Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1039
Article Thumbnail

Photoshop CC 2017.1.1 CPU Performance: Core i9 7940X, 7960X, 7980XE

Written on September 25, 2017 by Matt Bach
Share:

Introduction

Photoshop has long been known to be an application that doesn't scale well with high core count CPUs, so it might seem odd that we are benchmarking the new Intel Core i9 7940X 14 core, Core i9 7960X 16 core, and Core i9 7980XE 18 core CPUs in Photoshop. The reason behind this testing is not because we think they will be particularly good, but rather because Photoshop is very commonly used in a secondary role. While graphics designers who use Photoshop as their primary tool may not particularly care about these results, video editors, 3D modelers, and many others might be curious to know what kind of performance they will be giving up by using one of these high core count CPUs.

If you are interested in how the new CPUs perform in some of these other applications, you can view our full list of Skylake-X articles here.

To see how these new CPUs perform, we tested how long it took to complete a wide range of tasks in Photoshop including:

  1. Launching Photoshop
  2. Opening/Saving large .PSD files
  3. Convert to CMYK/RGB
  4. Applying Blurs
  5. Noise Reduction
  6. Camera RAW Filter
  7. Resizing images
  8. Creating Panorama image
  9. Creating HDR image
  10. More - see full list in the Benchmark Results section

If you would like to skip over our test setup and individual benchmarks, feel free to jump right to the conclusion section.

Test Setup

Our test platforms for the Skylake-X and Threadripper CPUs are listed below, but we did want to point out that the RAM configuration changed a little bit depending on the CPU. For the majority of the CPUs we used DDR4-2666 RAM but since the Core i7 7800X only natively supports DDR4-2400 we tested with that RAM instead.

As we hinted at in the introduction, Photoshop actually performs best with a lower core count, high frequency CPU so these new 14, 16, and 18 core CPUs (or any of the above CPUs for that matter) are unlikely to be the absolute best in terms of raw performance. To see how they compare against what is currently the fastest overall CPU for Photoshop, we also decided to include the Intel Core i7 7700K in our testing.

Intel Core i7 7700K (Z270) Test Platform
Motherboard: Asus PRIME Z270-A
CPU: Intel Core i7 7700K 4.2GHz
(4.5GHz Turbo) 4 Core

~$339
RAM: 4x DDR4-2400 16GB
(64GB total)
GPU: NVIDIA GeForce GTX 1080 Ti 8GB
Hard Drive: Samsung 960 Pro 1TB M.2 PCI-E x4 NVMe SSD
OS: Windows 10 Pro 64-bit
Software: Photoshop CC 2017.1.1

The images we used in our testing (and their source) are:

Photoshop Actions

360MP (21500x16718)
Scaled up from the Hardware Heaven Photoshop Benchmark V3 (no longer available)

Resized to 38MP (7000x5443) for Smart Blur

Merge to HDR

5x 18MP TIFF (5184x3456)

Camera: Canon EOS REBEL T3i

5x 80MP TIFF (7760x10328)

Camera: Phase One IQ180
Courtesy of Mark McGilvray Photography

Settings: Default

Photomerge

6x 18MP TIFF (5184x3456)
Merged to 98MP (26144x3759)

Camera: Canon EOS REBEL T3i

6x 80MP TIFF (7760x10328)
Merged to 177MP (18552x9552)

Camera: Phase One IQ180
Courtesy of Mark McGilvray Photography

Settings: Auto Layout. Blend, vignette removal, geometric distortion correction, and content aware fill enabled.

Benchmark Results

Due to the sheer number of actions we tested and the fact that the performance differences were remarkably consistent, we are going to forego our typical action-by-action breakdown. Feel free to analyze the individual results below in more detail if you wish, but due to the consistency of the results, we encourage you to skip straight to our conclusion for our benchmark analysis.

