Intel i9 9900K processor review using Vegas Pro

Comments

Peter_P wrote on 11/18/2018, 2:37 AM

I did a quick HEVC render using nvenc on both VPX and Vegas with the same settings and source clip (4K 25P into HD 25P), render speed was the same by the way.

@bitman

What about preview smoothness of UHDp50 footage ?

Former user wrote on 11/18/2018, 4:25 AM

Does anyone know for sure that Vegas pro is using AVX instruction or how this can be detected with the running program?

AVX2 is said to be 15-20% faster on intel cpu compared to AMD cpu, when encoding h.265. This same performance advantage is not seen with h.264. So if you had the resources, you'd render the same project without filters/plugins (as filters increase statistical noise as those cpu cycles won't use AVX2) to both h.264 & HEVC on an intel cpu, and repeat process on AMD cpu. If vegas is using AVX2 you should see a greater performance gain only on HEVC render with the intel CPU.

SonyVegas13 did not use AVX2 so nothing has probably changed?

Bencharking of cpu's using handbrake (uses AVX2) to render h.265 file always puts AMD in a bad light.

h264 encode

h.265 encode

Peter_P wrote on 11/18/2018, 7:20 AM

Thanks, where are these chart from? Would be interesting how much power was drawn in the 'unlimited' tests.

I normally use the Intel HEVC with QSV. This is rendering close to realtime UHDp30 to UHDp30 HEVC.

Former user wrote on 11/18/2018, 8:28 AM

They were video captures, but the full review is at https://www.techspot.com/review/1744-core-i9-9900k-round-two/

I know handbrake will do AVX512 instructions, and that will use a huge amount of power and reduce your clock speed, but I don't know the technicalities of when it uses it or not. For desktop single cpu use, popular wisdom with typical 16gb(or even 32gb?) ram was that you did not use avx512 instructions as it was not beneficial.

I believe you can see from the 95watt h.265 render that is performed better than the h.264 render in relation to amd cpu, so seems more efficient.

Peter_P wrote on 11/18/2018, 8:29 AM
 

idle power

 

@bitman

Your idle power consumption seems to be very high.

To make it comparable I have taken this original UHDp24 'runner' clip and put it 3x onto the Vp15 time line. First I run the preview of this area and following I renderet it to UHDp24 Intel HEVC 40Mbps.

Here you can see the CPU is drawing about 15W in idle mode, 35W in Best(Full) UHDp24 internal preview and ~75W while rendering  to UHDp24 Intel HEVC 40Mbps.

How does this look with the MAGIX VPX on your i7-8700k system ?

 

Peter_P wrote on 11/18/2018, 8:36 AM

They were video captures, but the full review is at https://www.techspot.com/review/1744-core-i9-9900k-round-two/

Thanks, a very interesting test. However, taking about the i9-9900k is not only a matter of CPU load. For preview and HEVC QSV rendering the iGPU takes  a big part of the workload. I’m not sure this is represented in this test.

 

bitman wrote on 11/18/2018, 2:48 PM

@Peter_P

The tool I used to measure Power is corsair link 4, it interfaces with my corsair power supply via an internal USB. Although I presume this measurements are quite accurate, it unfortunately is measuring the complete PC power draw, with power in and power out, so it can also calculate the PS efficiency. "Unfortunately" in the sense this does not separate the CPU power draw from the rest. I suppose this also explains my high powerdraw in idle as all PC components and case fans are included.

APPS: VIDEO: VP 365 (22 build 93, 21 - build 315), VP 365 20, VP 19 post (latest build -651), (uninstalled VP 12,13,14,15,16 Suite,17, VP18 post), Vegasaur, a lot of NEWBLUE plugins, Mercalli 6.0, Respeedr, Vasco Da Gamma 16 HDpro XXL, Boris Continuum 2024, Davinci Resolve Studio 18, SOUND: RX 10 advanced Audio Editor, Sound Forge Pro 17, Spectral Layers Pro 10, Audacity, FOTO: Zoner, DXO, Luminar, Topaz...

  • OS: Windows 11 Pro 64, version 23H2
  • CPU: i9-13900K (upgraded my former CPU i9-12900K), Air Cooler: Noctua NH-D15s
  • RAM: DDR5 Corsair 64GB (5600-40 Vengeance)
  • Graphics card: ASUS GeForce RTX 3090 TUF OC GAMING (24GB) 
  • Monitor: LG 38 inch ultra-wide (21x9) - Resolution: 3840x1600
  • C-drive: Corsair MP600 PRO XT NVMe SSD 4TB (PCIe Gen. 4)
  • Video drives: Samsung NVMe SSD 2TB (980 pro and 970 EVO plus) each 2TB
  • Mass Data storage & Backup: WD gold 6TB + WD Yellow 4TB
  • MOBO: Gigabyte Z690 AORUS MASTER
  • PSU: Corsair HX1500i, Case: Fractal Design Define 7 (PCGH edition)
  • Misc.: Logitech G915, Evoluent Vertical Mouse, shuttlePROv2

 

 

Former user wrote on 11/18/2018, 9:56 PM

For preview and HEVC QSV rendering the iGPU takes  a big part of the workload. I’m not sure this is represented in this test.

