Benchmarking results continued

JN- wrote on 1/18/2020, 12:56 PM

Continued from the now locked “Benchmarking” thread ...

Moderator's Note August 2021:

Project sample for this benchmarking project can be downloaded here:

https://www.dropbox.com/s/3i4xgnhl9fmc0je/BenchmarkProject4K.zip?dl=0

(Project works in Vegas Pro 16 and newer).

The participant's benchmark entry has been changed via Google Form:

Submit your result here : https://forms.gle/2gnzmv93Ar8P1tkx8

And view the result here: https://docs.google.com/spreadsheets/d/1b3ggVifKsuT-cp2kQHjum_KnQ4-2jBmUIvzmu7BQZ34/

Thank you user @RogerS for continuing mantaining this benchmark system.

Please message him if you have questions or problems regarding the render benchmark.

----------------------

Note that the render time results are for the full project duration, the average FPS value is for the “Region 1” duration only. Please enter only the same "Average FPS" value for all of your different encoder and frame size entries.

Don't change the Project properties. Whether checking your FPS/rendering UHD or FHD, leave as is.

Templates and user data input with examples … Use similar for Nvenc, QSV, VCE, and CPU encoders.

The data rates of the 2 templates below is 24/12 for FHD and 50/28 for UHD. So irrespective of which encoder you're using, just use those data rates..

Set the Intel QSV quality slider to 4, default. If this range of 1 to 7 changes in the future leave it at whatever the default is.

My rendered out file sizes on disc. You should get somewhat similar, but not necessarily the exact same results.

FHD size = 48.9 mb ..... Data rate = 9,583 kbps ... Nvenc encoding

FHD size = 50.8 mb ..... Data rate = 9,971 kbps ... Cpu encoding

UHD size = 114 mb ..... Data rate = 22,611 kbps ... Nvenc encoding

UHD size = 117 mb ...... Data rate = 23,308 kbps ... Cpu encoding

If your output file sizes and data rates are not similar, then check that you are using the correct render template settings.

Sample User Data Input with example, text file is in the ZIP file …

Use the instructions in the screenshots below to get the Average FPS value …

If possible set the Dynamic ram Preview to zero while doing so.

Note that the region 1 section above is only used to get the average FPS value, do render out the full length project to get the various render times.

The older "Benchmarking" thread.

Back to post

Comments

CarlM wrote on 1/31/2020, 3:05 PM

User name       : CarlM
   Machine           : OB
   VP version       : 17
   VP Build number       : 387
   CPU           : AMD 3970X
   Cores           : 32
   GPU           : GTX 1080
   iGPU           : N/A
   OC           : No
   Render frame size   : UHD
   HW Acc.           : NVidia
   Render As       : Magix AVC
   Encode mode       : NV Encoder
   Render time       : 1:35
   FPS           : 9.15
Comments : Preview RAM set to 0 for FPS test and 200MB (default) for Rendering test. System RAM 64GB CL14, NVMe SSD storage. Benchmark project seems significantly GPU bound, as observed CPU load hit only 5-10% except for brief burst to 15% at end of the project. Such low load however allowed native AMD Turbo function to run the CPU at 4.1-4.2 Ghz, where as under full load it settles down to about 3.9 Ghz. FHD render speed was 56 seconds.

Instrumented desktop recording of rendering the benchmark project:

More on this system build is over in this thread:

https://www.vegascreativesoftware.info/us/forum/amd-threadripper-3970x-w-vegas-pro-edit-17--118631/

Video on the system build:

Video on the storage sub-system:

JN- wrote on 1/31/2020, 3:39 PM

@CarlM Thanks Carl, I've updated your data in the screenshots in the first post in this thread and in the ZIP with updated spreadsheets.

TheRhino wrote on 1/31/2020, 5:46 PM

I get burned-out editing clients' work in my studio, which has a 9900K workstation, so I've been editing family videos from the comfort of my living room Laziboy recliner using a new $999 Walmart Evoo 17 Gaming Laptop w/17" IPS 144hz screen, 6-core i7-9750H CPU, GTX 2060, 16GB DDR4, 1TB M.2 for the OS, another empty M.2 slot for my source video & empty SSD slot for my target video. (Which I plan to fill as soon as I find a good deal...) I added a $88 Walmart 3 year protection plan for peace of mind but online reviews confirm that this is a good bang/buck laptop based on a popular frame...

Here are my benchmark numbers. Not too shabby against much pricier systems including my $1350 9900K / VEGA 64 upgrade from last April...

User name:           TheRhino
   Machine:           Walmart Evoo 17 Gaming Laptop
   VP version:       17
   VP Build:           387
   CPU:                 Intel i7-9750H
   Cores:               6
   GPU:                 GTX 2060
   iGPU:                630
   OC:                   No
   Render frame: UHD
   HW Acc.            NVidia
Render As:        Magix AVC
   Encode mode:   NVENC
   Render time:     1:41 (my 9900K is 1:39)
   FPS:                  7 (6-8 range) vs. 9900K/VEGA @ 13.25
   Render frame: FHD
   HW Acc.            NVidia
   Render As:        Magix AVC
   Encode mode:   NVENC
   Render time:     1:02 (my 9900K is 0:48)
   FPS:                  7 (6-8 range) vs. 9900K/VEGA @ 13.25

Comments: $999 Walmart Evoo 17 (same quality frame as Eluktronics); 16GB, 1TB M.2, empty
M.2 & SSD slots for source & target drives. Intel 6373 iGPU & AMD 441.87 drivers.

