4K Video Editing PC Build

[EEVblog]

At performance / price paid for mb+cpu , AMD wins.  At performance / electricity used, amd probably wins or is neck and neck, depending on encoding settings.

You'll get no argument from me, I agree.

[wraper]

You would save about 150 on the CPU, 50 on the cooler and perhaps another 50 on the motherboard or so. Not better performance *and* price, but price gains for sure.

Could save more than $100 on motherboard if go for B350 chipset.

[EEVblog]

The Ryzen 2700x has roughly the same performance as your CPU. Some wins, some losses as can be expected.

And both the 2700X and 8700X have only 16 PCI lanes, somethign which people have taken me to task about in the 7820X choice that has 28 lanes.

Asking us to beat *both* price and performance is just dirty fighting IMO 

That's the point, there is nothing wrong with my choice apart from it's not the best overall bang-per-buck.

[Ice-Tea]

You would save about 150 on the CPU, 50 on the cooler and perhaps another 50 on the motherboard or so. Not better performance *and* price, but price gains for sure.

Could save more than $100 on motherboard if go for B350 chipset.

My appologies. When I looked up the price of his Mobo apparently the first hit was a used board. So, yeah, 100 if not 150$.

[wraper]

The Ryzen 2700x has roughly the same performance as your CPU. Some wins, some losses as can be expected.

And both the 2700X and 8700X have only 16 PCI lanes, somethign which people have taken me to task about in the 7820X choice that has 24 lanes.

Ryzen has 16 PCI-E 3.0 lanes available from CPU to directly connect GPU, 4 PCI-E 3.0 lanes for NVMe SSD + additional lanes from chipset.

[wraper]

BTW threadripper has 64 PCI-E lanes if that's what you are after  .

[AmericanLocomotive]

And both the 2700X and 8700X have only 16 PCI lanes, somethign which people have taken me to task about in the 7820X choice that has 28 lanes.

No one "took you to task" for choosing a processor that only has 28 lanes. We critiqued the choice of buying a processor that only has 28 lanes, on a platform designed for a processor with over 40 lanes. Putting that CPU in an X299 board gimps a bunch of the PCI-E slot - which is one of the reasons why you'd buy the platform in the first place.

That's the point, there is nothing wrong with my choice apart from it's not the best overall bang-per-buck.

If saving close to $400 is not worth doing 30 minutes of basic research, then so be it. Your time is clearly far more valuable than mine.

[Monkeh]

I'll let you in on another Youtuber secret and why a Youtuber might respond to question or comment in what might seem like a harsh way;
When you ask a question, you are often not aware that a dozen or a hundred other people might have asked the same question, or made the same comment. So it's often not the first time the Youtuber has seen that question or comment, it could be the 10th time, or even the 100th time. That can be kinda frustrating sometimes.

Not a secret. Yes, that's right, I'm not a big Youtuber but I do have a clue.

So you just pulled the most expensive i7 you could see out of a hat and went with it.
Fine. Okay. 10 minutes research would've saved you some money - and four pages of this.

I could have saved money, at the expense of performance. Please show me an overall more powerful CPU than the 7820X for the same price.

You've already admitted your main workflow is single threaded and you're comparing multithreaded performance - and the improvements are mostly marginal anyway.

You don't do 4K HEVC encoding on the CPU, you don't have fast enough storage or networking to take advantage of the faster compression or encryption.. and with the CPU you're really limited in adding both.

The Ryzen 2700x has roughly the same performance as your CPU. Some wins, some losses as can be expected.

And both the 2700X and 8700X have only 16 PCI lanes, somethign which people have taken me to task about in the 7820X choice that has 28 lanes.

Because you've used a crippled CPU which limits the platform, not because you need the lanes. The entire purpose of that platform is to allow very high multithreaded performance and/or I/O capability - neither of which you need or use.