Scroll right for more results -->

Time to complete Photoshop action (Seconds) Intel Core i7 7700K 4 Core
4.2GHz (4.5GHz Turbo)
$339
Intel Core i7 7800X 6 Core
3.5GHz (4.0GHz Turbo)
$389
Intel Core i7 7820X 8 Core
3.6GHz (4.3/4.5GHz Turbo)
$599
AMD TR 1920X 12 Core
3.5GHz (4.0GHz Turbo)
$800
Intel Core i7 7900X 10 Core
3.3GHz (4.3/4.5GHz Turbo)
$999
AMD TR 1950X 16 Core
3.4GHz (4.0GHz Turbo)
$1000
Intel Core i9 7920X 12 Core
2.9GHz (4.3/4.4GHz Turbo)
$1199
Intel Core i9 7940X 14 Core
3.1GHz (4.3/4.4GHz Turbo)
$1399
Intel Core i9 7960X 16 Core
2.8GHz (4.2/4.4GHz Turbo)
$1699
Intel Core i9 7980XE 18 Core
2.6GHz (4.2/4.4GHz Turbo)
$1999
General Photoshop Actions                
PS Start Time 1.7 2.2 2.1 2.5 1.9 2.8 2.7 2.6 2.2 2.8
Convert to CMYK 2.7 2.2 1.9 2.8 2.2 2.9 2.4 2.5 2.6 2.8
Convert to RGB 4 3 2.2 2.6 2.4 2.7 2.6 3.7 4.4 4.3
Rotate 38deg 4.5 5.1 3.6 3.9 4.5 3.7 4.3 4.2 4.2 4.1
Smart Sharpen 4 4.6 4.2 4.6 4.8 4.7 4.9 4.9 4.9 5
Field Blur 26.8 30.3 25.1 31.9 29.5 32.4 28.1 29.9 28 29
Iris Blur 30 32.9 29.2 32.9 30.4 32.7 29.2 30.6 28.8 29.1
Tilt-Shift 29.8 32.6 29.3 33.2 30.6 33.1 29.3 31 28.7 29.4
Lighting Effect 14.2 18.3 20.7 15.7 20.4 16.8 16.3 16 15.9 16.3
Motion Blur 6.3 7.1 7.0 5.7 6.5 5.1 6.2 5.9 5.8 5.6
Water Color 44.5 50.5 45.4 63.9 44.9 64.6 46.7 46.9 47.2 51.8
Pallette Knife 73.1 89.2 79.7 82.8 78.6 84.1 82.6 82.6 82.9 90.9
Stained Glass 137.2 161.4 164.8 100.8 150.2 108.3 147.6 157.6 149.6 163.6
Liquify 17.8 25.8 21.3 22.2 23.3 22.2 21.5 18.8 18.9 19.7
Reduce Noise 52.5 59.2 51.9 48.9 51.2 48.8 54.2 53.6 53.9 58
Camera Raw Filter 13 11.6 10.1 11.5 9.4 11.7 9.2 9.2 9.4 10.3
Generate Normal Map 10.3 13.1 11.7 13.2 11.6 13.1 12 12 12 12.9
Lens Correction 36.8 50.9 40.5 43.3 39.6 43.1 41.4 41.5 41.8 44.6
Adaptive Wide Angle 138.4 189.9 198.8 203.9 258.6 224.6 241.2 243.8 247.1 256.5
Resize to 109MB 7.1 9 8.1 8.7 8.1 8.7 8.1 8.3 7.8 8.2
Smart Blur 20.8 26.4 22.7 22.8 22.7 22.9 22.8 23.2 22.4 24.4
Save 1.17GB PSD 16 21.1 17.1 20.7 19.7 19.9 19.5 22 19.7 20.4
Open 1.17GB PSD 5.5 6.3 5.4 5.9 5.5 5.9 5.6 5.8 5.6 5.9
Photomerge                  
Photomerge - 6x 18MP Images 64 64.5 56.1 58.7 55.3 57.8 56.6 55.6 55.4 59.7
Photomerge - 6x 80MP Images 248.7 276.5 248.6 253.9 243 258.5 251.1 250 247.1 270.3
HDR Creation                  
HDR Analysis - 5x 18MP Images 14.2 20.4 16.4 38.4 18.2 20.5 25.4 15.5 20.3 23.7
HDR Creation - 5x 18MP Images 7.7 8.8 7.7 9.2 7.7 9.4 8.1 8 8.1 8.8
HDR Analysis - 5x 80MP Images 33.8 32.8 32.0 37.7 32.9 39.5 32.1 33.7 34 36
HDR Creation - 5x 50MP Images 30.3 34.2 30.3 35.9 29.8 36.9 31.4 31.7 31.8 34.5

Conclusion

Photoshop Skylake-X Core i9 7940X 7960X 7980XE Overall Benchmark Results

Overall, we were actually a bit surprised at how well these new CPUs did in Photoshop. We wouldn't recommend them for people who are primarily concerned about Photoshop performance since the Intel Core i7 7700K (or even the Intel Core i7 7820X) is going to be overall faster and much less expensive, but it is good to see that there is not a huge performance loss. HDR creation definitely got a bit slower as we went up in core count but for general actions you are only looking at about 9-14% lower performance compared to the fastest CPU available for that task.

Are the Core i9 7840X, i9 7960X, or i9 7980XE a good choice for Photoshop? Probably not if Photoshop is where you really need the performance. However, if you just need to make some materials in Photoshop to use before rendering a scene in Keyshot or Vray, these CPUs should work out just fine. They are not the best, but for light or even medium Photoshop loads you would probably be hard pressed to tell the difference between them and a CPU like the Core i7 7700K without the use of a stopwatch.

Tags: Photoshop, Skylake-X, Threadripper, 7980XE, 7960X, 7940X
Eloi

It would be nice to see some CPU/GPU comparative for Capture One.

