multi core performance vs single core (vs nMP)

For video editing and conversion (encoding etc) multiple cores are used at once. The new iMacs with greater multicore scores will definitely speed up your render/conversion time. Although this would depend on what your Mac Pro specs are.

 

A 4Ghz retina iMac with m295x GPU is faster than a quad-core nMP at most things, whether single or multi-threaded. It is vastly faster at single-pass transcoding or rendering video to H.264, since the Haswell CPU has Intel Quick Sync: https://en.wikipedia.org/wiki/Intel_Quick_Sync_Video The Xeon CPU in the nMP doesn't have this.

However this depends on whether your software uses Quick Sync; FCP X and Handbrake do, I don't know about other software.

So basically unless you're considering a 6 or 8-core nMP a 4Ghz retina iMac will be faster at most things. It's even faster than the 6-core at some things.

Re 2011 MP single-threaded performance vs a 4Ghz retina iMac, I thought there was a 2010 and 2012 MP, not 2011? But I guess they're pretty similar.

The fastest CPU in the 2010/2012 MP was 3.33Ghz, and it is several generations behind Haswell from instructions per cycle. You'll pick up 20% just from the clock rate increase, and maybe another 20% from IPC improvements. Any software that uses the new AVX vector instructions will be faster; I don't have a list of that.

 

It also depends on the work load. The longer the processor has to work the hotter it gets and it slows down but from what you say you are doing that probably won't be an issue.

 

Well that's a closer call. Video editing and conversion is typically heavily multi-threaded and uses all available cores a fair % of the time. If the nMP has D700 GPUs that would be another nice thing.

If you edit in FCP X, it is very render-oriented and you spend lots of time looking at red render bars in the timeline. That process is multi-threaded, so the more cores the better. FCP is very good at background rendering, so it doesn't slow you down that much, but sometimes you just have to wait.

Premiere is a lot better at using GPU acceleration for common effects but that's with CUDA, meaning nVidia. I've read that CC uses OpenCL effectively so it achieves the same result on ATI GPUs, including dual GPUs on the nMP.

OTOH if you do lots of rendering or transcoding to MPEG-2 or single-pass H.264, the iMac will be much faster (if using FCP X or Handbrake, don't know about Premiere).

Supposedly multi-pass rendering (which cannot use Quick Sync acceleration) is better quality but I really can't see much difference, so I usually use the faster single-pass version.

You could devise a typical workload, put that on a thumb drive and try it out on both retina iMac and nMP.

 

For video editing, in general, multi core performance is more important than single core performance.

If you bought your MP in 2011 (I assume it's the 2010 single CPU version), all you need to do is just spend $300 on eBay, get a W3690, then you will have the same (or a little bit better) multi core performance as the Max out RiMac.

Of course, that's the old technology, no quick sync. However, for 2 pass (or even 3 pass) H264/H265, all you need is just raw power, this $330 chip perform more or less the same as the modern chip.