Intel i9 9900K processor review using Vegas Pro

Comments

Peter_P wrote on 11/23/2018, 1:13 AM

@Former user

Using this RX10M4 original clip 3 times in the time line I rendered it with this template

in 26s with a peak CPU package of ~71W on my i7-8700k (no OC)

Thanks for your help.

 

Edit/Note:

As I just saw in the media info, Vp15 does not output HEVC as I had expected by the name of the template. It outputs AVC which takes less time.

Peter_P wrote on 11/23/2018, 1:57 AM

BTW rendering these 3 events with Vp16 B 307 takes 34s and CPU Package draws ~74W

In both cases So4 reader is set to FALSE.

Former user wrote on 11/23/2018, 5:59 AM

@Peter_P

Using VP16 b307.

Running Man test, origional ... (3 pieces of it end to end, 37s 12f duration.)

HW Acceleration with either Nvidia, Intel graphics or none.

Render with either Nvenc, Intel QSV or CPU only.  Render templates set at default for Nvenc and QSV.  The Hevc template render values are in brackets, {Hevc}.  The OC values are in brackets, (Overclock).

New Intel Graphics driver 444 in brackets [(444  ) ]

HW Acceleration = Nvidia

Render with Nvenc .......  0:23s ... ({0:29s} ….. 0:20s) .. [(444 0:23s)]

Render with QSV ..........  0:22s ... ({0:34s} ….. 0:18s) … [(444 1:09s)]

Render with Cpu only ...  1:55s ... ({1:04s} … 1:48s) … [(444 1:54s)]

 

HW Acceleration = Intel Graphics

Render with Nvenc ....... N/A …  ({0:30s} … 0:21s) … [(444 0:23s)]

Render with QSV .......... N/A …   ({0:33s} … 0:19s) … [(444 1:08s)]

Render with Cpu only … N/A …   ({1:05s} … 1:51s) … [(444 1:55s)]

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

 

HW Acceleration = None

Render with Nvenc .......  0:23s ... ({0:29s} ….. 0:21s) … [(444 0:23s)]

Render with QSV ..........  0:22s ... ({0:34s} ….. 0:19s) … [(444 1:09s)]

Render with Cpu only ...  1:55s  ... ({1:04s} .. 1:51s)  Max CPU temp about 74 degrees. [(444 1:54s)]

I rendered using 3 variations of each of these 2 templates, (“Profile “ had only one option, default) avc mp4 and also {Hevc}

So Peter, if you upgrade you’ll lose one second in render times, using Hevc, Qsv render, VP16, i.e. your 34s QSV render times vs my {35s} times. Really hard to justify an upgrade🤣.

I used the "Asus Dual Intelligent Processor 5" app to monitor a few peak CPU watts. No OC.

I modified this later, December 1st.

Peter_P wrote on 11/23/2018, 7:31 AM

Using VP16 b307.

Running Man test, origional ... (3 pieces of it end to end, 37s 12f duration.)

H  The Hevc template render values are in brackets, {Hevc}.

HW Acceleration = Nvidia

Render with QSV ..........  0:22s ... {35s}

@Former user

Thank you very much for making these tests.

 

So Peter, if you upgrade you’ll lose one second in render times, using Hevc, Qsv render, VP16, i.e. your 34s QSV render times vs my {35s} times. Really hard to justify an upgrade🤣.

Please don’t tell my wife ... 😉

That is very surprising but shows how much workload is taken by the iGPU (QSV) and this is the same on both CPUs. Since the video (preview) acceleration on my system can not be set to the Intel GPU and you getting render error when set to this option I assume QSV can not accelerate the preview but your additional 2 cores and a little higher clock might help UHDp50/60 preview a bit.

Thanks again, Peter

Peter_P wrote on 11/23/2018, 8:35 AM

@Former user

If you want to really stress your CPU rendering with Vp16 use the MC HEVC option :

In this case the CPU package draws ~100W on my i7-8700k system. This is already a bit over the 95 TDP. Running the Intel Processor Diagnostic tool goes even above 120W for a short while.

bitman wrote on 11/23/2018, 1:15 PM

@Former user and @Peter_P

I did some "running man" tests on my system, with some strange findings and a possible a Vegas bug. Settings as posters. My INTEL iGPU is disabled in BIOS (I prefer nvenc).