I was a bit disappointed by the power supply though. It works, yes, but I'm sure you'd point and laugh if a power supply like that appeared in a mailbag.

I would have no clue at all unless I opened it and inspected it. It's that horrible? Like so bad it's going to fail in a year when fairly lightly loaded?

No, it's just not the class of power supply you'd expect for a machine like this. It's the type of PSU I'd compromise on for a $500 all-in build, or keep as a spare, not what I'd put in a working machine with a CPU which costs more than the average PC.

[EEVblog]

New video coming on the main channel, with commentary on my decisions and reasoning so that people (will hopefully) stop asking.
Ok, they won't stop  , but it should be less volume.

I'm done discussing this, I want to get on with other things.

[Howardlong]

In my comparatively limited video editing experience, recently mostly with PowerDirector 16 at 4k50p, performance depends on:

(a) Software (different s/w, and different versions, support different h/w better than others).
(b) Hardware (hardware my support encode/decode, but limited codecs, resolutions, hybrid?),
(c) Workflow (effects-heavy workflow such as transitions, PIP, text etc is very different to assembling simple cuts).

My Youtube videos are almost all simple 4k50p cuts re-assembled (i.e. transcoding rather than rendering). On PD 16, use of a Kaby Lake Quick Sync is the fastest method to produce by some distance, being about 1.2x to 1.3x faster than real time on either an i7-8700k or i7 7700k.

NVENC on a GTX1070 or 1080 I found was about the same speed, being about 80% of real time on PD 16 for the same production.

Similarly, running a 24C/48T dual Xeon E5-2696V2 software encode on PD 16 was roughly the same speed on the NVENC on GTX10xx hardware. I found that a Ryzen 1800X software encode was of similar speed.

So, in PD 16, for 4k50p, simple cut re-assembly with no effects, iGPU Quick Sync on a Kaby Lake processor with iGPU came out about 1.5x faster than software or NVENC rendering.

I'm not aware of any socket 2011 v1 to v4  processors supporting Quick Sync as it's an iGPU function, and as far as I know no socket 2011 processors have an iGPU.

I also find that NLE on PD 16 is a very, very much better experience on a machine with an iGPU as it doens't necessarily need to transcode content to a lower resolution for editing.

[YouTube recommends h.264 at between about 53 and 68Mbps variable bit rate in 4:2:0 for 4k50p, and so I use a pre-baked PD 16 AVC preset that presents at 50Mbps, close enough, and upload the produced video without further transcoding. The source files are mostly at a significantly higher bitrate, 150Mbps 4:2:0, but at the same resolution and frame rate.]

[EEVblog]

So, in PD 16, for 4k50p, simple cut re-assembly with no effects, iGPU Quick Sync on a Kaby Lake processor with iGPU came out about 1.5x faster than software or NVENC rendering.

I have got PowerDirector and it's reported to be quite fast rendering, but haven't tried it on the new machine. IIRC is was quite a bit faster than Vegas. First impression usability wasn't that great, but better than other editors I've tried. No plans to switch editors though, just from a productivity point of view, but I'm always curious.

I also find that NLE on PD 16 is a very, very much better experience on a machine with an iGPU as it doens't necessarily need to transcode content to a lower resolution for editing.

You don't want to have the extra step to transcode to a smaller size just for editing, at least not if you are doing it every day.

[Fungus]

So, in PD 16, for 4k50p, simple cut re-assembly with no effects, iGPU Quick Sync on a Kaby Lake processor with iGPU came out about 1.5x faster than software or NVENC rendering.

I have got PowerDirector and it's reported to be quite fast rendering, but haven't tried it on the new machine. IIRC is was quite a bit faster than Vegas. First impression usability wasn't that great, but better than other editors I've tried. No plans to switch editors though, just from a productivity point of view, but I'm always curious.

I've used PowerDirector and it is fast at rendering. It uses the Intel Quick Sync video compression instructions and should be very fast on your new 16-thread CPU.

