TOPIC:

pszilard wrote:

It maybe that some configuration in Motion, may be required. …

No config options whatsoever in Motion. Your info about 20% CPU is probably of interest.

… Thanks!

I worked for +20y in marketing, so I'm adopting that Phil-Schiller-lingo (awesome, groundbreaking, earth shattering) very well…

pszilard wrote:

It maybe that some configuration in Motion, may be required. If it did this with 20% cpu, then 100% would by 5x faster - perhaps.

not necessarily... with Motion CPU is an auxiliary to GPU, and rendering a particles based project : GPU might be taking the charge

Max Yuryev just posted an extensive performance test of the 8-core Vega56 iMac Pro vs a top-spec 2017 iMac 27, emphasizing FCPX video editing:

Many of those tests were disappointing. Compared to a top-spec 2017 i7 iMac 27, the 8-core Vega56 iMac Pro slower at all these tests:

- About 1/2 as fast on FCPX stabilization
- 17% slower at rendering 1080p H264 with two LUTS and film grain applied
- 28% slower at rendering 4k H264 with two LUTS and film grain
- 20% slower at rendering to H264 4 x 4k with two LUTS and film grain
- 7% slower at encoding a pre-rendered 4k H264 file from a Sony A7RIII
- 27% slower at exporting a non-rendered 4k H264 file from a Sony A7RIII

The iMac Pro was much faster at encoding to H265, and vastly faster at editing RED RAW video. It was quieter than the iMac.

Max Yuryev's are the first real-world extensive tests involving one of the most common video codecs, which is H264. This apparently includes H264 variants such as Sony's XAVC-S and maybe XAVC-L. Prior to Max's tests, based on the very sketchy preliminary tests it looked like the iMac Pro either had a custom Xeon with Quick Sync or was using the encoding hardware on the Vega GPU. However that no longer appears the case.

Since more FCPX editors are now getting their hands on the iMac Pro, we'll have more test data soon. But if Max's tests hold true, it is disappointing that Apple could not make a machine which runs their own FCPX editing software faster on H264 media than a regular iMac.

This will not affect those with an all-ProRes workflow from acquisition to delivery, nor those acquiring and editing raw video. It appears the iMac Pro is somewhat faster than the trash can Mac Pro on H264, so those upgrading from the nMP will benefit no matter what the codec. However those editing H264 on a 2017 iMac using FCPX may be disappointed if they upgrade to the iMac Pro.

As I said before, there is a huge difference between the performance of the 8-core Vega56 and the 10-core Vega64 iMacPro. I have worked with both, and I can confirm what these tests reveal:

- When you work with regular footage on regular projects (as in this test) you need to get the right configuration for the iMacPro (which he hasn't), or you won't see much difference with a top specced 2017 iMac.

- When your workflows get more demanding (e.g. handling 5K and higher RAW footage or complex multicam projects) you WILL see a lot of difference, even with the 8-core Vega56 model. This is also confirmed in this test.

The sweet spot for modern FCP X workflows is the 10-core Vega64 model with either 32 or 64GB RAM. Get that and you will see huge benefits over any other Mac, no matter what kind of work you do. People who only deal with HD or 4K projects and who already own a recent top specced iMac or a MacPro should not buy a base model iMac Pro. That's why this test is kinda useless.

- Ronny

Did you export to ProRes HQ, or just ProRes 422? I've been testing for a couple of days now exporting Karsten's Motion Stress Test to ProRes HQ. Neither iMP (10-core Vega 64) nor MP (8-core D700) does it in less than an hour.

I like to test with real-world projects. I have just finished a complex 35' project with native 6K RED RAW on a 4K timeline. Edited natively to preserve the possibilities of tweaking the RED RAW settings while editing. The timeline played un-rendered at Best Quality with LUT and several color corrections applied to it. Shared to ProRes 422 and H.265. Had no issues whatsoever. Huge difference with my 2016 fully specced 27" iMac. Looking forward to seeing the results of your tests.

