[Erik Lindahl]

I’m not sure the notion of “if you rely heavily on effect and animation – go with adobe” I correct or even true. Dynamic link is extremely flawed and lacking, consistency between AE and PrPro is poor and just working in PrPro in general give sub-par results to FCPX’s engine.

Yes, for animation, compositing, motion graphics etc AE rules but in this area I really don’t see PrPro very strong. Adobe real f:ed up not taking their amazing animation package into their editor.

That said FCPX isn’t perfect and I’d always go to AE for detailed controlled work. But for example color correction is vastly better in FCPX given it does have its limitations.

I think the main / only hurdle for FCPX is it’s new workflow. That can set a lot of people off – my self included.

[Erik Lindahl]

All true and sound. But I see the good scaling is in Episode Engine. I wonder how their desktop product will fair. I don’t think it scales as well due to lacking the split and stich tech. Interesting never the less.

[Erik Lindahl]

Very interesting in deed. Depending on your needs however, two 6-core systems would possibly fair better than this and land in the same ball-park cost. I know certain codecs scale much worse than h264 / x264. If you could do some WMV, MPEG2 and general QuickTime (ProRes, Uncompressed, possibly DV) codec tests on top of the above that would be awesome.

I wonder if we’ll see Telestremes GPU accelerated x264 codec for Episode by NAB? They’ve already released it for Vantage which is much more expensive system.

[Erik Lindahl]

Nice. Video encoding can be very multithreaded. Was the above 12 core vs 8 core or 6 core?

[Erik Lindahl]

No worries. I also think moving forward, even if we get better threading in apps (RAM-previews in AE or dynamic link rendering in PrPro CC are a joke today) I think the main boost will come from GPU-acceleration. In video-land FCPX, PrPro CC and Resolve have shown this clearly.

Moving on things like RAW-decoding will move to the GPU I’m quite sure of. Already today RED CINE-X has support for this for RED RAW (given it’s still BETA) as well as Resolve has GPU-support for Sony F65 decoding.

Other features should also move there in due time. That’s where by far you can get the biggest boost in performance. Even a maxed out wintel-Xeon today might have 2X the CPU power than a maxed out MacPro 2013 in an optimal multithreaded scenario. If you instead run computational tasks on the GPU’s you’re talking multiple times greater increase in performance. It won’t work for everything but it does work for a lot.

[Erik Lindahl]

Well, mutli-threading has always been a problem. A lot of apps don’t scale well or scale to 3-4 cores. Here the 12-core machine will suffer.

What’s more, currently the multi-core machines take a bigger hit in general performance due to poorer base and turbo speeds per core.

I’d say the 8-core is the current-gen sweet spot if you can afford it. Otherwise the 6-core. The D700 are a given to me as well as at least 32 GB of RAM (we went with 64).

[Erik Lindahl]

[Ben Mullins] “I noticed that too, that the more cores you want the slower the processor speeds are. Why?”
Basically each CPU as a threshold of how much power (W) it’s allowed to use before it starts to damage the system. Hence a 12-core machine has to be clocked lowered than a 4-core machine.

However, all CPU’s have the option of using their “turbo boost” feature which will change the picture above quite a lot.

4-Core Machines scale from 3.9 Ghz with 1 core to 3.7 Ghz with 4 cores active
6-Core Machines scale from 3.9 Ghz with 1 core to 3.6 Ghz with 6 cores active
8-Core Machines scale from 3.9 Ghz with 1 core to 3.4 Ghz with 8 cores active
12-Core Machines scale from 3.5 Ghz with 1 core to 3.0 Ghz with 12 cores active

As I wrote in another reply, the 4, 6 and 8-core machine will be very similar when 1-4 cores are used. The 6-core will have a slight edge at 6-core activity, the 8-core at 8-core. The 12-core machine will only be favorable when 10 or more cores are being used.

[Erik Lindahl]

[Mark Dobson] “I’m still waiting for a New Mac Pro vs a topped out 2012/13 27” iMac comparison. “
We have a pretty much maxed out 8-core MacPro on order I will test against a MacPro 2008 that it replaces as well as an iMac 2012 I use at my home studio. Sadly this will be weeks or even months away seeing how Apple is having some production issues with the new MacPro.

[Erik Lindahl]

[Oliver Peters] “Or, if you don’t believe in the future of 4K mastering and don’t use FCP X near-exclusively, you could go with the 6-core and D500. Faster processors than the 8 or 12-core.”
This isn’t really true. Even if it gets a bit messy with base clock and turbo clock speed of the new MacPro’s it virtually looks as follows:

If 1 core is used
– 4-Core = 100% CPU speed
– 6-Core = 100% CPU speed
– 8-Core = 100% CPU speed
– 12-Core = 90% CPU speed

If 2 cores are used
– 4-Core = 100% CPU speed
– 6-Core = 97% CPU speed
– 8-Core = 103% CPU speed
– 12-Core = 86% CPU speed

If 4 cores are used
– 4-Core = 100% CPU speed
– 6-Core = 97% CPU speed
– 8-Core = 95% CPU speed
– 12-Core = 86% CPU speed

If 6 cores are used
– 4-Core = 100% CPU speed
– 6-Core = 146% CPU speed
– 8-Core = 138% CPU speed
– 12-Core = 122% CPU speed

If 8 cores are used
– 4-Core = 100% CPU speed
– 6-Core = 146% CPU speed
– 8-Core = 184% CPU speed
– 12-Core = 162% CPU speed

If 12 cores are used
– 4-Core = 100% CPU speed
– 6-Core = 146% CPU speed
– 8-Core = 184% CPU speed
– 12-Core = 243% CPU speed

The 8-core will perform similarly to the 4- or 6-core offerings. There might be cases where the 6-core is faster than the 8-core but quite seldom.

The 12-core will only be fastest if 10 or more cores are fully saturated. This is quite uncommon in most tasks but for example video encoding or 3D-rendering can, with relative ease, saturate the machine I think.

Price / performance I think the 6-core probably beats the 8-core but on the 8-core you have similar performance in most tasks and decent multithreaded tasks should be upwards 25% faster.

[Erik Lindahl]

We went with an 8-core system. Better bang for the buck, in theory faster in many situations. At single threaded apps the 8-core will be much faster, where if 11 or more core are used in the 12-core machine it should have an edge but below that probably not. Primarily that means 3D and possibly video transcoding, but most tasks don’t scale that well sadly.

If we need more power we’ll add a secondary 6- or 8-core system. Being able to completely offload work has its benefits.

For certain tasks the 12-core will rule. RED RAW on the CPU for example. But RED will have proper OpenCL acceleration in their SDK soon. My guess in NAB.