Usability is OK for splicing bits of video together. Some functions are quite well hidden though, be prepared to use google search until you figure out where they are.

[EEVblog]

I've used PowerDirector and it is fast at rendering. It uses the Intel Quick Sync video compression instructions and should be very fast on your new 16-thread CPU.

Oh, it looks like SkyLake has quicksync, I didn't think it had it.

[Monkeh]

I've used PowerDirector and it is fast at rendering. It uses the Intel Quick Sync video compression instructions and should be very fast on your new 16-thread CPU.

Oh, it looks like SkyLake has quicksync, I didn't think it had it.

afaik it's part of the GPU, which you don't have.

[Fungus]

afaik it's part of the GPU, which you don't have.

Wikipedia says: Certain low-end and high-end parts (including multi-socket Xeons, and some Extreme Edition CPUs expected to be used with a dedicated GPU) do not contain the hardware core to support Quick Sync.

Because ... servers shouldn't be expected to compress video? 

[rx8pilot]

I have been in the movie business for decades.....that may be the silliest thing yet. Here in Hollywood.... 4k is overkill for most everything. Beyond that, it is only useful for bkg plates and visual effects elements.
YouTube videos? Just plain funny to bother.

Countless people have said it's a big improvement in video quality. Not just in 4K but in 1080p and lower.

The most popular camera in TV and Features is the ARRI Alexa - a 2k camera (the 'k' arguments are best left for another forum altogether)
The reason it has been so remarkably successful when Sony had the F65 (8k) and RED was pushing 4k and 6k, and for the past few years Panasonic has been offering 4k in the Varicam.....the Alexa still outnumbers them all put together. Why - because it looks good to the end user. The colorimetry and the delicate handling of its dynamic range results in a pleasing image that is also easy to work with in post-production.

As far as I know....watching screenings in laser projection at the Academy of Motion Picture Arts and Sciences with ASC cinematographers...I have never once heard a comment that sounding anything like  - "I wish they would have shot this on a higher resolution camera" when speaking of 2k acquisition on the Alexa.

Netflix and Amazon is the primary driver of 4k acquisition since they arbitrarily made it a requirement. To managment....more K's is better. So they pushed a lot of shows into Panasonic, Sony, and RED cameras to the overall detriment of the final product.

No matter how you slice it.....your content is not judged on image resolution. It is first judged on the message or story followed by editing, followed by color and contrast, followed by audio, followed by spacial resolution. I have no issue with high-resolution of course....I just don't think it should slow down or negatively impact any other element. Personally - I would never spend the money when .0001% of the audience can even tell the difference.

[wraper]

afaik it's part of the GPU, which you don't have.

Wikipedia says: Certain low-end and high-end parts (including multi-socket Xeons, and some Extreme Edition CPUs expected to be used with a dedicated GPU) do not contain the hardware core to support Quick Sync.

Because ... servers shouldn't be expected to compress video? 

Because in say 8700K GPU takes around 30% of the die area. Also Quick Sync encodes in crap quality. When it comes for free with iGPU, it's OK. But I doubt you would want Quick Sync instead of a few CPU cores. Moreover, I think most people would want 8700K with 3 additional cores instead of iGPU they likely would never use.

[Muxr]

I just want to say to Dave. Sorry if I came off a bit harsh. I think we all need to take it down a notch.

It's awesome that you did a new PC build, and you really don't need to justify yourself to anyone. I only brought up AMD Threadripper because I think it's an amazing product (price/perf wise) for workloads similar to yours, and since you influence many people's decisions I think it was important to bring it up so that those watching know there are other [potentially better] options out there, that's all.

That being said PC building is about choice, and you can build your computer from any components you want. It really doesn't matter in the end as long as you're happy.

[mtdoc]

No matter how you slice it.....your content is not judged on image resolution. It is first judged on the message or story followed by editing, followed by color and contrast, followed by audio, followed by spacial resolution. I have no issue with high-resolution of course....I just don't think it should slow down or negatively impact any other element. Personally - I would never spend the money when .0001% of the audience can even tell the difference.