Vegas 16, So4 enabled. {Hevc} in brackets. HW Acceleration = Nvidia

  • Render with Nvenc .......  0:31s ... {37s}
  • Render with QSV ..........  {3:03s} Bug? not really QSV (disabled in BIOS), but just CPU???
  • Render with Cpu only ... 

Vegas 15,  {Hevc} in brackets. HW Acceleration = Nvidia

  • Render with Nvenc .......  0:29s … {36s} Vegas 15 is slightly faster than Vegas 16 ???

Vegas 16, now enabled INTEL iGPU in bios, HW Acceleration = Nvidia

  • Render with Nvenc .......  0:26s ... {33s} ??? faster with Nvenc with intel igpu enabled in BIOS !!!
  • Render with QSV ..........  0:23s ... {34s}   
  • Render with Cpu only ...  3:05s (comparable time as "false" QSV when disabled in BIOS)

Conclusion:

  1. Vegas 15 Nvenc render is slightly faster than Vegas 16 (on my system) - at least with Bios disabled iGPU. Update: Rendering with Bios igpu enabled: Vegas 15 render time is similar as Vegas 16...
  2. Bios setting Intel iGPU enabled/disabled messes up your render templates choice if disabled. If disabled you can still choose Intel QSV
  3. Nvenc encoding is faster with iGPU enabled in BIOS. I wonder why?

 

 

 

 

Last changed by bitman on 11/23/2018, 1:58 PM, changed a total of 4 times.

Current system: VP 17 post (build 387), VP16 (suite) build 424, (uninstalled VP 13,14,15), Vegasaur, Magix Video Pro X (VPX11), Corel VS ultimate 2019, a lot of NEWBLUE plugins, Titler Pro 6, Mercalli 4.0, Respeedr, Vasco Da Gamma 12, VASST stuff, Production Assistent pro3, Boris Continuum 2020, Davinci Resolve Studio 16,...

  • OS: Windows 10 Pro 64, version 1909
  • CPU: i9900K stepping R0 (since October 2019), previously, der8auer i7-8700K (advanced edition), default speed (no overclock), Cooler: Noctua NH-D15s
  • RAM: G.Skill Trident Z 3200C14 DDR4 64GB, XMP set to profile 1 in BIOS
  • Videocard: NVIDEA RTX 2080Ti (Founders edition), NVIDEA studio drivers
  • Monitor: LG 38 inch ultra-wide (21x9) - Resolution: 3840x1600
  • C-drive (games & APPS): Samsung NVMe SSD 2TB 960 pro

  • Current Video source work drive: Samsung NVMe SSD 2T 970 EVO plus

  • Mass Data storage & Backup: WD gold 6TB + WD Yellow 4TB

  • MOBO: Gigabyte Z370 Aorus Gaming 7, BIOS F14
  • PS: Corsair HX1200i, Case: Silverstone fortress 2,
  • Misc: Logitech G910, Evoluent Vertical Mouse, shuttlePROv2

 

Former user wrote on 11/23/2018, 2:52 PM

@bitman 

One of the problems using this benchmark is that it doesn’t show the benefits of a better gpu, the Red car test would differentiate the hardware out better.

As mentioned earlier by bob-h Vegas may not yet properly support Intel 630 graphics.

I think that Magix needs to look again at the QSV layout, as you mentioned in conclusion item 2, I had noticed this a while back, it’s really bad programming to display an option thats not in reality available, needs to be cleaned up, options that are not going to work need to be greyed out, made inactive.

bitman wrote on 11/23/2018, 3:31 PM

@Former user

red car test results (render same as project settings) on Vegas 16, So4 enabled. {Hevc} in brackets. HW Acceleration = Nvidia

i7 8700k with RTX 2080ti

1:18s      Cpu only
0:19s {23s} QSV intel
0:21s {23s} Nvenc 

It is interesting to see my test results (above) and yours (below) are totally equal when using nvenc even tough I use a 6 core and you an 8 core (and I presume a faster cpu), also the QSV is near equal, only the CPU only run is (obviously) showing it's power on the i9. This means that if you use hardware accelerated encoding the i7 8700k still has some headroom and is not fully stressed in Vegas yet.

