MP 7,1 Puget Photoshop Speed Comparison 7,1 to 5,1

[choreo]

I just ran the Puget Photoshop Speed Test on my new 7,1 Mac Pro (see report for specs). I also ran the same test on my 2012 5,1 Mac Pro in High Sierra with a 580X Graphics card several weeks ago in preparation to compare to the new 2019 MacPro. Same test, but they upgraded their interface in the meantime. Latest versions of Photoshop CC at the time:

2012 6-core 5,1:
Overall Score - 494.6
GPU Score - 53.5
General Score - 54.1
Filter Score - 43


2019 12-core 7,1:
Overall Score - 3,115
GPU Score - 135.3
General Score - 218
Filter score - 404.9

 

[OkiRun]

I just ran the Puget Photoshop Speed Test on my new 7,1 Mac Pro (see report for specs). I also ran the same test on my 2012 5,1 Mac Pro in High Sierra with a 580X Graphics card several weeks ago in preparation to compare to the new 2019 MacPro. Same test, but they upgraded their interface in the meantime. Latest versions of Photoshop CC at the time:

2012 6-core 5,1:
Overall Score - 494.6
GPU Score - 53.5
General Score - 54.1
Filter Score - 43


2019 12-core 7,1:
Overall Score - 3,115
GPU Score - 135.3
General Score - 218
Filter score - 404.9

Can you provide a quick synopsis of how the test works and how it measures results? Its a speed test of what function?
Thanks!

 

[choreo]

Can you provide a quick synopsis of how the test works and how it measures results? Its a speed test of what function?
Thanks!

You can download it from Puget with instructions. It opens a series of images and performs various functions and filters on them in about 20-30 minutes.

 

[sirio76]

Looks like their benchmark is running 3x faster on MP compared to all the other Windows systems in the list(select PS benchmark):

PugetBench List

Build your own computer, workstation, mini pc, or server at Puget Custom Computers.

www.pugetsystems.com

 

[OkiRun]

Did you have to follow this step?

====

1. Download a ZXP (Adobe plugin) installer and use it to install the plugin. We recommend using aescripts ZXP installer to install the .zxp file

 

[choreo]

Did you have to follow this step?

====

1. Download a ZXP (Adobe plugin) installer and use it to install the plugin. We recommend using aescripts ZXP installer to install the .zxp file

Yes - that is what I used.

 

[DFP1989]

So I'm really confused by your results, we have almost the same system (I'm 16-core v 12-core), and I'm getting a score of less than 1/3 what you did.

 

[choreo]

So I'm really confused by your results, we have almost the same system (I'm 16-core v 12-core), and I'm getting a score of less than 1/3 what you did.

What drive is your Photoshop application on? Any scratch disk set? Amount of Ram allocate to Photoshop? GPU Rendering on?

 

[AidenShaw]

Looks like their benchmark is running 3x faster on MP compared to all the other Windows systems in the list(select PS benchmark):

PugetBench List

Build your own computer, workstation, mini pc, or server at Puget Custom Computers.

www.pugetsystems.com

One should be suspicious of accepting this as a blanket statement.

If the base system is the same, 3x is absurd. If the test is using accelerators (CUDA, AfterBurner, GPUs,...) those need to be mentioned.

Apple's Xeon systems aren't three times faster than other Xeon systems. For some particular combinations of applications and hardware - it could be true. But not in general.

 

[DFP1989]

I totally missed sirio76's reply before. My results would seem to be along the correct lines then, a 3x difference doesn't make sense.

@choreo App and scratch set to internal SSD, 70GB allocated to PS, all GPU settings on default (aka. on).

I'm about to go out so will set it running again with the scratch on a blank NVMe SSD (rated for over 3GB/s) and see if there is any difference. I would be surprised if there was, and even more so if it's 3x.

 

[choreo]

I totally missed sirio76's reply before. My results would seem to be along the correct lines then, a 3x difference doesn't make sense.

@choreo App and scratch set to internal SSD, 70GB allocated to PS, all GPU settings on default (aka. on).

I'm about to go out so will set it running again with the scratch on a blank NVMe SSD (rated for over 3GB/s) and see if there is any difference. I would be surprised if there was, and even more so if it's 3x.

My settings look just like yours.

I just re-ran the test to see what I got...

 

[DFP1989]

Only difference this time was using the separate NVMe SSD as scratch rather than internal.

I wonder if the 5700XT in my system is having an affect? Still seems odd that my system would sit right along with other powerful systems, and be over 60% slower than yours.

 

[choreo]

Only difference this time was using the separate NVMe SSD as scratch rather than internal View attachment 939907

I don't think the test is using the scratch disk (mine is set to the internal SSD). Your GPU Score is not too far off from mine, but the general score and filter score is quite a bit different?

 

[sirio76]

One should be suspicious of accepting this as a blanket statement.

If the base system is the same, 3x is absurd. If the test is using accelerators (CUDA, AfterBurner, GPUs,...) those need to be mentioned.

Apple's Xeon systems aren't three times faster than other Xeon systems. For some particular combinations of applications and hardware - it could be true. But not in general.