Posted on 2017-10-03 21:36:58

We've been hearing more and more people mentioning Capture One, so that is definitely something on our list to look at. We have our testing prioritization already pretty set through the end of the year, but hopefully we'll be able to get to it relatively soon.

Posted on 2017-10-03 23:09:02
Umano Teodori

It's a good idea, a lot of ppl now are complaining about macs even the most faithful customers. But in the photography most ppl use laptops (even a lot of retouchers) so a good custom laptop, with full aRGB matte panel, maybe even a 17" for some, can be awesome and that kind of customers are accustomed to very high price tag 3/3.5k, But ofc the support service has to be apple eq or even better

Posted on 2018-01-26 22:48:25
Christian Alexandrov

Hello guys,

thanks for the benchmark comparison, this is very helpful. I am curious about the X399 platform for AMD. I know that multicore CPU like Threadripper is not the opmimal choice for editing photos in Photoshop, but for larger projects I can add more memory than on the current mainsteam Intel motherboards. The interesting part is that on X399 you can combine two or more NVMe devices in RAID 0, which will be powered directly by the CPU lanes. I think this will be very beneficial in terms of fast scratch disk writes. Can you make a RAID 0 benchmark on a X399 platform, using your usual actions?

Posted on 2017-10-18 15:47:11

There may be times that a NVMe RAID 0 scratch drive will be useful, but honestly if you have the money for that you should just get more system RAM so that you don't even really need a scratch drive. Even in RAID 0, a pair of NVMe drives is going to be much slower than system RAM so simply having enough RAM is almost always going to be better.

We haven't tested this is Photoshop specifically, but in almost everything we have tested, once you get up to the speeds of a single M.2 drive you actually become CPU bottlenecked before storage speed is an issue. It may be different in Photoshop (you never know without actual benchmark data) but I suspect that RAIDing a pair of NVMe drives for scratch won't actually be any faster than just using a single NVMe drive.

Posted on 2017-10-19 16:56:48
Christian Alexandrov

Well, AMD enabled NVMe Raid on the X399 platform and thus the bottleneck of the DMI 3.0 bus is eliminated (this is used in Intel platforms where the disk controllers are routed through the chipset and the link is much slower - x4) The X399 has the benefit that all M.2 connections have direct link to the CPU, which cannot be a bottleneck for the drives throughput.
The first benchmarks after the BIOS update and AMD driver installation show incredible speed of 18-20 GB/s (achieved with 5 or 6 NVMe drives in Raid 0) and this is quite a leap in performance. But how Photoshop will handle this speed and if there will be any actual benefit when reading and writing small chunks of data in the scratch file is quite another matter...
Such a benchmark will be very interesting to see.
Thanks for your answer.

Posted on 2017-10-19 22:22:27
Rebecca Wayne

Are these benchmarks, and in fact all your benchmarks, just based on each CPU's base operating frequency? Just from looking at your numbers, it would appear that those CPUs with the highest base frequencies have the highest performance, which is expected since clock frequency still reigns over memory speed, GPU speed, and sometimes can also negate the difference between 4, 8, and 10 cores if the higher-core CPUs naturally run at a lower base frequencies. But especially for Intel CPUs that often have high overclocking headroom, you really need to normalize all your benchmarks to the same frequency to make a valid comparison between CPUs. For example, if each CPU was clocked to 4.0 GHz, which all Intel and AMD CPUs can attain, that would not skew your results to favor the lower-core CPUs. So I would like to see these benchmarks normalized so the 4, 6, 8, 10, 16, and 18 core CPUs are all run at the same frequency during your testing.

Posted on 2017-11-02 20:49:16

Our testing is always done at the stock speeds, but that isn't actually the CPU's base frequency. Pretty much all Intel and AMD CPUs (minus the Core i3) these days have a form of Turbo Boost that increasing the frequency depending on the number of cores that are being used. Honestly, the "base frequency" is pretty much meaningless these days, but we still list it in the CPU name since Intel and AMD continue to use it in the CPU's official name.

Most of the time, the CPU will be running somewhere between the base and Max Turbo frequency depending on the number of cores being used. If one or sometimes two cores are used, you get the max Turbo frequency but the more cores you use, the lower the frequency drops.

As for normalizing the frequency between the CPUs, that is actually similar to what we used to do in our multi-threading tests like this https://www.pugetsystems.co... . A little different since we used a single CPU and limited the number of cores rather than limiting the frequency, but essentially the same idea. Unfortunately, we found that the Multi-Core testing didn't actually mean much in the real world. It is interesting, but it doesn't work to accurately estimate the performance of different CPUs due to all the Turbo/speedstep stuff. Because of this, we transitioned to simply testing the majority of Intel/AMD CPUs that make sense for that application. That way we have real results that we can use to directly compare performance across different CPU models.

Posted on 2017-11-02 21:05:59