Yeah good point, It's IGPU uses a maximum of 15w. The handbrake test should have been software encoding & the 'use quicksync for decoding' option should have been turned off, as the benchmarks would not have been valid due to the AMD's lack of QSV.

 

MikeyDH wrote on 11/19/2018, 1:44 AM

I just loaded a 5 minute clip to render in Magix AVC at 1080P 29.97 Thee are only cross fade transitions and one swish pan. One line No titles or extreme FX. Why should this take 1 hour and twenty minutes to render? Vegas 16 build 307 using 11 1080 AVCHD files

That is a very old processor. You're going to need to upgrade that a bit if you want better rendering times with 1080p media.

You should invest in a good GPU if you really want to speed things up, though when pairing it with a CPU that old, it might be best to just invest in a new system. Many of the Intel CPUs come with built in hardware acceleration for encoding that Vegas supports, that's something to consider.

If you upgrade, I'd recommend at least 16GB RAM as well.

Yes, I realize there is a need for an upgrade to the processor. It just seems with every software upgrade that you need a hardware upgrade of one thing or another. Might as well just go in debt for a workstation dedicated to video only. I put a lot into this machine over the years and it handles most things pretty well.

I believe there is 16 gigs of ram in it already so apparently there might be another issue

 

john_dennis wrote on 11/19/2018, 2:19 AM

"I believe there is 16 gigs of ram in it already so apparently there might be another issue"

?????????????????????????

MikeyDH wrote on 11/19/2018, 3:20 AM

Thanks, I realized that after I posted. I have to pay a little more attention to what the heck I'm looking at. looks like a trip to the memory store.

Former user wrote on 11/21/2018, 3:31 PM

Some “Red car” benchmark tests with this cpu ...

Oveclocking in brackets (Overclocked). Defaults for all render templates, 24/12 etc. see below. All output is FHD.

[ non OC 4790k values]

 

HW Acceleration = Nvidia

Render with Nvenc .......  0:21s …..... [4790K ... 0:31s] .... (19s) … Hevc is 0:23s {(20s)} [444 (19s)]

Render with QSV ..........  0:18s ……………………………. (15s) … [444 (44s)]

Render with Cpu only ...  0:59s ......………………………... (53s .. 70 deg) .. [444 (53s)] ...[4790K ... 2:11s]

 

HW Acceleration = Intel Graphics

Render with Nvenc … N/A ….......  (34s)

Render with QSV … N/A ..........  (48s)

Render with Cpu only … N/A ….  (54s .. peak cpu temp 67 degrees)

 

HW Acceleration = None

Render with Nvenc .......  1:10 ………………………………… (1:05s)[444 (1:05s)]

Render with QSV ..........  1:05 ………………………………… (0:58s) [444 (1:18ss)]

Render with Cpu only ...  1:33s ......…………………………… (1:35s .. 71 deg) .. [444 (1:31s)]..[4790K ... 3:08s]

Unfortunately no results for the middle Intel QSV section above, when I attempted to render, the time counter just kept moving up, no rendering, each time I had to do a forced PC shutdown.

Update … The new Intel graphics driver 444 fixed the middle section issue.  I've added in values post Intel driver update in brackets [444 ].  Note that the render times are now longer for QSV renders, more than double where rendering using QSV and Nvenc is HW Acc.  Non QSV render types are the same or slightly quicker.

N/A …  Simply no tests done here for non Oveclock

 

 

Former user wrote on 11/21/2018, 4:13 PM

dle Intel QSV section above, when I attempted to render, the time counter just kept moving up, no rendering, each time I had to do a forced PC shutdown.

So vegas16(15?) not compatible with Intel UHD Graphics 630 when used for graphics acceleration?

 

 

Former user wrote on 11/21/2018, 5:34 PM

I used Intel 100.6373 and Nvidia 4.1694 driver versions, also tested on VP15, it went a bit further than VP16 then froze the render, had to force close. Referring to the “middle” test section only.

Peter_P wrote on 11/22/2018, 1:17 AM

Some “Red car” benchmark tests with this cpu ...

No overclocking on cpu or memory, just defaults. Defaults also for all render templates, 24/12 etc.

@Former user

What encoder did you use for that test ?

And could you please use AIDA64 - Tools/Systemstability to monitor the CPU Package (W) ?

Would be very nice if you could also get us the CPU package (W) when rendering this UHDp24 sample to MAGIX HEVC with QSV to 2160p.

Former user wrote on 11/22/2018, 4:58 AM

There are two 4K samples at that link, the original and other, which one or does it matter?

I'm sure I can probably find AIDA64 online but what is the cpu package (W)?

What encoder did I use. Magix avc, using any of QSV, NVENC or Cpu.

Peter_P wrote on 11/22/2018, 5:44 AM

There are two 4K samples at that link, the original and other, which one or does it matter?

