I finally have my Mac Pro updated to where I want with the extra RAM I ordered so this morning I ran the Puget Systems Photoshop benchmark. At the moment, it's the only "use case" benchmark vs. synthetic benchmark that I could find that a) works cross-platform and b) has recent data from other systems to compare to. My results in the graphic attached, Puget's most recent data at the following link:
Photoshop typically doesn't require a HEDT processor, but with Intel drastically reducing the price of their new X-series CPUs and AMD greatly improving single threaded performance in their 3rd Gen Threadripper CPUs, is this finally going to change?
www.pugetsystems.com
A few points of comparison: My overall score of 926.2 is less than 5% lower than the highest overall score which went to one of the newest AMD processors, the AMD Ryzen 9 3950X 16-core. Impressive performance for the AMD at it's price point but it's also a consumer chip and not entirely comparable (there are limitations in consumer chips). On the other hand, given that it has a almost 10% higher base clock speed, almost 7% higher turbo clock speed and 92% more cache (73MB vs. 38MB), one could argue that the only 5% total performance increase is a bit underwhelming. But the price! Very inexpensive, so kudos AMD (and a very low TDP, thanks to that 7nm).
On the "General Score" my system actually gets the highest score of any of the Puget Systems: 115.3, vs the top score of 110.3 for the AMD TR 3960X 24-core--and with 33% fewer cores, too boot.
On the filter score, I only got a 76.7, only good enough for 5th place, but the top score is only 81.1, so a little less than 6% different--and the top score in that category goes to an Intel 9900K 8-core anyway, not an AMD chip. The best AMD chip is the Ryzen 9 3950X 16-core, at 79.9, or just about 4% better than my system.
The Photomerge score I am getting is bizarrely low. In fact, without that being so low my system may have been the top overall performer. I suspect something is wrong here as it scores a 79.1, so low it's off the bottom of the chart. On the bright side, I never use Photoshop to merge photos, although I don't know if this is actually a Photoshop issue or something in Puget's benchmark, which is still listed as a "beta."
Finally the GPU score; they use the same Nvidia RTX 2080 Ti card with 11GB of GDDR6 memory vs. my system, which has the Radeon Vega II with 32MB of HBM2 memory. Scores are tight there, not surprising. My 99.8 score is the third highest behind an Intel 9900K and an AMD Ryzen 7 system, but not sure we can read much overall into this score. The differences across the top ten scores are negligible and of course Photoshop doesn't leverage the GPU as heavily as many other applications do.
Some caveats: It seems that they may have run all of their benchmarks on version 20 of Photoshop; I am on version 21.0.2. I could downgrade just to run the tests, but the oldest version I can install is actually 20.0.8, so I'm not sure what difference it would make. It could be why the photo merge score is odd though. Also, Photoshop is not very good at all in terms of using multiple cores/threads. That means the raw speed of the CPU and overall CPU/system architecture have a greater impact than core number. This was obvious as I ran the test as the CPU monitor showed the system barely doing anything....CPU usage rarely spiked to more than 50%, and typically was way down in the 10% range. Same with turbo boost, it mostly sat at the base Ghz and I only saw it spike as high as 4.1Ghz briefly.
So, a real world benchmark of a sort. Performance of the Mac Pro seems to be excellent, on par with the best systems Puget lists, with the exception of that off Photomerge score. I'm going to email Puget about that, as it seems like it has to be bug. Would an AMD processor be cheaper? Yes. Are they "40-50% better" as we've seen claimed here? Certainly not in this benchmark.
Meanwhile, if you intend to use Photoshop heavily with your Mac Pro, these results might be interesting and useful to you.
View attachment 885864