4K Video Editing PC Build

[Dubbie]

>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.

And this is where I seem to hold an extremely unpopular opinion. I would LOVE it if feature films were 50-60fps. I find low framerate artifacts incredibly distracting and disruptive to the photography and story.

Don't ask me to do vfx at 60fps tho! 

[GeorgeOfTheJungle]

Interlaced 50i and 60i are a disaster for fast moving objects, much worse imo than 30p, 25p or even 24.

[Dubbie]

I hate the look of interlaced video. It's a nasty fudge and I'm glad it has mostly disappeared. At least as far as I am concerned.

[EEVblog]

For me the jury’s still out on whether I prefer 1080p50 or 2160p25.

I shoot and upload in 2160p30 not 25fps

[EEVblog]

To take advantage of any 4k content, I have to have a 4k display AND I have to very close to it.

Actually, no, you get a benefit without even realising it. Uploading in higher bitrate 4K to Youtube allows Youtueb to produce better quality 1080p downsampled content.
You don't have to watch in 4K to get the benefit.

[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.

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

Damn, I think you are right.
The 7820HQ/HK/EQ have it, but the 7820X does not 

[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.

Damn, I think you are right.
The 7820HQ/HK/EQ have it, but the 7820X does not 

And those are Kaby Lake, and mobile units. The 7820X is a Skylake-W Xeon W-2145 with some of the PCIe lanes chopped off.

[wraper]

Damn, I think you are right.
The 7820HQ/HK/EQ have it, but the 7820X does not 

Don't bother with that. It's only good if you are OK with making video files with sub par compression/quality ratio. Also quality wise CPU encoding still is better that GPU, so if you are primarily after quality, CPU encoding on threadripper is the way to go.

[rx8pilot]

To take advantage of any 4k content, I have to have a 4k display AND I have to very close to it.

Actually, no, you get a benefit without even realising it. Uploading in higher bitrate 4K to Youtube allows Youtueb to produce better quality 1080p downsampled content.
You don't have to watch in 4K to get the benefit.

This is true, but small and depends on the entire chain from camera to end viewer. For example, my phone will record 4k video but an 18 year old Sony 1080p HD camera still looks considerably better in almost every respect.

Human vision is VERY sensitive to temporal resolution, interlacing, color, dynamic range compared to resolution. In fact, the resolution is so hard to perceive that Sony's HDCAM tape format sampled 3:1:1 meaning that 1920 pixels were only recorded at 1440 and no one could see the difference outside of VFX processing. Many feature films were recorded on that tape format and only a tiny number of industry professionals can see the difference when compared to full resolution 2k.

VFX are still done in 2k, even on 4k movies because it saves time ($CASH$) and almost no one can see the difference. There is an inside joke that if the viewer is worried about low resolution - you are doing everything wrong.


When the biz was all analog SD - we still shot commercials on 35mm and 16mm film. Most people could quickly point out 16mm over 35mm even when seen on 270i VHS tapes. So I fully understand the concept of oversampling. Just wanted to point out that other elements tend to drastically drown out the benefit of high resolution.

[Dubbie]

One area where high resolution really helps is when Dave shows on-screen material like datasheets, graphs etc. It's nice being able to read them properly.

[EEVblog]

Damn, I think you are right.
The 7820HQ/HK/EQ have it, but the 7820X does not 

Don't bother with that. It's only good if you are OK with making video files with sub par compression/quality ratio.

I've heard that but not tested it myself.

[Razor512]

How about we just say that Dave may not have made the best choice, but he didn't make a bad choice (especially considering the circumstances) ... and move on.

Can anyone actually tell me, ignoring "bang-per-buck", if there is anything fundamentally "wrong" about my build?
Could I have gotten *better* performance (for my needs) for a *lot less* cost?
Is my CPU or motherboard or memory or drive or cooler fundamentally *unfit for my purpose* in some way?
Because I have yet to hear this in any of the comments.

The arguments all seem to be based on "bang-per-buck", with, at least on the forum here, advocating for an essentially lower performance system.

There is nothing fundamentally wrong, The only major bottleneck you will run into is on the side of single threaded performance. It is a fast CPU, but with a modern intel CPU, nvenc really shines when you can get the CPU to around 5GHz. The issue is that with that CPU, it is difficult to keep from throttling when pushing over 1.3V.

