were designed for would make a fairer comparison? DDR4-2666 was for the first generation of Ryzen chips and the new ones support DDR4-3000 (although this might be limited to 64GB, I'm not quite sure). Or at least keep that in mind when testing the next generation to come!

In any case, thank you for the benchmark.

Posted on 2018-11-09 19:43:57

2nd Gen Ryzen (and Threadripper) is rated for a max of 2933MHz, so using DDR4-2666 like we are is only one step down from the absolute max you are supposed to use with them. The most up to date info we have from AMD is detailed here https://www.pugetsystems.co... . That was before the latest Threadripper came out, but as far as I understand from our contacts at AMD, it is still generally correct for those CPUs as well. So you would need to use single rank RAM and populate only half the RAM channels to use 2933MHz RAM (if you want to stay in spec). Most RAM these days is dual rank, and we are using all the RAM slots so most likely we are slightly outside of spec even with this configuration.

We are pretty comfortable with DDR4-2666, however. Right now, we feel like it is the most stable frequency for DDR4 RAM in general. As long as you get high quality RAM that isn't crazy on the timings, you probably would be OK with using DDR4-2933MHz, but definitely don't try to go for budget RAM at that speed. I don't think it is going to be a huge performance increase, however, so if it was my personal system I would stick with DDR4-2666.

Posted on 2018-11-09 19:56:07
Rex Lajos

Tr 2990wx is suffering from a windows scheduling bug, not low core speed.

Posted on 2019-03-01 00:25:45

This article is a bit old at this point and was before AMD released their "Dynamic Local Mode" software that was supposed to fix the single threaded performance issues. This is the most up to date results https://www.pugetsystems.co... using DLM. Unfortunately, performance isn't much better even with DLM.

There are a few people trying to improve on AMD's implementation as well (https://bitsum.com/portfoli... for anyone who wants to look into it), but while it looks promising, from what I've heard in their updates, it is still a bit inconsistent. I think the biggest issue at this point is that the 2990WX has been out for 6 months now and AMD still hasn't done anything beyond DLM to try to fix the issues with the "WX" Threadripper CPUs. That really suggests to me that it isn't just a simply bug, but something inherent with the architecture itself.

Posted on 2019-03-01 00:38:44
Rex Lajos

Matt, Dynamic Local mode can only do so much when Windows will bounce a task around different threads on different ccx with separate cache. L1 tech has been working on a solution for a while now and most of the aforementioned issues are mainly a Windows issue as the are not present in Linux.

Posted on 2019-03-01 22:25:36

Yep, the Coreprio software I linked to is what Wendell and the rest of L1Tech have been helping develop as a intermediate fix for the issue. The last I heard, AMD is saying that the underlying issue is Microsoft's fault and the hope is that a fix will be in place sometime in the first half of this year. Once there is an official fix, we will of course implement it in our testing.

Posted on 2019-03-01 22:37:22