Best CPU for UF: i7 or Ryzen?

You can add this to the results:

Ryzen 9 3900X 2x12 @ ~4.05 GHz: 2m51s

It doesn't quite beat a 2950X, but it's extremely close.

22.09.2019 16:14:15: Starting job Benchmark.
22.09.2019 16:14:15: Anti-aliasing off.
22.09.2019 16:14:15: Motion blur off.
22.09.2019 16:14:15: Rendering in PNG format.
22.09.2019 16:14:15: Calculating Benchmark in C:\Users\Phillip\Documents\Ultra Fractal 6\Fractals\bench.upr.
22.09.2019 16:17:06: Finished calculation. Time: 0:02:51.79.
22.09.2019 16:17:06: Saving to I:\jpgs\Benchmark.png.
22.09.2019 16:17:11: Job finished.

My results:
9/24/2019 8:25:58 AM: Starting job Benchmark.
9/24/2019 8:25:58 AM: Anti-aliasing off.
9/24/2019 8:25:58 AM: Motion blur off.
9/24/2019 8:25:58 AM: Rendering in PNG format.
9/24/2019 8:25:58 AM: Calculating Benchmark.
9/24/2019 8:32:42 AM: Finished calculation. Time: 0:06:43.69.
9/24/2019 8:32:42 AM: Saving to C:\Users\jamad\Documents\Ultra Fractal 6\Fractals\Benchmark.png.
9/24/2019 8:32:46 AM: Job finished.

MSI GS75 Stealth, Windows 10 Home, Intel Core i7-9750H @ 2.6 GHz, 16 Gb DDR4 RAM.

Added these – thanks!

But, @Phillip: Did you overclock your (enviable) Ryzen 9? AMD gives 3.8 GHz as its base clock, and hence should only go beyond this when fewer cores are used...

No, thats just the stock boost behavior. 3.8 GHz is the base clock but given enough power and cooling the CPU boosts higher than that. 4.05 isn't even that impressive, I've heard of others whose CPU boosts to over 4.1 on all cores.

It should boost higher for fewer cores used, up to 4.6 at one core only but i have never seen more than 4.5.

Ah, okay. Thanks for the clarification!

Speaking of overclocking... I clocked the CPU to 4.3 GHz on all cores and this is the result:

03.11.2019 12:48:47: Starting job Benchmark4300.
03.11.2019 12:48:47: Anti-aliasing off.
03.11.2019 12:48:47: Motion blur off.
03.11.2019 12:48:47: Rendering in PNG format.
03.11.2019 12:48:47: Calculating Benchmark in C:\Users\Phillip\Documents\Ultra Fractal 6\Fractals\bench.upr.
03.11.2019 12:51:27: Finished calculation. Time: 0:02:39.51.
03.11.2019 12:51:27: Saving to I:\pngs\Benchmark4300.png.
03.11.2019 12:51:31: Job finished.

The gains are obviously minimal, but it's enough to pull ahead of the TR 2950X by about half a second.

//edit:
Just for completeness' sake, here's a run from my Laptop, an Intel Core i7-4710HQ @ 2.5 GHz:

03.11.2019 20:11:09: Starting job Benchmark.
03.11.2019 20:11:09: Anti-aliasing off.
03.11.2019 20:11:09: Motion blur off.
03.11.2019 20:11:09: Rendering in PNG format.
03.11.2019 20:11:09: Calculating Benchmark in C:\Users\Phillip\Documents\Bench.upr.
03.11.2019 20:25:54: Finished calculation. Time: 0:14:45.68.
03.11.2019 20:25:54: Saving to C:\Users\Phillip\Desktop\Benchmark.png.
03.11.2019 20:26:00: Job finished.

Thanks! You've really passed the Threadripper, though rather by 0.13s – in the table I rounded all the figures.

11/25/2018 1:44:41 AM: Finished calculation. Time: 0:02:39.64.

vs.

03.11.2019 12:51:27: Finished calculation. Time: 0:02:39.51.

Apple M2 3.3ghz 8x8 cores:

11-11-22 16:47:49: Starting job Benchmark.
11-11-22 16:47:49: Anti-aliasing off.
11-11-22 16:47:49: Motion blur off.
11-11-22 16:47:49: Rendering in PNG format.
11-11-22 16:47:49: Calculating Benchmark.
11-11-22 19:11:37: Finished calculation. Time: 2:23:47.83.

Slowest result ever for the UF benchmark?

Meanwhile, when I run Cinebench 23 this Mac scores 7955 multicore/1557 single core, which seems to be typical for M2. My Mac is faster at rendering than the Intel i7 9750h which did the UF benchmark in only 6m44s.

Threadripper Pro 5975WX (32/64C), 4.2-4.3GHz when running all cores. 512GB of PC3200 ECC RDIMMs / 8 channel, Windows 10 Pro Workstation.

I'm still messing with the machine and I've had it run up to 10s faster than this on NPS1 but I may have also had some other setting changed related to power management at the time so take that with a grain of salt.

12/15/2022 8:47:25PM: Starting job Benchmark.
12/15/2022 8:47:25PM: Anti-aliasing off.
12/15/2022 8:47:25PM: Motion blur off.
12/15/2022 8:47:25PM: Rendering in PNG format.
12/15/2022 8:47:25PM: Calculating Benchmark.
12/15/2022 8:48:39PM: Finished calculation. Time: 0:01:14.08.

I'm on Windows 10 so I needed a workaround using SysInternals "psexec -a 0" (I don't know exactly why it works since it's forcing cpu affinity and the 0 can be any number but I gave it a shot) to run on all cores in L3 as SRAT & NPS4 (4 numa nodes on this processor) mode, or it just works in NPS1 / no L3 as SRAT mode which is less optimal on higher core count AMD. I don't know how many cores per node this will make things work right on.

@Eric B : I'm working on a native M1/2 version of Ultra Fractal... would be interesting to see where you end up then. Although I wouldn't really expect too much: in my testing so far the M1 is never better than a comparable Intel processor.

Thanks Frederik. Looking forward to it.