JN's i9 9900k :

  • Render with Cpu only ...  0:59s ......... [4790K ... 2:11s]
  • Render with QSV ..........  0:18s
  • Render with Nvenc .......  0:21s …..... [4790K ... 0:31s] .... (Hevc is 0:23s with this i9 cpu)

 

 

 

Last changed by bitman on 11/23/2018, 3:55 PM, changed a total of 1 times.

Current system: VP 17 post (build 387), VP16 (suite) build 424, (uninstalled VP 13,14,15), Vegasaur, Magix Video Pro X (VPX11), Corel VS ultimate 2019, a lot of NEWBLUE plugins, Titler Pro 6, Mercalli 4.0, Respeedr, Vasco Da Gamma 12, VASST stuff, Production Assistent pro3, Boris Continuum 2020, Davinci Resolve Studio 16,...

  • OS: Windows 10 Pro 64, version 1909
  • CPU: i9900K stepping R0 (since October 2019), previously, der8auer i7-8700K (advanced edition), default speed (no overclock), Cooler: Noctua NH-D15s
  • RAM: G.Skill Trident Z 3200C14 DDR4 64GB, XMP set to profile 1 in BIOS
  • Videocard: NVIDEA RTX 2080Ti (Founders edition), NVIDEA studio drivers
  • Monitor: LG 38 inch ultra-wide (21x9) - Resolution: 3840x1600
  • C-drive (games & APPS): Samsung NVMe SSD 2TB 960 pro

  • Current Video source work drive: Samsung NVMe SSD 2T 970 EVO plus

  • Mass Data storage & Backup: WD gold 6TB + WD Yellow 4TB

  • MOBO: Gigabyte Z370 Aorus Gaming 7, BIOS F14
  • PS: Corsair HX1200i, Case: Silverstone fortress 2,
  • Misc: Logitech G910, Evoluent Vertical Mouse, shuttlePROv2

 

Former user wrote on 11/23/2018, 3:48 PM

I just, for fun, did an overclock test, added to my main AVC results above in () brackets. OC was just done automatically.

bitman wrote on 11/23/2018, 3:57 PM

@Former user

I just re-edited my last post, and added your red car test results to compare.

Last changed by bitman on 11/23/2018, 4:02 PM, changed a total of 1 times.

Current system: VP 17 post (build 387), VP16 (suite) build 424, (uninstalled VP 13,14,15), Vegasaur, Magix Video Pro X (VPX11), Corel VS ultimate 2019, a lot of NEWBLUE plugins, Titler Pro 6, Mercalli 4.0, Respeedr, Vasco Da Gamma 12, VASST stuff, Production Assistent pro3, Boris Continuum 2020, Davinci Resolve Studio 16,...

  • OS: Windows 10 Pro 64, version 1909
  • CPU: i9900K stepping R0 (since October 2019), previously, der8auer i7-8700K (advanced edition), default speed (no overclock), Cooler: Noctua NH-D15s
  • RAM: G.Skill Trident Z 3200C14 DDR4 64GB, XMP set to profile 1 in BIOS
  • Videocard: NVIDEA RTX 2080Ti (Founders edition), NVIDEA studio drivers
  • Monitor: LG 38 inch ultra-wide (21x9) - Resolution: 3840x1600
  • C-drive (games & APPS): Samsung NVMe SSD 2TB 960 pro

  • Current Video source work drive: Samsung NVMe SSD 2T 970 EVO plus

  • Mass Data storage & Backup: WD gold 6TB + WD Yellow 4TB

  • MOBO: Gigabyte Z370 Aorus Gaming 7, BIOS F14
  • PS: Corsair HX1200i, Case: Silverstone fortress 2,
  • Misc: Logitech G910, Evoluent Vertical Mouse, shuttlePROv2

 

Former user wrote on 11/23/2018, 4:16 PM

I updated the Red car with OC also, not such a difference, but still some for CPU render.

I wasn't expecting to see such improvement for the 4K "Running man" material, nice since I mostly use UHD. Update 24th Nov. .. it was incorrectly done, virtually no improvement.

Meant to say earlier that So4 enabled also in all tests.