You should use the original file which is an XAVC-S 100Mbps.

 

I'm sure I can probably find AIDA64 online but what is the cpu package (W)?

With the mentioned tool of AIDA64 Extreme, you can monitor not only tempertures of CPU cores but also the power that the CPU packge is drawing. Use the 'Unified' tab and select the items in the preference. Refresh should be set to 1s.

P.S. PLS use the portable *.zip file. that does not need to be installed.

 

What encoder did I use. Magix avc, using any of QSV, NVENC or Cpu.

So it is the 1080p : MAGIX AVC/AAC MP4 – Internet 4k 2160p 29.97 fps (Intel QSV)

This takes with Vp15 B416 23s / 62W on my i7-8700k.

 

Please use for the '3xRunner' the Vp15 'MAGIX HEVC/AAC MP4 – Internet 4k 2160p 23.97 fps (Intel QSV)' template.

Thanks for your help to find out if an upgrade to the i9-9900k could make sense for me.

 

Former user wrote on 11/22/2018, 6:16 AM

 

"So it is the 1080p : MAGIX AVC/AAC MP4 – Internet 4k 2160p 29.97 fps (Intel QSV)”

Don't understand any of this.  Its a FHD project, not 4k.

 

I’ve posted a Red Car Hevc time, 23s, plus render templates used in my main post above. I’ve also included, in brackets, a few comparison i7 4790K values.

The main big improvement you may get is in CPU render, not Nvenc or QSV, thats to be expected, I got about half the render time improvement compared to the 4790K times. So really worthwhile if you render for best quality using CPU with HW Acc.  Not saying theres anything wrong with Nvenc/QSV rendering, just probably don't upgrade if you are normally rendering using Nvenc or QSV.

Peter_P wrote on 11/22/2018, 6:32 AM

"So it is the 1080p : MAGIX AVC/AAC MP4 – Internet 4k 2160p 29.97 fps (Intel QSV)

Sorry my fault. I did a C&P from the wrong line in my notes

This is the FHD render template for the Red Car test I used with Vp15 B416:

 

 

The main big improvement you may get is in CPU render, not Nvenc or QSV, I got about half the render time improvement compared to the 4790K times.

But I'm comparing to an i7-8700k with 6 cores and newer iGPU. My most projects are UHDp30 so I render to UHDp30 with HEVC and QSV. Quality is perfect.

 

 

Former user wrote on 11/22/2018, 6:54 AM

Hi Peter, surely the best way of getting an idea of the difference between what you have and the i9 would be for you to do the same red car tests that I did, with your existing setup. Then compare with my i9 results, granted your gpu etc may be different to mine, but it will still get you most of the way there.

I also shoot nearly exclusively in uhd, if you want to also post your i7 times for the sample mentioned earlier I’ll then download it and post times, as mentioned above I can only render 6 of the possible 9 options.

I doubt if its going to be worthwhile going from your 6 core to the 8 core.

Quality is perfect”

I agree, the most, if any improvement will be 1 or 2 seconds using Nvenc or QSV renders, maybe none if you have a high end Ti?

Peter_P wrote on 11/22/2018, 7:23 AM
if you want to also post your i7 times for the sample mentioned earlier I’ll then download it and post times, as mentioned above I can only render 6 of the possible 9 options.

It would be fine for me, just to get the MAGIX HEVC 2160p 40Mbps QSV render time, because this is the output I mostly use. Since there are CPU tests with high temperatures based on high power consumption on the i9-9900k, it would really by good to know if there is any problem, when rendering with Vegas pro.

I only rendered the Red Car to 1080p with MAGIX AVC and it took 23s while your I9-9900k was done in 18s.

 

Peter_P wrote on 11/22/2018, 8:50 AM
Red Car 1080p

HW Acceleration = Nvidia

Render with Cpu only ...  0:59s ......... [4790K ... 2:11s]

HW Acceleration = AMD R7 on i7-8700k Vp15 B416

Render with Cpu only ...  0:59s ......... [4790K ... 2:11s] /  [8700k … 1:15 / 88W]
So the i7-8700k takes 27% longer with CPU only 1080p30 AVC rendering.

Here the power consuption (~88W) and CPU ( 55°C) / core #1 (~66°C)

 

Peter_P wrote on 11/22/2018, 11:23 AM
 

Unfortunately no results for the middle Intel QSV section above, when I attempted to render, the time counter just kept moving up, no rendering, each time I had to do a forced PC shutdown.

@Former user

On my i7-8700k system I can not select the Intel iGPU for video acceleration.

Not in the latest Vp16 nor Vp15. So this rendering problem does not show up here. Probably Vegas pro not jet detects the i9-9900k/GPU properly in this aspect.

 

 

Former user wrote on 11/22/2018, 1:13 PM

@Peter_P “It would be fine for me, just to get the MAGIX HEVC 2160p 40Mbps QSV render time, because this is the output I mostly use.”

Hi Peter, if you can post your results of the above with a screen shot of the render template used i’ll follow suit.