Usually I agree with you and in general when something is too good to be true it isn’t. This though is not a synthetic benchmark, it just measure the time it would take to complete a series of real tasks on PS and if you go to compare the time in seconds(under raw results) you will see that effectively many operation runs much faster. Probably it’s just a matter of optimization of the resources under MacOS.

 

[DFP1989]

Would be interesting to see someone else here run the benchmark. Still don't understand why my system aligns with other fast systems in the table, and yours is a substantial outlier.

FWIW my resource usage is very low during the benchmark.

 

[choreo]

Would be interesting to see someone else here run the benchmark. Still don't understand why my system aligns with other fast systems in the table, and yours is a substantial outlier.

I don't have an answer. If it makes you feel any better, I have had so many crashes today, it offsets any speed gains!

How long does it take for your copy of Photoshop to launch (pallets to appear after double-clicking the app)? Mine is taking about 7 seconds on average.

 

[th0masp]

How long does it take for your copy of Photoshop to launch (pallets to appear after double-clicking the app)? Mine is taking about 7 seconds on average.

Not sure that is a factor. I just counted 7 seconds to launch PS into useable state too - on my Late 2011 MBP, loading in from a SATA-SSD.

 

[bsbeamer]

This test is HEAVILY influenced by your GPU settings within PS. Preferences > Performance needs to be setup correctly. METAL within PS is often disabled by default as well.

There's a thread comparing data from MBP16,1. Might be worth looking at some of those results as well:

Can someone with a 16” 8-core/8GB 5500m/32 or 64GB Ram please run this benchmark?

Hi, I’m gathering info to help me decide on an upgrade from my late 2013 rMPB and 2013 MP. I’ve found the Puget benchmarks to reflect the type of real world scenarios I’m interested in. So I’m very curious how a high spec 16” MBP would do on this particular one...

forums.macrumors.com

 

[choreo]

This test is HEAVILY influenced by your GPU settings within PS. Preferences > Performance needs to be setup correctly. METAL within PS is often disabled by default as well.

There's a thread comparing data from MBP16,1. Might be worth looking at some of those results as well:

Can someone with a 16” 8-core/8GB 5500m/32 or 64GB Ram please run this benchmark?

Hi, I’m gathering info to help me decide on an upgrade from my late 2013 rMPB and 2013 MP. I’ve found the Puget benchmarks to reflect the type of real world scenarios I’m interested in. So I’m very curious how a high spec 16” MBP would do on this particular one...

forums.macrumors.com

Interesting. These are my Photoshop Performance settings (OpenCL not checked)...

 

[bsbeamer]

"Use native operating system GPU acceleration" is the METAL setting. Using this does seem to speed up PS on SOME machines.

If you're driving multiple monitors from a single GPU, PS performance in testing will likely take a hit/reduction. Some of the items within the Puget test seem to be influenced by total available landscape/resolution and those are the only that would get a few extra total points driving multiple monitors. GPU performance itself will slow (at least slightly) when driving multiple monitors.

Rarely do I see any test results translate to an identical real-world experience. Once you're above a certain point, it's not impacting your day-to-day operations. FWIW, more than half of the Puget effect tests within PS are never used in real life either.

 

[choreo]

"Use native operating system GPU acceleration" is the METAL setting. Using this does seem to speed up PS on SOME machines.

If you're driving multiple monitors from a single GPU, PS performance in testing will likely take a hit/reduction. Some of the items within the Puget test seem to be influenced by total available landscape/resolution and those are the only that would get a few extra total points driving multiple monitors. GPU performance itself will slow (at least slightly) when driving multiple monitors.

Rarely do I see any test results translate to an identical real-world experience. Once you're above a certain point, it's not impacting your day-to-day operations. FWIW, more than half of the Puget effect tests within PS are never used in real life either.

I am driving (3) 27" NEC Monitors (but they are all QHD resolution - no 4K or above). Not sure why the scores I am getting are so much higher than some others with "Use native operating system GPU acceleration" unchecked?

 

[darthaddie]

I ran the benchmark

Here are my specs

Mac Pro 7,1
12 Core
96GB RAM (6 slots) - This could be a major factor as 6 six channel had 50% boost for me in Lightroom Bench.
AMD Radeon Pro VEGA II
Catalina
2 monitors - LG ultrafine 5k, Ben SW271 connected via TB3 and USB-C
Internal 256GB Samsung 960 Pro SSD for PS and Lightroom scratch
Full acceleration in Photoshop performance settings with everything turned on
Internal T2 SSD used for storing the benchmark
5 external USB and TB3 drives - Not is use during benchmark
one internal 512GB SSD connected via a generic M.2-PCIE card
No apps running except Google Backup Sync

* I listed my full specs to be better able to identify comparisons between similar machines

Results are the same as @choreo

 

[choreo]

I ran the benchmark
Results are the same as @choreo

Almost identical! I suspect (as you do) that having the same RAM configuration may be part of the reason?

 

[darthaddie]

Almost identical! I suspect (as you do) that having the same RAM configuration may be part of the reason?

I am pretty sure. My Lightroom export times reduced by almost 50% when I matched 6 DIMMS compared to 4.

 

[OkiRun]

Almost identical! I suspect (as you do) that having the same RAM configuration may be part of the reason?

The 7,1 film is finished and its open for a day. i would run this test but the second step download is not understandable to me.