Yeah, I don't get it. Who cares if Dave's videos are in 4K?  I don't need to see the pores in his skin or the length of individual nose hairs.  Circuit board details?  Those are best left to high resolution still photos. 1080p (or even 720p!) is more than adequate for any Youtube video IMHO.

[Fungus]

Yeah, I don't get it. Who cares if Dave's videos are in 4K?

Posterity.

In 100 years time people will look back at Dave's 1080p videos and think how low resolution they are.

[rx8pilot]

Yeah, I don't get it. Who cares if Dave's videos are in 4K?

Posterity.

In 100 years time people will look back at Dave's 1080p videos and think how low resolution they are.

So they say.....in reality only a tiny percentage of motion picture content value is locked in its resolution. IMAX is a good example - the most marketable part of the end product IS resolution and the controlled display of it.

I would argue that Dave would lose close to zero subscribers if it was entirely 720p end-to-end with everything else remaining the same. Even cinematography nuts like me would not even notice. I am here for the content which is only marginally enhanced by increased resolution. To take advantage of any 4k content, I have to have a 4k display AND I have to very close to it. That is a big ask for YouTube viewing. For me personally - the majority of my YouTube consumption is on a 1080p monitor and far enough away that it is effectively 720p or 480p.

I get nothing more out of the videos at the higher resolution. It's not Blade Runner.

[Howardlong]

afaik it's part of the GPU, which you don't have.

Wikipedia says: Certain low-end and high-end parts (including multi-socket Xeons, and some Extreme Edition CPUs expected to be used with a dedicated GPU) do not contain the hardware core to support Quick Sync.

Because ... servers shouldn't be expected to compress video? 

Because in say 8700K GPU takes around 30% of the die area. Also Quick Sync encodes in crap quality. When it comes for free with iGPU, it's OK. But I doubt you would want Quick Sync instead of a few CPU cores. Moreover, I think most people would want 8700K with 3 additional cores instead of iGPU they likely would never use.

After about a year of pain trying to settle on my daily driver desktop, I went through many iterations, in an effort to find THE ONE machine where I could do my development work, day to day business, and 4k video editing, in approximate order...

E5-2670v1 x2
I7-6800k
I7-5820k
E5-2696v2 x2
I7-6700k
Ryzen 1800X
Ryzen 1700
I7-7700k
I7-8700k

By quite a way, the best processors for 4k video editing for my workflow were the two Kaby Lakes with iGPUs. They aren’t bad dev boxes either, particularly the i7-8700k with its 6 cores.

In the end, I have ended up with two day to day machines, the dual E5-2696v2 for development and admin chores, and the i7-8700k for video editing.

The Kaby Lake i7-7700k is now in my NAS that doubles up as one of my VM servers, and occasionally I use it remotely to produce videos if the 8700k is busy.

The Skylake i7-6700k just doesn’t seem to have the right iGPU to do h264 4k encoding like the Kaby Lakes chew through on PD 16.

The main down side of the dual Xeon is the boot time, but wow do 24c/48t speed up your dev builds! Its mobo also has a uefi bios that takes months to figure out. The retail i7-8700k on the other hand boots in about 4 seconds and is the most snappy machine I have.

The Ryzens I’ve found really a little disappointing. They are slower to boot compared to their mainstream Intel counterparts, and there’s something about them that simply makes them feel not as snappy as the Intels. Like the Intel socket 2011s, there’s no iGPU in the two Ryzen 7s I use. Both Ryzen builds live in a cupboard switched off if that’s any testament.

