This is something I couldn't believe at first, but checked it enough times to know it's a hard fact. Please take a look at my systems specs,; you can see that relatively speaking, both my editing platforms have slow CPUs and "not the worst" GPUs (according to the consensus in this forum, and my own belief so far). So I tried Oldsmoke's 2 short XAVC S 100 Mbps 4K clips from the AX100 he posted, as well as their XAVC, 10 bit 4:2:2 mxf counterparts (created using Catalyst Browse). I observed the consumer "S" version barely touching 23.937 fps on my i7 system with what is considered "the last really working CUDA" card - my Fermi GTX 580; the same system only achieved some 15 fps with the mxfs... Well, I thought - time for a new system :( But then - out of curiosity - I turned my GPU card of in video preferences (I never did that before with my 1080p material as I was totally satisfied with my editing experience on this PC, and having believed the full fps I was getting was in large part the result of my GTX 580 card acceleration)...
Guys, was I shocked! Without any help (or should I say - interference) from the GPU, I got rock solid 24p at Best/Full with the original "S" clips (even with 3 FXs), and almost the same fps (only sometimes slower) with their mxf counterparts!
So, I tried the same with my System #2 (the same generation Sandy Bridge, 4 years old laptop) - and.... almost the same results after switching my Quadro M4000 off! Of course this was a bit slower - especially the full XAVC mxfs only played at some 20-21 fps - but this is just a laptop after all, and the CPU speed is rated 4x 2.7 GHz vs 4x 4.4 GHz on my slightly OC'ed desktop!
So really - I'm not sure about other formats, as with 1080p/25fps I mainly used for years I didn't complain - but with 4K XAVC (S) at 24p, both my CPUs are enough for quite convenient editing a simple project (no MC of course), while the mere presence of what I thought were competent GPU accelerators is slowing it all down to a crawl....
Now my questions is: where is all that power going, and why does it slow down pure-CPU performance so badly? After all, when the GPUs are used, the GPU-Z utility shows solid 80% of GPU load!
This makes me think some magic toggle within internal settings must exists, as this test defies all logic :(
Piotr
Guys, was I shocked! Without any help (or should I say - interference) from the GPU, I got rock solid 24p at Best/Full with the original "S" clips (even with 3 FXs), and almost the same fps (only sometimes slower) with their mxf counterparts!
So, I tried the same with my System #2 (the same generation Sandy Bridge, 4 years old laptop) - and.... almost the same results after switching my Quadro M4000 off! Of course this was a bit slower - especially the full XAVC mxfs only played at some 20-21 fps - but this is just a laptop after all, and the CPU speed is rated 4x 2.7 GHz vs 4x 4.4 GHz on my slightly OC'ed desktop!
So really - I'm not sure about other formats, as with 1080p/25fps I mainly used for years I didn't complain - but with 4K XAVC (S) at 24p, both my CPUs are enough for quite convenient editing a simple project (no MC of course), while the mere presence of what I thought were competent GPU accelerators is slowing it all down to a crawl....
Now my questions is: where is all that power going, and why does it slow down pure-CPU performance so badly? After all, when the GPUs are used, the GPU-Z utility shows solid 80% of GPU load!
This makes me think some magic toggle within internal settings must exists, as this test defies all logic :(
Piotr