Um ... not at all clear why you wish to avoid proxies ... that seems well, silly to me. Unless you're doing a lot of time-ramping at least.
The media you're talking about is very long-GOP and highly compressed, which rags the crud out of a CPU for playback. Basic players can get by because they don't place any other load whatever on the computer subsystems ... while an NLE is by nature itself a heavy demanding app for subsystems. Running the NLE through the CPU/cores/threads/RAM, and then throwing highly compressed 4k long-GOP media at it, is by definition about the worst "stress" you can put on a system.
Cineform proxies or DNxHD proxies are quick & easy on the system, playback's a breeze, and you never have to worry about the proxy/original media for nearly anything but time-ramping or merged files. For that, it's best to transcode the original media to Cineform or DNxHR so the playback is vastly improved, and also use perhaps a Cineform preview format for the sequence so you can render sections for best playback.
Other than that, check on Bill Gehrke's PPBM8 page, he's got one of the Ryzens as a very high-ranked CPU, certainly on a bang-for-buck rating by far the best.
And he's also tested the EVGA 1060/6GB GPU, and found that two of those blow away a single 1080 with max vRAM. I've got the same 1060, and it's been a huge upgrade for my i7/6-core machine. Playback & renders (when involving GPU accelerated effects) are much nicer, "red" sections of sequences all gone to yellow, it's been a big help.
I don't know the comparison of the 1060 vs the 1050ti though.
Thanks for the response Neil. I'm aware of the high compression of h.264 and the fact it's very taxing on hardware. I produce long event videos such as weddings, and do not care to wait hours to render proxy files if there is an affordable configuration to avoid it. I think the idea of having to care for such a thing, as well as having two versions of the same 100+ video clips is an inefficiency in it's own respect. To me, having to create a proxy is a bandage to help with the issue of either insufficient hardware, or inefficient software. All philosophy aside, all I'm looking for is a simplified drag-drop-edit workflow, no need for proxy rendering, and I don't care if it causes my computer to have 1/2 of it's lifespan really. That's all I'm looking for an answer to. Thanks again for your input!
1 person found this helpful
Double the RAM.
1060 instead of the 1050i.
I do some event work myself, and even get some projects from ... 4k phone media. Oh ... joy. So I do understand the wish to be able to just sit & edit.
When I come back from such a job with my regular cameras, I upload the media to the computer, dropping it into a watch-folder for AME and go home for the night. Next morning, it's all in (typically) Cineform ready for me to ingest and create proxies if needed.
For the 4k phone media, I've gotten very familiar with using Handbrake's queue batching capabilities, and also the watch-folder setups in Media Encoder, so that I can set a folder or more of 4k media to go through Handbrake turning it into 4k CFR at 48Mbps, exported to an AME watch-folder where it gets transcoded to Cineform YUV 4k media. All done overnight.
So I do that before leaving at night, and next morning, my transcoded media is ready for ingesting into PrPro. After the project's completed, I dump the Cineform transcodes as they are large files and I don't need them as I can always make more from the "original" files I created in Handbrake. Or ... do the whole thing from the phone media files at need. Doesn't take any of my time, I'm sleeping.
And as Peru Bob says, you can try the hardware changes. I've just installed an EVGA 1060/6GB replacing the 970GTX I had, and it's doing a much better job on sequences with Lumetri or other GPU accelerated effects present. That might get you up to acceptable speed.
Might help the proxies get done quicker too ...
The main thing is just to get to a mix of hardware & workflow to spend the least time messing with stuff.
I think a computer upgrade is well worth it to avoid the proxy workflow. You'll need to double your ram and most likely you'll need to update your gpu as well. I have an i7-5820k cpu, a GTX 970 GPU, 32 gigs of ram and a Thunderbolt3 raid which allows me full res 4k playback without issue.
You should consider removing h264 from your workflow. While you're getting a great image with your camera, you're not getting a great format.
For the kind of performance that you're looking for, you want to be working with a mezzanine codec.
Try adding this to your workflow to avoid h264:
I've strongly considered it, but for the type of event work I do, capturing in-camera at h.264 is what works best. It would be great to just record Prores like that though. But my delivery is in h.264 anyway, so it's going to get compressed either way, right?
The difference on capturing via the Atomos & then editing that in PrPro versus H.264 ... you'll get a better image into PrPro that will pass through the system for editing much better than any H.264. So, you'll have a better image with more latitude for corrections to tonality & chroma plus better performance editing ... to turn into the sausage that is H.264. Your exports could easily be faster also ... it's that much improvement in the processing chain.
So ... your final sausage may be a nicer sausage, and your process will be nicer for you.
As a delivery format H.264 can be a pretty decent sausage, very usable. As a capture it's fast & aggressive for purpose-designed chips to encode for writing to card. For anything else, it's ... sausage. And note, I do love a good sausage myself ... so that isn't a slight, just noting that a sausage is rather more processed than meat cut into a steak.