i'd like to add my experiences/observations as well as recommendations based on my personal quest for the "best" system:
TL;DR:
-
only for perfectly isolated and prioritized use cases you can have the
"best" system. (one tool for the job), however you have options that can
provide an adequate balance of .e.g multicore vs. single core as well
as GPU compute performance based on your usage profile.
- PTGui
by now is utilizing almost all of the critical/limiting components of a
system (also depending on your choice of workflow) and i consider it
worst case to cater for in a custom build since you need to test/max out
and figure out the weakest link (which for me turned out to be
CUDA/OpenCL compute power).
- custom build implies PC platform -
not exactly for the cost/benefit ratio vs. apple, which ist amost equal
when picking high quality components (minus the pain of using windows),
but the flexibility in shaping a hardware config to a certain processing
load.
verbiage:
- since i had the possibility to get
my hands on a few systems i did benchmark them with the same set of
input data: HDR pano, exposure brackets (3 shots), 93 x 46MP of 8bit
JPGs(for portability of the fileset), resulting in a 440MP pano. again,
this is a very specific use case taxing primarily RAM and SSDs. by no
means this is representative for your particular workflow and filesizes
nor does it want to be. but it it helped me to put things into
perspective(PT11, latest release including universal binary for M1)
system 1: mac mini (late 2018, 6core, 3.2GHz, 32GB RAM)+ blackmagic eGPU (radeon 580 pro)
align: 14 seconds
optimize (vignetting, camera curve): 6,8 seconds
render TIFF, 16bit, no alpha, 100% (440MP): 5 min 16 seconds
system 2: macbook air M1 (16GB, 8GPU cores)
align: 40 seconds
optimize: 6 sec
render: cancelled after 7 minutes
system 3: TR3970x (32 cores), 128GB RAM, NVMe, RTX 2080ti
align: 15 seconds
optimize:1 sec
render: 44 seconds
GPU: max 20% CPU load, RAM, SSD negligible, GPU hitting 99%
CPU only, CPU load hit 65%, render took 2 minutes 16 seconds
after
having used system 3 as my main workstation for about 6 months now i
decided to "downgrade" the system to a ryzen 9 5990x (16 cores)due to:
-
my usage profile does only benefit a fraction from the massive
multicore processing power as well as PTGui mostly benefiting from more
CUDA/OpenCL compute power on my system.
- heat and cooling are
immense and to actually take advantage of a durable multi core
performance you need to go big gun custom water loop
- leave
alone a sustainable energy profile when NOT being in need of 32 cores at
the same time AND having software maximizing/balancing GPU/CPU in this
case. HEDT my a** - this thing belongs in a server room.
- single
core performance (most of a photographers software) is average at best
with the TR and i'll gain more efficiency from the ryzen9 cpu while
still maintaining a very reasonable performance for the multicore
portion of my usage profile.
while i'll get back some of my
initial invest i'll wait for apple to actually come up with a
workstation SOC along with adequate GPU compute power - it will be there
and killing - however for now i need to get sh** done.
best,
t