The ultimate issue is that nvenc is still not taking full advantage of many CPU cores, and thus gets a larger benefit from higher single threaded performance over overall multithreaded performance.

If using NVMe storage, then having more PCI express lanes is a must to get the full benefit. While the bandwidth to the chipset is decently fast, it is also heavily shared with a wide range of IO, thus you will see noticeable speed drops when trying to use the NVMe SSD while also doing something like running a 10GbE adapter or multiple SATA hard drives. Additional lanes = you not having to share the DMI3.0 as much.

The CPU is a good tradeoff between single threaded performance, and multithreaded performance for video editing, especially if you decide to try davinci resolve for certain post processing.

The power supply should be fine, I had a thermal take branded one with worse components last about 10 years.

[Razor512]

Wanted to also ask, when will we see 8K 60p videos

RED has a decent 8K camera and the video clips are rather small thanks to a rather low data rate of around 270-290MB/s, a mailbag episode would only use up about 800-900GB of storage.

[NiHaoMike]

What about HDR? The camera supports it. It is often said that HDR gives a more obvious improvement in quality than 4K does. (I don't have a HDR display so I won't be able to take full advantage yet.)

[EEVblog]

There is nothing fundamentally wrong, The only major bottleneck you will run into is on the side of single threaded performance. It is a fast CPU, but with a modern intel CPU, nvenc really shines when you can get the CPU to around 5GHz. The issue is that with that CPU, it is difficult to keep from throttling when pushing over 1.3V.
The ultimate issue is that nvenc is still not taking full advantage of many CPU cores, and thus gets a larger benefit from higher single threaded performance over overall multithreaded performance.

NVENC is in the GPU, not the CPU
"NV" stands for NVidia

[Stefan Payne]

I don't need the other PCI-E slots.

Then why go for the LGA2011 Plattform? With that darn inefficient CPU?

And a more expensive PSU would have gained me what exactly?

working protection against overload for instances or that it won't go out of spec when loaded.
Higher quality Components, better efficiency...

You are the Electronics guy. And you bought an old double Forward, Group Regulated PSU when the LLC-Resonant Mode Converters are only around 20 bucks more...


Things I'd get (don't know if to get in Australia:
be quiet Pure Power 10, 400W
be quiet Straight Power 11, 450W
Bitfenix Formula 450W
Bitfenix Whisper M, 450W
Cougar GX-F 550W

Before you complained my processor wasn't fast enough, now you are complaining that I bought too high end a processor. Do make up your mind.

No, we are complaining that you bought a processor that doesn't make any sense or benefit at all.

Its the same with the Ryzen Threadripper 1900x as well. That processor makes little sense for I'd say 95% of all people.

The ONLY Reason to go for entry level High End Desktop Processors is:
a) you need the Memory and populate all 8 Sockets
b) you need the PCIe Lanes (wich is around 64 for Threadripper and only 28 for the Entry 2011 but 44 for the higher end ones)

No other reason to go HEDT with an 8 core CPU exist.

So to be blunt:
A Ryzen 2700x would have had about the same or better performance for half the price.

[Stefan Payne]

You don't have to. TomHardware and JonnyGuru have already done all the legwork. TomsHardware even has a handy ranking guide for their PSUs. You should have made the time, because the PSU is one of the most important parts of a PC. I've seen many computers killed by cheap power supplies that grenaded.

ARGH

The THG List is garbage. As far as I remember its totally outdated and puts an ancient, almost 10 year old, group regulated unit that isn't any better than what Dave has here, just with better caps into Tier one.

And that is the best way to see if a PSU Tier List is good or not. Look for Seasonic S12II-Bronze or the M12ii_Bronze variants with 620W and below.
If they are somewhere at the top, the guy making the list doesn't know what he's doing.

The List at Linus Tech Tipps is much better...

Oh and Dave:

YOU don't have to do any Research.
You have US!

You could have just asked, Link to a shop, set a price and tell people to build something for you.

If you had a Forum that allowed for "Thanks!" Postings, you could have said something like "I take the configuration with the most votes". And be done. Why do the research YOURSELF when you have people here that know their stuff around PC-Hardware in their sleep??