"It is interesting to see my test results (above) and yours (below) are totally equal when using nvenc even tough I use a 6 core and you an 8 core (and I presume a faster cpu), also the QSV is near equal, only the CPU only run is (obviously) showing it's power on the i9. This means that if you use hardware accelerated encoding the i7 8700k still has some headroom and is not fully stressed in Vegas yet."

Both GPU's may be using similar asic for HW encoding?

 

bitman wrote on 11/23/2018, 6:39 PM

Both GPU's may be using similar asic for HW encoding?

If you mean the intel iGPU QSV, that could be the same, both are coffee lake, but if you mean the nVENC asic in the RTX2080ti vs 1080, I vaguely recall reading somewhere there was a NVENC difference in the RTX (some sort of improvement). Better quality (IQ) for the same file size or something, I must look it up. But if it is a quality improvement it does not necessary mean a speed improvement.

Current system: VP 17 post (build 387), VP16 (suite) build 424, (uninstalled VP 13,14,15), Vegasaur, Magix Video Pro X (VPX11), Corel VS ultimate 2019, a lot of NEWBLUE plugins, Titler Pro 6, Mercalli 4.0, Respeedr, Vasco Da Gamma 12, VASST stuff, Production Assistent pro3, Boris Continuum 2020, Davinci Resolve Studio 16,...

  • OS: Windows 10 Pro 64, version 1909
  • CPU: i9900K stepping R0 (since October 2019), previously, der8auer i7-8700K (advanced edition), default speed (no overclock), Cooler: Noctua NH-D15s
  • RAM: G.Skill Trident Z 3200C14 DDR4 64GB, XMP set to profile 1 in BIOS
  • Videocard: NVIDEA RTX 2080Ti (Founders edition), NVIDEA studio drivers
  • Monitor: LG 38 inch ultra-wide (21x9) - Resolution: 3840x1600
  • C-drive (games & APPS): Samsung NVMe SSD 2TB 960 pro

  • Current Video source work drive: Samsung NVMe SSD 2T 970 EVO plus

  • Mass Data storage & Backup: WD gold 6TB + WD Yellow 4TB

  • MOBO: Gigabyte Z370 Aorus Gaming 7, BIOS F14
  • PS: Corsair HX1200i, Case: Silverstone fortress 2,
  • Misc: Logitech G910, Evoluent Vertical Mouse, shuttlePROv2

 

Former user wrote on 11/23/2018, 6:46 PM

@bitman

“But if it is a quality improvement it does not necessary mean a speed improvement.”

Thats really what I meant. Maybe the asics are similar for speed? The results between our tests seems to point to that. There are always improvements in the codecs with each generation though.

Peter_P wrote on 11/24/2018, 2:01 AM
Nvenc encoding is faster with iGPU enabled in BIOS. I wonder why?

@bitman

Did you check the GPU usage with the task manager ? I'm quite sure you will see both GPUs beeing unsed.

Peter_P wrote on 11/24/2018, 2:14 AM
Running Man test, origional ... (3 pieces of it end to end, 37s 12f duration.)

Render with either Nvenc, Intel QSV or CPU only.  Render templates set at default for Nvenc and QSV.  The Hevc template render values are in brackets, {Hevc}.  The OC values are in brackets, (Overclock).

HW Acceleration = Nvidia

Render with Nvenc .......  0:23s ... {31s} ….. (8s)

Render with QSV ..........  0:22s ... {35s} ….. (7s)

 

Red Car 1080p

HW Acceleration = Nvidia

Render with Nvenc .......  0:21s …..... [4790K ... 0:31s] .... (22s) … Hevc is 0:23s with this i9 cpu.

Render with QSV ..........  0:18s ……………………………. (18s)

 

@Former user

At what clock is your i9-9900k now running in OC mode ?

I assume, the (OC) values are for AVC output ? The real strange thing is that there is nealy no benefit of the OC on the Read Car 1080p rendering if GPUs are used, but significantly with UHD rendering.

Would you mind to check also the (OC) rendering with UHDp24 HEVC output ?

Peter_P wrote on 11/24/2018, 4:44 AM

Since you both are running with So4 enabled, I did a retest with Vp16 B307 and So4  reader : TRUE and AMD acceleration.