JN- wrote on 1/31/2020, 6:42 PM

@TheRhino Data is updated now. Seriously good bang for your buck, no doubt. Incredible value really.

CarlM wrote on 1/31/2020, 6:53 PM

For fun here's an instrumented desktop recording of the benchmark project:

TheRhino wrote on 2/23/2020, 11:40 AM

New Benchmarks for V17 Build 421 with better AMD support... The BIGGEST change is that I can use the latest Intel iGPU drivers & latest AMD drivers & get slightly better results. Previously, if I didn't use the older Intel 6373 drivers, my render times increased. In comparison, for my Evoo 17 laptop with Nvidia RTX 2060 I get about the same render times as previously listed...

User name:           TheRhino
   Machine:           DIY 9900K, 8 cores, OC'd to 4.8 ghz (dropped from 4.9 for silent running...)
   VP version:       17, Build 421
   GPU:                 VEGA 64 LQ w/LATEST 20.2.1 Driver
   iGPU:                630 w/LATEST 7870 Driver vs. very old 6373
   Render frame: UHD
   HW Acc.            AMD
Render As:        Magix 4K AVC
   Encode mode:   AMD VCE
   Render time:     1:34 (was 1:39) QSV=1:34
   Loop FPS:         13.25 average

   Render frame: FHD
   HW Acc.            AMD
   Render As:        Magix AVC 1080p
   Encode mode:   AMD VCE
   Render time:     0:45 (was 0:48) QSV=0:45
   Loop FPS:        13.25 average

Vegas 11 Red Car Test = 0.14 at 4.8 ghz (silent) & 0.13 at 5.0 ghz (but fans spin-up...)

So even though I dropped my overclock from 4.9 to 4.8 ghz, and didn't stop any background processes, etc., my render times have decreased. These settings keep my fans quiet so I can record nearby....

JN- wrote on 2/23/2020, 11:47 AM

@TheRhino I'll update shortly, I'm away from my PC at the moment.

”These settings keep my fans quiet”

Doesn’t work for me, no matter what I do they keep applauding 🤣

JN- wrote on 2/23/2020, 1:41 PM

@TheRhino Is updated now. As an aside from this benchmarking, I think only yourself and @BruceUSA with his rig achieved 14s for the red car test, well done.

I got 15s (Red Car) with the older Intel drivers for QSV, but their newer ones changed all of that. I find VP17 slower rendering than VP16 for these render tests. I must update to the latest Intel drivers soon. Anyway thanks for all of the data, I'm sure it'll be of help to other users looking at upgrading etc.

TheRhino wrote on 2/23/2020, 3:59 PM

I just updated my results to show that I get 0:13 (every time with AMD VCE) on the Red Car Test if I set my overclock to 5.0 ghz. The other tests are exactly the same as the 4.8 ghz setting, as well as my day-to-day work. The 3 fans spin-up on my H150i liquid cooler at 5.0 ghz, so I have been keeping my 9900K at 4.8 ghz for a silent studio... (I also have several quiet Noctua fans throughout the case...)

EDIT: V17 VCE produces 0:13 (100% of the time) with the LATEST AMD 20.2.2 & Intel 7870 drivers & I still get 0:14 for QSV renders with the new drivers...

JN- wrote on 2/23/2020, 4:04 PM

@TheRhino Well, congratulations on that, it will be some time before thats overtaken. Oh sorry, maybe not, I can just hear @BruceUSA ordering a new rig🤣.

Reyfox wrote on 3/9/2020, 1:13 PM

I hope I did this right. I rendered the small area only.

Username: Reyfox

Machine: Ryzen

VP version: 17

VP Build number: 421

CPU: AMD 9 3900X

Cores: 12

GPU: RX480

iGPU: None

OC: No

Render frame size: UHD

HW Acc.: AMD

Encode mode: AMD VCE

Render As: Magix HEVC/AAC MP4 - Internet UHD 2160p 25 fps (AMD VCE)

Render time: 0:24 (only between the markers)

FPS: 7-7.9 average

@james-ollick when I rendered at 1080HD 25fps (AMD VCE) I get these numbers:

Render time: 11 seconds

FPS: 15.7

JN- wrote on 3/9/2020, 1:31 PM

@Reyfox Thanks for taking the time to do the benchmarking test. Unfortunately you do need to render out the full project length for your render times. Region 1 section is only for playback fps testing.

I'll post your results when you’ve updated the render times, and a single fps value.

Also, just supply a single fps value, for region 1, its independent of your frame size render output, thanks.