[EEVblog]

Why do the research YOURSELF when you have people here that know their stuff around PC-Hardware in their sleep??

Because they all give me a different answer. Seriously, I've been asking for help in various computer and other fields as a long as I've had an audience and the result is always the same. All the "experts" have different opinions. Be it computer hardware, video editing/encoding, networking, web and server stuff etc. There is almost never a consensus.
I spend just as much time if not more reading, digesting and investigating other peoples opinions.
Your spat with AmericanLocomotive in this very post is classic example.
Oh, and I can't stress this enough, it's fun to do it myself.

[Stefan Payne]

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?

Dave, I linked you a Review with Pictures of the Insides of the Units. With a bit of Capacitor Manufacturers.

It uses JunFu WL and WR Capacitors as well as CapXon GF.

Has _NO_ OCP on any Rail and UVP on +12V is probably also not there.
UVP on +5V is something like 3,5V or so, if I remember correctly.

I haven't seen a decent 8pin Supervisior for ATX PSU. They are either bad or worse.


And there is Voltage Regulation.
Try loading this PSU with  1A max on 5V and everything else on +12V. YOu'd probably get something around 11,6V on 12 and 5,5V on 5V...

[EEVblog]

Dave, I linked you a Review with Pictures of the Insides of the Units. With a bit of Capacitor Manufacturers.

I'm done in this, I'm not going to argue.

[Stefan Payne]

Why do the research YOURSELF when you have people here that know their stuff around PC-Hardware in their sleep??

Because they all give me a different answer. Seriously, I've been asking for help in various computer and other fields as a long as I've had an audience and the result is always the same. All the "experts" have different opinions. Be it computer hardware, video editing/encoding, networking, web and server stuff etc. There is almost never a consensus.
I spend just as much time if not more reading, digesting and investigating other peoples opinions.
Your spat with AmericanLocomotive in this very post is classic example.
Oh, and I can't stress this enough, it's fun to do it myself.

Yes and if you had a Plugin for your Forums that Allows people to thank other for Posting something, you don't need to have that Problem. You just get the one with the most votes/thanks and misuse that feature for other things.


As for your Question for the Problem with the CPU/Plattform:
doesn't support ECC -> corrects 1bit errors and detects 2 bit errors
darn high Power Consumption
https://www.anandtech.com/show/11550/the-intel-skylakex-review-core-i9-7900x-i7-7820x-and-i7-7800x-tested/8
But the Biggest Problem is the Spectre/Meltdown situation.
In short to somewhat fix that, some OS crippled the I/O Performance drastically. Like 50% less than intended or way higher CPU Load when accessing LAN/Drives.

the old Ryzen 7/1700x is at 82,5W and your 7820 is somewhere around 150W.
An i7-8700 would have been around 100W as well (they aren't that great these days).

And about the PSU:
http://www.realhardtechx.com/index_archivos/SILVERSTONE_Essential_series_ET650_B_650W_4.html

The big black ones on the secondary side should be CapXon KF, the green ones with the cross CapXon GF
And the blue and green ones with the thre cuts JunFu WG (blue) and WL (green)


And one thing we missed:
The FAN!

On good quality PSU you have an advanced Sleeve Bearing like those Fancy Fluid Dynamic Types that are also used in good old harddrives for years.
Those should last for forever. And outlast the PSU in some cases as well...

While the one in your PSU will die sooner rather than later.

[Razor512]

There is nothing fundamentally wrong, The only major bottleneck you will run into is on the side of single threaded performance. It is a fast CPU, but with a modern intel CPU, nvenc really shines when you can get the CPU to around 5GHz. The issue is that with that CPU, it is difficult to keep from throttling when pushing over 1.3V.
The ultimate issue is that nvenc is still not taking full advantage of many CPU cores, and thus gets a larger benefit from higher single threaded performance over overall multithreaded performance.

NVENC is in the GPU, not the CPU
"NV" stands for NVidia

The issue is that nvenc does not rely 100% on the GPU, it still has a CPU aspect to the process. If needed, use process explorer to view the threads that it launches. The encoder can only do the heavy lifting as fast as the CPU can feed it the prepared data.

[Monkeh]

Yes and if you had a Plugin for your Forums that Allows people to thank other for Posting something, you don't need to have that Problem. You just get the one with the most votes/thanks and misuse that feature for other things.

