Seems Apple is getting more and more open. Reviews of the iMac Pro with FCPX 10.4 before official releases. That's the fourth one I've seen today. Third party benchmark test results are what I'm really wanting to see. And what about heat issues in the long run?
Vincent Laforet said FCPX on the 10-core iMac Pro was about 2x faster at transcoding from H264 to ProRes optimized media than his 2017 top-spec iMac 27. He also posted numbers from various editing software.
This was a key number because it indicates the iMac Pro has some kind of hardware encode/decode acceleration, despite using a Xeon CPU.
I think the video said US$5,100 for the shown config.
It is so funny to me, as yesterday I ordered a specced out iMac 27" from Apple! But the joke is on them, as I can return/cancel within 14 days. So now I am awaiting the official price tomorrow, before deciding if I want to switch to the Pro.
FCPX.guru wrote: ...if it's only "up to 45% faster than the 2013 Mac Pro", is that 45% boost worth $5K?
What about heat issues when pressed to export a hour TV shows as MXF?....
I know from my own testing of a 12-core D700 Mac Pro vs my top-spec 2017 iMac, the iMac is 2x faster than the nMP on certain tasks -- import and transcode to ProRes, export to H264, and smoother on scrubbing a 4k H264 timeline. So for these tasks the 10-core iMac Pro might be 4x faster than a nMP.
However for other things the nMP is pretty fast, such as stabilizing a Pro Res clip -- about 2x faster than my iMac. Likewise on certain effects such as Aged Film, Add Noise, or Sharpen, the nMP is about 2x faster (if using ProRes) than my iMac.
So even today the "ancient" nMP is pretty fast on a ProRes workflow. It's also very quiet and can handle being pushed to the limit for days at a time. I've had my 2017 iMac transcoding for several days straight and so far it's held up OK, but this is a valid concern. How will the iMac Pro handle this for sustained periods such as weeks/months on end?
The "45%" figure is merely comparing the GeekBench multi-core results of a 12-core nMP vs a 10-core iMac Pro. As shown above on real-world FCPX tests, the difference is all over the map depending on your codec, plugins, workloads, etc. The 45% figure is also not comparing the top nMP to the top 18-core iMac Pro.
When Apple shortly releases the price for each config we can better decide whether it's worth it.
While comparisons are to the current MacBook Pro in the above linked article both the current MBP and iMac Quad i7 handle H.264 encode decode much faster than the late 2013 MP (number of cores dependent).
My poor laptop has 16 GB not 128GB of RAM. It has 4 cores not 10, and it has the Pro 460 graphics with 8GB of graphics memory. It is designed around compact size, light weight and low power consumption/battery life, factors that just don't apply to the iMac Pro. As a result transcoding half an hour of video on the MacBook Pro takes a quarter of the time on the iMac Pro. Four times speed difference is really noticeable on practical jobs, (A little over 3 mins on the iMAC Pro vs around 12 mins on the MacBook Pro, Handbrake File HD conversion - latest release). daVinci Resolve is also showing similar 4x to 6x speed ups. This machine is fast.
1- iMac's have that CPU with H.264 encoding acceleration, the 2013 MP lacks it, so it encodes that class of codec slower, yes.
2- There was nothing "real world" about any testing/comparisons I've read anywhere. No one edited a full production start to finish, only small segments of existing timelines.
3- No one ran a test of a 1 hour TV show exporting to MXF or something similar (not H.264) to see if the fans rev up too loudly or if there are heat issues.
Yes, we'll have to wait for folks to get them and work with them and hear what "real world" FCPX and Motion work on them is really like. But for a generalized 45% bump, I think $5K plus is a steep price to pay...
Believe that all 64 bit "Pro" software supports full usage of all multi cores/processors/GPUs.
Apple rewrote Grand Central Dispatch earlier this year and I think Apple stated at WWDC 2017 that code takes advantage of the new 18 core Xeons. 36 cores is the virtual core count.
AFAIK FCPX has always supported using all available cores, I think??
Doesn't mean that all cores are used all the time.