Unlike the dual socket 2011 Xeons, the two socket 2011 i7s, 6800k & 5820k, both 6 core, turned out to be a lot of trouble. They are both power hungry which makes a really compact ITX solution pretty much impossible, plus you need an external GPU. For my work, I found no real performance benefit of an external GPU for video editing, but handy for a bit of mining. I did try an ITX build with the i7-6800k, after thinking it was THE ONE, but with just two weeks in the role, it just stopped working, it powered itself off and couldn’t switch it on. I now believe that the VRMs blew the processor as it wouldn’t work in another board I had. Trouble was, in trying to diagnose the problem, I put the 5820k into the mobo that stopped working... and I’m pretty sure it blew that up too. Now neither worked in my spare board anymore. A bloody expensive mistake.

One other thing. Beware of overclockable CPUs and the temptation to overclock! You’ll end up burning days and weeks of your life on it, when you could’ve been productive instead.

For testing, all of these machine builds were populated with at least 32GB RAM in XMP profile on Intel platforms, and used either Samsung 850 EVO SATA-3 or 950 PRO NVME drives, I couldn’t tell any difference but I didn’t time it. I did try a hybrid drive a couple of times, frankly I wouldn’t bother unless it’s all you have and are desperate, they’re very disappointing.

[Howardlong]

Yeah, I don't get it. Who cares if Dave's videos are in 4K?

Posterity.

In 100 years time people will look back at Dave's 1080p videos and think how low resolution they are.

For me the jury’s still out on whether I prefer 1080p50 or 2160p25. The 50fps I find to be more “real”, and for me that 50p has always been a differentiator particularly for technical and factual content.

I realise you have more options in post with 4k like cropping, zooming and panning, but realistically I wonder how often that’s going to be used.

The videos I do nowadays are 4k50p, with mostly 4k50p sources, in an effort to cover all bases. There’s also a school of thought that says for Youtube there’s some benefit to upconverting higher bit rate 1080p to Youtube bit rate 2k or 4k to improve streaming quality. It’s also noticeable to some 50fps vs 60fps when played on a different refresh rate momitor and particulalry if the content is panning. I learned from a forum member this can be mitigated by choosing the right shutter speed/angle, but this limits correcting options for low light conditions often found inside. For the factual content done here, I can’t say I particularly notice it.

A downside of 50p is that you need more light, or higher ISO, or larger aperture. Some like the bokeh effect of a larger aperture, but frankly it adds little value in the kind of factual content we’re talking about, and is detrimental when trying to maintain a large depth of field on a close up.

[Dubbie]

In our studio of about 20 people, I think I am the only person who prefers 50/60fps content.
I can't stand the juddery artifacts on 24p. It makes me think that it might be a difference in perception. Not just a preference.

Regarding what RX8Pilot said about Arri Alexa, Another big attraction for us is how well everything is documented. As far as we are concerned, there is no "secret sauce" or other proprietary nonsense when it comes to their colorimetry etc. They publish all their matrices and transfer curves etc, which makes them a dream to work with from a VFX point of view. RED and Sony aren't so forthcoming and it makes it not so nice to work with.

[rx8pilot]

In our studio of about 20 people, I think I am the only person who prefers 50/60fps content.
I can't stand the juddery artifacts on 24p. It makes me think that it might be a difference in perception. Not just a preference.

It seems like it is a generational thing (not exclusively, but generally)

24fps was chosen as a compromise between reasonable perception and practical reality. It makes a surreal motion where your brain is required to fill in the gaps.
25 and 30 fps were defined at the early stages of TV, but then we got interlaced as yet another technology inspired compromise. That gave us far more real motion that we perceived as 50fps and 60fps, albeit at half resolution.

Now that we are in the mega-bit-rate-mega-pixel world.....60fps at ridiculous resolution is totally possible. It is perceived as hyper-realistic. The surreal dreamy look is traded for a bombarding of visual information. For some content it works well....other stuff it is simply distracting or otherwise ruins the show. For example....I would not even consider seeing a feature film that was shot/displayed at 30-48- or 60fps. It looks like a soap opera on daytime TV.