-
iMac Pro
-
A blown out HP Z840 is expensive. (And all the guys will jump in and say “but I can build my own PC”) –
In the real world, a pro company is going to buy a Z840 or a Dell Precision top of the line.
Add the NVidia GTX 1080ti (the new Radeon card is said to be the equivalent of the GTX 1060 – honestly I don’t know the difference).
Now, add the amazing HP Z31X monitor ($4000 retail) that they showed at NAB (auto calibration, selectable LUTs)now you have an expensive hi end workstation. And I am sure that Boxx make an equivalent, or even faster system.
I don’t want to hear “but I can build my own PC for a fraction of the price”. Good for you.So these solutions are both around the $10,000 or a little more per workstation.
The flip side of this is the people that remember that you could buy a cheese grater Mac Pro Quad Core with 4 Gig of RAM, and it would run FCP 7 perfectly for $3000. These days, with Resolve, CC2017, Smoke – and no need for the “video gear” – I guess this price range is ok.
Bob Zelin
Bob Zelin
Rescue 1, Inc.
bobzelin@icloud.com -
Since the iMac Pro will use Xeon CPUs does that mean it won’t have Quick Sync to speed up H.264 encoding?
-
[Andrew Kimery] “Since the iMac Pro will use Xeon CPUs does that mean it won’t have Quick Sync to speed up H.264 encoding?”
Intel does not currently put Quick Sync on CPUs with more than four cores, so unless this is changed the iMac Pro will not have Quick Sync. This means it will not have hardware-accelerated encode/decode for H.264, H.265, VP9, AV1 or any wrapper format using those codecs such as AVCHD and XAVC-S. The Mac Pro doesn’t have Quick Sync either but that was designed in the pre-4k era and before H.265 was looming on the horizon. H.265 is much more compute-intensive than H.264, therefore needs hardware acceleration even more.
OTOH since Apple makes hardware, OS and (in the case of FCPX) the application, in theory they could rapidly add code to use AMD’s hardware acceleration VCE (Video Coding Engine): https://en.wikipedia.org/wiki/Video_Coding_Engine
VCE like nVidia’s NVENC resides on the GPU but is functionally separate from the GPU — it is a “bag on the side” from an architectural standpoint. Long GOP encode/decode cannot be meaningfully accelerated using traditional GPU methods, it requires fixed-function logic, IOW an ASIC (Quick Sync being one example).
Whether VCE is actually practical would depend on an extensive technical evaluation of performance, image quality, and how many software changes are required at the OS and application layers. It would also depend on how functional and stable the software development framework which supports VCE. If the hardware works but the development libraries and other software does not, or is not available in macOS, then it’s not an option.
The lack of Quick Sync in most Xeon CPUs has long been an issue for the video world. It potentially impacts all editing software but some software like Premiere either doesn’t use it, doesn’t use it on Mac or doesn’t it effectively. FCPX has long used Quick Sync very effectively so the impact is significant.
I’ve never seen a good explanation from Intel or anybody else why Quick Sync is withheld from CPUs with more than four cores. A possible explanation may be it requires GPU on-chip resources, even though the actual encoding logic is separate. This would require Xeon to have the full on-chip GPU which is a significant hit to the transistor budget. Xeon is mostly a server and workstation chip. Servers don’t need an integrated GPU and workstations will have a discrete GPU, so putting the integrated GPU on chip would waste transistors. The only loser is the subset of the video editing community that relies on Quick Sync. It was probably a tradeoff.
-
sorry Surface Studio you are right
-
The Radeon Vega is more closely comparable to a 1080ti just to FYI
Radeon Vega:
Better floating-point performance 13,107 GFLOPS vs 10,609 GFLOPS Around 25% better floating-point performance
Significantly wider memory bus 2,048 bit vs 352 bit More than 5.8x wider memory bus
More shading units 4,096 vs 3,584 512 more shading units
More texture mapping units 256 vs 224 32 more texture mapping unitsGTX 1080ti
Much higher effective memory clock speed 11,008 MHz vs 2,000 MHz More than 5.5x higher effective memory clock speed
More memory 11,264 MB vs 8,192 MB Around 40% more memory
Higher pixel rate 130.2 GPixel/s vs 102.4 GPixel/s More than 25% higher pixel rate
More render output processors 88 vs 64 24 more render output processors
Higher memory clock speed 1,376 MHz vs 1,000 MHz Around 40% higher memory clock speed -
It’s like the traschcan but it has a screen. Wheee…
-
[Scott Witthaus] ” but if you are really doing this as a “pro”, it’s not that big a cost. One or two jobs?”
The issue for me isn’t the cost as much as it is sinking that cost into a non upgrade-able system. If I’m investing 5K+ I want to know that I won’t regret it next year, which is not possible with any closed system, be it Surface or Imac.
Herb Sevush
Zebra Productions
—————————
nothin\’ attached to nothin\’
\”Deciding the spine is the process of editing\” F. Bieberkopf -
Not sure if this has been mentioned yet, but people that got a look at it in person said there’s no apparent door/hatch to access the RAM like on other iMac’s so RAM might not be upgradable.
Apple confirming official support for eGPUs is great news, but that feature isn’t expected to be released to the masses until Spring 2018 (and it requires Thunderbolt 3).
-
No else said. I will:
2018: iMac Pro running iMovie Pro. wink wink
Reply to this Discussion! Login or Sign Up