You won't get a lot (any?) benefit from RAIDing as the benchmark is designed to stress graphics cards. Very little writing is done,. I recall we did that and it made no difference with real hard disks as opposed to SSDs,
iMac Pro, 10 core, 3.0GHz, 64 GB RAM, Vega 64
Exporting to Promise Pegasus 32 TB (28TB usable), 8 bay, Thunderbolt 3, RAID 5
Using the version of the test currently linked in the OP. Is there an update to this?
24.33 seconds avg
20.8 seconds avg
Just for fun, H.264 exported to the internal 2TB SSD drive, 25 seconds. So, never edit on your system drive, it's not necessarily the fastest.
What I find interesting is that I do 30 second TV commercials in 4K, export to ProRes, in much less time. Which is real-world work, so I base my opinions on that. But this is a fun test. I wonder if folks hit the start on their stop watches when they hit the save button or when the background task clock starts moving.
My latest tests on 2017 top-spec iMac 27 and 10-core Vega 64 iMac Pro, all running macOS 10.13.6 and FCPX 10.4.3. Each test was done five times on five fresh libraries to avoid caching effects, and the average taken.
Background rendering was off, timeline was not pre-rendered, Time Machine was shut down. Ideally when doing benchmark testing Spotlight should be shut down but I didn't do that.
To improve timing precision, I would suggest everyone subsequently copy and paste the BruceX timeline to itself 4 times and divide the resulting execution time by 4. However -- these numbers aren't directly applicable to the previous single-pass numbers, but somewhat faster. This might imply a significant % of BruceX timing is startup overhead and isn't really measuring the core task. Conversely it might imply the 4x timeline is benefiting from cache effects of the repeating test. This shows how complex a simple-looking benchmark can be. Until this is understood or Alex posts an "official" new test procedure, I recommend we keep running it as usual.
However there is considerable run-to-run variation, so it's a good idea to run several iterations and take the average. Definitely background rendering should be disabled.
Time was started when hitting the share button and stopped when the file loaded in Quicktime Player. I tested both ProRes 422 and H264 4k using "fast" export option. Note the H264 export is 4k not 5k.
Quick question for everyone. I am planning to build a new Hackintosh. For best editing in a 4K workflow with FCP X. Would sticking with the two current 7850 be good or would any of you upgrade to two RX560s? I could also go with a single 7950 GPU, if that would be the better bet?
I mostly shoot with my iPhone at 4k. So, I don't necessarily need the faster most powerful GPUs.
My guess the reason for the slow time is the video card. I'm running dual monitors from the card, I'm sure that doesn't help.
I just replaced the above EVGA GeForce GTX 560 2048 MB card with a Sapphire R9 280X 3GB graphics card. Ran BruceX and as always followed the instructions. The R9 280X is over 4 times faster than the GeForce!
zzchu wrote: MacPro 5,1 (Mid 2010) 3.46 Ghz 6-core
Sapphire Radeon RX 580 Pulse 8gigabyte
FCPX 10.4.2 and Mac OS 10.13.4.
Export to ProRes:
Haven't done this test in over a year... but
Final Cut Pro X 10.4.7 update seems to have cut the time from 22 seconds to about 17 seconds with same hardware (Mac Pro 5,1 - 6 core, AMD RX580 and Mojave 10.14.6)
Model Identifier: iMacPro1,1
Processor Name: 10-Core Intel Xeon W
Processor Speed: 3 GHz
Number of Processors: 1
Total Number of Cores: 10
Memory: 64 GB
Boot ROM Version: 1037.60.58.0.0 (iBridge: 17.16.12551.0.0,0)
BruceX residing on and exporting to the system drive, 9 seconds.
BruceX residing on and exporting to Pegasus RAID, 8 seconds.
We have exactly the same iMac Pro model and config, however mine ran BruceX much faster. I don't know why. Was your output H264 or ProRes 422? Mine was running FCPX 10.4.8 and Mojave 10.14.6, Pro Video Formats 2.11.
Since BruceX runs so fast on modern hardware/software, I appended the timeline 5 times and divided the running time by 5.
My BruceX numbers for 10-core Vega 64 iMac Pro below.