- Ronny

Using Karsten's project, my iMac Pro was MUCH slower, but only with Motion. I don't buy that it's that specific Motion project. Faster should be faster, period.

I need this:
Karsten's Motion project on a late 2013 Mac Pro, 8-core, D700's, do a RAM Preview, then export as ProRes HQ.

Seriously, it can NOT be this specific project. That means there are functions in that project that the iMac Pro can't handle well. Or is there another logic explanation?

ronny wrote:

People who only deal with HD or 4K projects and who already own a recent top specced iMac or a MacPro should not buy a base model iMac Pro. That's why this test is kinda useless.

- Ronny

true that ! except that it is useful for them to be confirmed : save 2 000 $ !

problem being ... to get such a CG into a extra thin computer it seems they castrated it ! what if the same applies to the 2018 iMac Pro yet to come ? so the computer won't melt ... still limited performances . Design is a bitch

Stu wrote:

...problem being ... to get such a CG into a extra thin computer it seems they castrated it ! what if the same applies to the 2018 iMac Pro yet to come ? so the computer won't melt ... still limited performances . Design is a bitch

This is commonly said, but there's no evidence the iMac Pro's performance -- IF it's significantly limited in any area -- is related to the form factor. I haven't observed other manufacturers demonstrate the engineering skill to make an all in one computer that thin, light, quiet and powerful using high-core-count Xeon CPUs.

But just because nobody else has achieved this doesn't mean the iMac Pro is somehow handicapped by the form factor. None of the tests to date indicate a significant problem with thermal throttling or acoustic noise on the iMac Pro.

Just watched Max's video and I am seriously upset!

If I was to swap my iMP to a maxed out iMac I would save Au$3,770 which is a significant amount!

I don't need:

• 10Gbps LAN
• 2 TB3 connections are enough
• I do not want to work on RED footage. H265 probably from my iPhone, but not for huge files
• Max I plan on i s 4k video
• TB3 external GPU still opens opens for later upgrade

I've had the iMP for 3 days, so can easily return it, however the iMac won't be here in time without a gap so I would need to trust TM and cloned HDD to do the migration.

I don't know what to do! Help.

Seeing your 'I don't need' list, I think you have every reason to be upset. Not with the iMacPro, but with your decision to buy a more expensive machine while your list clearly shows that you don't need anything that it has to offer over a less expensive fully specced iMac.

We will see more of this in the coming months. If you need real horse-power, the iMacPro will be the better choice over any other current Mac. If you don't need that extra power, the iMac or even the MacBookPro will be more cost-effective. Now that we have more options, it's up to us to make the right decisions based on our real needs. My advice: 3,770 AUS$ is indeed a significant amount. If you can still swap your iMacPro to a maxed out iMac, I would do it.

- Ronny

I have more trust is BAREFEATS than in Max.

barefeats.com/imacpro_vs_pt2.html

I have in order iMP 64 GB 10c Vega64 Hope to be not disappointed

I am having a look now...

I agree that the Barefeats tests are more accurate. I think you have made the right choice regarding the configuration of your iMacPro. Hope you wil enjoy your new machine!

- Ronny

My choice was driven by the (mis)understanding that the iMP will be faster for working on 4k video edits and "normal" use. I was also expecting my Promise Pegasus R3 to deliver up to 1,500MBps (as claimed) when on TB3. This also hasn't happened!

I can order an iMac on AmEx and hope that it arrives before my return date on the iMP so I can compare them here myself.

Still thinkin'

Just read the Barefeats page. Seems a bit more encouraging re the iMP. Also saw OWC Teardown, indicating that the iMP "could" be upgraded both for cpu and ram, so at least that would leave the door open for the future.

I had a friend run BlackMagic Speed Test on the SSD of his new iMac and the iMP was 50%. Geekbench 4 multi-core is also 50% faster, but single core is slower.