.. have you actually used this forum? Like, ever? Looked around?

doesn't support ECC -> corrects 1bit errors and detects 2 bit errors

And Ryzen doesn't have validated ECC.

There is nothing fundamentally wrong, The only major bottleneck you will run into is on the side of single threaded performance. It is a fast CPU, but with a modern intel CPU, nvenc really shines when you can get the CPU to around 5GHz. The issue is that with that CPU, it is difficult to keep from throttling when pushing over 1.3V.
The ultimate issue is that nvenc is still not taking full advantage of many CPU cores, and thus gets a larger benefit from higher single threaded performance over overall multithreaded performance.

NVENC is in the GPU, not the CPU
"NV" stands for NVidia

The issue is that nvenc does not rely 100% on the GPU, it still has a CPU aspect to the process. If needed, use process explorer to view the threads that it launches. The encoder can only do the heavy lifting as fast as the CPU can feed it the prepared data.

Now awaiting reputable benchmarks showing nvenc is held back by a high end CPU when the entire fucking purpose is to offload the CPU for other tasks.

[gnif]

Hard to believe it takes 16 CPUs running at 60%+ just to feed the GPU with data. 

This is often misunderstood, high CPU load doesn't mean the CPU is actually busy with processing anything useful. For example, lets look at the OpenGL call `glFinish` on NVidia hardware. This API asks the driver to finish the current frame, and blocks until it is done, normally games etc do not call this as they don't care about when the frame is done as when it's done the card will just automatically flip buffers and put it on screen, however when rendering or encoding, the CPU needs to know when the "frame" is done to read it back from GPU memory to local system RAM.

Btw, I know that I am using OpenGL as the example here and the application likely uses DX, but under the hood the driver implements these primitives the same. For the sake of simplicity I have used glFinish as an example, but there are other synchronization primitives such as sync fences that operate the same at the driver level.

So lets imagine the render pipeline.

• Application reads a frame off disk into RAM
• Application feeds the frame to the GPU to encode it
• Application waits for the GPU to finish encoding the frame
• Application reads the frame from the GPU RAM to local RAM
• Application writes the encoded frame to disk

For the application to wait the application needs a mechanism to synchronize with the card. I know for a fact that on NVidia hardware the glFinish and glSync calls, while blocking perform what is called a "SpinLock", which is essentially this:

Code: [Select]

while(frameNotReady) {}

Run that code and your core jumps to 100% usage as it is spinning in a tight loop polling for the completed frame. There is no event for this because windows is not a real time operating system (RTOS), even the shortest of sleeps would be too long (the OS scheduler wouldn't wake the process up again soon enough) and performance would suffer. For example, say you're rendering 1080p, the card will complete the frame extremely fast, and the CPU will not spin for long, showing lower CPU usage, however with a 4K frame, the CPU will spin for much longer, doing nothing, but reading high CPU usage.

I hope this makes it clear, that high CPU usage doesn't mean the CPU is actually doing much at all, 60% load seems high because it's just spinning for 99% of the time, likely the other cores that are seemingly not doing much are the ones writing to disk, etc. In reality they are doing more work then the CPUs that read high usage.

Btw, I have noted that AMD hardware behaves differently and seems to have better wait logic, likely using hardware interrupts rather then polling.

[gnif]

There is nothing fundamentally wrong, The only major bottleneck you will run into is on the side of single threaded performance. It is a fast CPU, but with a modern intel CPU, nvenc really shines when you can get the CPU to around 5GHz. The issue is that with that CPU, it is difficult to keep from throttling when pushing over 1.3V.
The ultimate issue is that nvenc is still not taking full advantage of many CPU cores, and thus gets a larger benefit from higher single threaded performance over overall multithreaded performance.

NVENC is in the GPU, not the CPU
"NV" stands for NVidia

The issue is that nvenc does not rely 100% on the GPU, it still has a CPU aspect to the process. If needed, use process explorer to view the threads that it launches. The encoder can only do the heavy lifting as fast as the CPU can feed it the prepared data.

I beg to differ, I have extensive development experience with the NVAPI, NvENC and NvFBC, the CPU spends 90% of it's time waiting on the GPU to complete each frame. A faster clock will just let you spin faster.