3xRunner UHDp24 to Intel HEVC 40Mbps default template takes again 34s but CPU Package is down to ~60W. The i7-8700k is running with 4.4GHz (AIDA64 data) while rendering.

bitman wrote on 11/24/2018, 6:02 AM

@Peter_P I re-tried the "running man" (Intel HEVC) with So4 reader False, and it is 1 s faster (33s) than True (34s) which is pretty much negligible. I always try a few times, and it is consistent 1s faster. I'll keep it enabled.

Update: Actually with So4 reader False it is a few seconds slower in the non-HVEC render, so I definitely keep it True

Last changed by bitman on 11/24/2018, 6:18 AM, changed a total of 2 times.

Current system: VP 17 post (build 387), VP16 (suite) build 424, (uninstalled VP 13,14,15), Vegasaur, Magix Video Pro X (VPX11), Corel VS ultimate 2019, a lot of NEWBLUE plugins, Titler Pro 6, Mercalli 4.0, Respeedr, Vasco Da Gamma 12, VASST stuff, Production Assistent pro3, Boris Continuum 2020, Davinci Resolve Studio 16,...

  • OS: Windows 10 Pro 64, version 1909
  • CPU: i9900K stepping R0 (since October 2019), previously, der8auer i7-8700K (advanced edition), default speed (no overclock), Cooler: Noctua NH-D15s
  • RAM: G.Skill Trident Z 3200C14 DDR4 64GB, XMP set to profile 1 in BIOS
  • Videocard: NVIDEA RTX 2080Ti (Founders edition), NVIDEA studio drivers
  • Monitor: LG 38 inch ultra-wide (21x9) - Resolution: 3840x1600
  • C-drive (games & APPS): Samsung NVMe SSD 2TB 960 pro

  • Current Video source work drive: Samsung NVMe SSD 2T 970 EVO plus

  • Mass Data storage & Backup: WD gold 6TB + WD Yellow 4TB

  • MOBO: Gigabyte Z370 Aorus Gaming 7, BIOS F14
  • PS: Corsair HX1200i, Case: Silverstone fortress 2,
  • Misc: Logitech G910, Evoluent Vertical Mouse, shuttlePROv2

 

Peter_P wrote on 11/24/2018, 6:25 AM

@bitman

... and you might notice, that the preview (fps) keeps (mostly) constant between UHD events with Vp16 B307 and So4 enabled. On the downside, larger (~40minutes) XAVC-S UHDp30 projects take longer to load and require more RAM.

Former user wrote on 11/24/2018, 6:35 AM

@Peter_P

@bitman

Ok, fullest apologies to you both, major cockup.  I did the OC tests with the single Running man, not the three pieces.  I only saw this when starting to do the OC Hevc test for Peter_P.

Values are now all updated correctly, BTW, all tests loaded and saved to spinning iron, checked with SSD this morning, no difference.

I did a single Hevc OC test … HW Acc = Nvidia, Render with Cpu only ...  1:55s ... {1:10s (1:07) } … (1:52s) … 1:07s vs 1:10s.

Peter_P, re: OC, I just did the in software Asus 5 way OC optimization. In observing the CPU speed while rendering it didn't exceed 4.9 ghz.

Peter_P wrote on 11/24/2018, 7:07 AM

Thanks again very much. I was wondering how a slite OC could get the increase of power.

So there still seems to be no real indication for us to upgrade to the i9-9900k - specially not when using a GPU for rendering.

Former user wrote on 11/24/2018, 7:13 AM

Indeed Peter, maybe your wife won’t have to be told at all now😀

I agree, hardly worth it, I use CPU rendering with HW Acc. for anything important and get a 100% decrease in render times with the upgrade, plus playback has to be better, for occasional video stuff I always use Nvenc, its great. Going from a 6 core to an 8 core and mostly HW render, just hard to make a case for it, IMHO.

Peter_P wrote on 11/24/2018, 8:35 AM

Up to now I could not see any reason for CPU rendering of HEVC UHDp30 output. Quality does not benefit from CPU rendering as it some time did in MPEG or AVC output in the past.

Former user wrote on 11/24/2018, 9:26 AM

Thats good to know.

Former user wrote on 11/24/2018, 1:19 PM

FWIW re: Intel Graphics 630, I found that the "Running man" project did indeed render out with Intel Graphics set for GPU Acc.  However neither the "Red Car" nor the VP 16 sample project would render, both freeze.

I've updated the "Running man" results accordingly.