The world’s most
frustrating 2-12
hour benchmark
It shouldn’t come as much of a surprise when
we say that assessing how these machines
handle image editing varies depending on what
you’re doing. When you’re packing loads of
memory, fast storage, the fastest processorsmoney can buy, and great graphics cards, many
tasks that take a while on a standard desktop
are too easy for our monster rigs. We searched
for workloads that actually made sense, but
kept coming up with operations that finishedwithin seconds, or were too specific to be of
much use. In other words, if you’re looking for a
workstation for image editing, both machines
fulfill the role competently.
Not all image-based tasks are created equal,
though, and if you’re looking for serious system
draining, image stitching is a good option. The
concept is to take a series of photographs that
cover a wide panorama, and stitch them
together to create a massive final image. In
order to get consistent coverage for these
images, a “robot” is often used, and that’s
exactly what was done to get the live data set for
our testing. The shots in question were
bracketed for improved contrast, and the final
image count came in at a cool 690 images, with
each measuring 5184x3456 pixels.
There’s some necessary overlap between
images, but the final panorama is
136,758x50,502 pixels, and weighs in at 22GB.
The software we used for this test, Autopano
Giga 4.4, uses as many CPU cores as you can
throw at it, loads of RAM, and requires fast
storage, too. No real surprises on the result
front – the added core count of the Intel build,
along with the larger capacity memory, makes a
real difference.IMAGE STITCHING
When considering CPU cores, how do you best
explain the proposition behind a workstation,
like the two builds we have here?
Going from a four-core, eight-thread
processor to an eight-core, 16-thread one can
cut render times in half. If you shave 15 minutes
off a 30-minute render, and do that twice a day,
five times a week, that’s 2.5 working hours
saved in a week. Multiply that out by a year, and
it’s around 125 man hours saved. Take the
average wage of an experienced video editor
(£20 an hour), and that’s a total of £2,500
saved, purely in time, in a single year.
What we’re getting at is that, for the
self-employed, or the corporate busybodylooking to improve their company’s efficiency,
investing in good systems, much like
infrastructure, will always pay for itself in the
long run. And that’s something you partially see
with the two systems we have here. Ryzen’s
value proposition in contrast to our Intel build is
huge. We knew from the start that Intel could
never win a fight based on trying to spec a
similarly priced machine, so performance was
the goal – smashing the opposition through
sheer brute force, graphically, computationally,
and in I/O support and memory was key.
We’ve gotta give credit to Ryzen – it’s
exceptionally well equipped at an outstanding
price in comparison to Intel, but it wasn’twithout its limits. Initial setup and getting
memory to operate above 2,133 was still
challenging, even with the latest BIOS updates,
and Gigapan, in particular, proved troublesome
(although that’s also true of the Intel rig).
Ultimately, recommending one rig over the
other is difficult. If you’ve got the cash to spare,
the Intel one is easier to set up, more powerful,
and far more stable than its Ryzen counterpart.
But, in all honesty, with AMD’s 16-core
Threadripper HEDT platform and Intel’s
Skylake-X right around the corner, it would be
better to see what both of those new platforms
muster up first, before committing to either
processor ecosystem.INTEL CONCLUSION
Pricey build, pricey prize
INTEL BUILD
Feature
AMD vs.INTEL
GIGAPAN IMAGE STITCHING
VIEWSET RYZEN WORKSTATION ZERO POINT INTEL WORKSTATION
Kolor Autopano
Giga 4.4 (Time) 1h 57m 34s (4%) 2h 1m 55s 1h 29m 6s (27%)
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%Our desktop zero-point has a Core i7-7700K running at a stock frequency of 4.5GHz, an Nvidia GeForce GTX 1080 8GB,
16GB of Corsair Vengeance LPX DDR4-2400, and a 250GB Samsung 850 Evo, mounted on an Asus Maximus IX Hero mobo.