I think I'll stick with the iMP, as I don't want to change in a year's time, and according to OWC, it can actually be upgraded. Plus I expect FCPX to make better use of the h/w in a future update. Of course, I may change my mind

… just to drop my 5¢ into that topic, the 'academic one' first:

benchmark tests test benchmarks tests.
who cares for superb synthetic numbers, if the software makes no use of the hardware? or, vice versa, if the software is 'better' fit to hardware …

practical comment:
Published TWO different stress tests here, and meant as 'real world' tests…

Anyhow, looking forward for profound tests … and conclusions.
…

joema wrote:

Stu wrote:

...problem being ... to get such a CG into a extra thin computer it seems they castrated it ! what if the same applies to the 2018 iMac Pro yet to come ? so the computer won't melt ... still limited performances . Design is a bitch

This is commonly said, but there's no evidence the iMac Pro's performance -- IF it's significantly limited in any area -- is related to the form factor. I haven't observed other manufacturers demonstrate the engineering skill to make an all in one computer that thin, light, quiet and powerful using high-core-count Xeon CPUs.

But just because nobody else has achieved this doesn't mean the iMac Pro is somehow handicapped by the form factor. None of the tests to date indicate a significant problem with thermal throttling or acoustic noise on the iMac Pro.

there doesn't seem to be a problem, absolutely... that "sheet of a computer " won't burn over hard work.
but , and that's the point, the "commonly said," is that they've limited the performances of Xeon and CG to avoid overheating.

The Barefeat is comforting though , at least it " Pro " and not Youtuber testing

ronny wrote:

I agree that the Barefeats tests are more accurate. I think you have made the right choice regarding the configuration of your iMacPro. Hope you wil enjoy your new machine!...

It's comforting that the Barefeats tests look good -- they cover a wide but sparse spectrum.

However their tests are not focused on video tasks using the most common codecs such as H264, and variants like Sony's XAVC-S and XAVC-L. Unlike Barefeats, Max Yuryev 's tests DO focus on those, so it's not a case that Barefeats is more accurate, but that Max is specifically testing things that Bearfeats is not -- real-world items using FCPX on the most common codecs in use today. It is expected their results may differ.

Max's H265/HEVC test showed the 8-core Vega 56 iMac Pro was 2x faster than the top-spec 2017 iMac on FCPX export. Due to the extreme computational difficulty of HEVC, this implies that FCPX on the iMP was using some form of hardware acceleration. Why similar gains did not appear on H264 is a mystery.

The problem is large-scale H264 production is common, especially in field documentary and reality TV. I am editing a doc right now that's composed of 232 hr (7,507 clips, 20 terabytes) of 4k H264 material. It has over 150 4k multi-camera interviews, some three and four camera. H264 workflow at this size is as demanding as smaller productions using RAW footage. The fact that FCPX can do this on a 2017 iMac is incredible. However more performance would be useful.

So everyone using H264 is not on the lower end of the production spectrum. If the iMac Pro mainly benefits those using RAW and ProRes acquisition, that's a limited user base. However the H265/HEVC results indicates FCPX on the iMac Pro is using hardware acceleration -- likely via AMD's UVD/VCE logic. If FCPX 10.4 on the iMac Pro isn't using this for H264, maybe it could in the future. It's possible the code path for HEVC export was newly written, and not applicable to H264, and Apple may intend to eventually harness the presumed UVD/VCE acceleration for H264.

But if it does not, here is the risk to Apple: The rumored 2018 six-core i7 iMac will probably use the i7-8700K which is six-cores and has a 4.7Ghz boost clock -- and Quick Sync. If by that time Apple has not figured out how to enable hardware acceleration for H264 on the iMac Pro, the inevitable benchmarks comparing the 2018 top-spec iMac with the iMac Pro on H264 will not be pretty.

Thank you for this clarification, very useful. With the iMacPro we have a lot more options now in the Mac line-up, and I don't see why there should be any competition between the iMac and the iMac Pro. Quite the opposite: both systems have their specific target groups. This confirms my former statement that, from now on, people should consider more carefully what kind of system will best benefit their workflows. Having more options is good, but this also leads to having to make the right choices.

- Ronny