Reyfox wrote on 3/9/2020, 1:59 PM

Rendered the entire timeline.

Render As: Magix HEVC/AAC MP4 - Internet UHD 2160p 25 fps (AMD VCE)

Render time: 2:50

FPS: 6.17

If I've made another error, remember you are conversing with an almost brain dead 68 year old...

Region 1 playback (Best Full): 5.6fps

JN- wrote on 3/9/2020, 3:02 PM

@Reyfox Well you’re several years younger than me, so make sure to check out my updates, just in case🤣. I’ll update later on, not at PC now.

Since theres only room for one fps entry i’ll enter your highest average value.

JN- wrote on 3/9/2020, 3:25 PM

@Reyfox Do you want to also include the render time for FHD?

Reyfox wrote on 3/9/2020, 5:29 PM

@JN- it's up to you what you include in the spreadsheet. Including the FHD would help someone, sure, but it was only for that short segment and not the entire video render at 1080. The one brain cell has shut off...

What is this, a senior citizen's home for video editors?? 😁

JN- wrote on 3/9/2020, 5:44 PM

@Reyfox What I meant to say was did you want to redo the FHD render for the full project length?

Reyfox wrote on 3/9/2020, 6:31 PM

@JN- these are the results since I retested again making sure of everything. Hardware is still the same. Use what you need. I hope it helps!

Render As: Magix HEVC/AAC MP4 - Internet HD1080p 25 fps (AMD VCE)

Render time: 0:55

FPS: 19

--------------------------------------------------------------------------------------------------

Render As: Magix HEVC/AAC MP4 - Internet UHD 2160p 25 fps (AMD VCE)

Render time: 2:14

FPS: 8

--------------------------------------------------------------------------------------------------

Render As: Magix AVC/AAC MP4 - Internet HD 1080p 25 fps (AMD VCE)

Render time: 0:57

FPS: 19

---------------------------------------------------------------------------------------------------

Render As: Magix AVC/AAC MP4 - Internet UHD 2160p 25 fps (AMD VCE)

Render time: 3:09

FPS: 5.56

JN- wrote on 3/9/2020, 7:08 PM

@Reyfox I've updated your data now, in the first post in this thread and in the zip file, 2 screenshots and 3 spreadsheets are updated. The data is useful I think for anyone considering either an upgrade or new PC. Quite a few, 21 unique users have now contributed, with a large variety of systems, take it easy.

TheRhino wrote on 3/9/2020, 7:11 PM

@Reyfox FPS is how well in FPS your preview screen displays the looped section at best settings... It is not how fast your system renders... The FPS will rise & fall as you watch the video. Take the average between the fastest FPS and slowest FPS as you watch the loop few times through...

JN- wrote on 3/9/2020, 7:27 PM

@TheRhino Indeed. I didn’t want to labour the point, but several contributors are supplying multiple fps values. Fortunately in the midst of many of @Reyfox's entries he had earlier on entered this “FPS: 7-7.9 average” which I have averaged and used. I assume its the fps average for the loop region only?

NickHope wrote on 3/10/2020, 6:00 AM

User name: NickHope
Machine: OB (Asus Deluxe-X99)
VP version: 17
VP Build number: 421
CPU: Intel i7-5960X
Cores: 8
GPU: AMD Vega 64
iGPU: None
OC: CPU overclocked to 4.3GHz. GPU not.
Render frame size: UHD
HW Acc.: AMD
Encode mode: AMD VCE
Render As: Magix HEVC/AAC MP4 - Internet UHD 2160p 25 fps (AMD VCE)
Render time: 1:40
FPS: 9-16, ave 12.5
Comments: so4compoundplug OFF. UVD OFF. GPU is ASUS ROG Strix RX VEGA64, driver 20.Q1.1. Windows 10 version 1511.

User name: NickHope
Machine: OB (Asus Deluxe-X99)
VP version: 17
VP Build number: 421
CPU: Intel i7-5960X
Cores: 8
GPU: AMD Vega 64
iGPU: None
OC: CPU overclocked to 4.3GHz. GPU not.
Render frame size: UHD
HW Acc.: AMD
Encode mode: AMD VCE
Render As: Magix HEVC/AAC MP4 - Internet UHD 2160p 25 fps (AMD VCE)
Render time: 1:40
FPS: 8-14, ave 11
Comments: so4compoundplug ON. UVD OFF. GPU is ASUS ROG Strix RX VEGA64, driver 20.Q1.1. Windows 10 version 1511.

User name: NickHope
Machine: OB (Asus Deluxe-X99)
VP version: 17
VP Build number: 421
CPU: Intel i7-5960X
Cores: 8
GPU: AMD Vega 64
iGPU: None
OC: CPU overclocked to 4.3GHz. GPU not.
Render frame size: UHD
HW Acc.: AMD
Encode mode: AMD VCE
Render As: Magix HEVC/AAC MP4 - Internet UHD 2160p 25 fps (AMD VCE)
Render time: 1:40
FPS: 5.3-6.3, ave 5.8
Comments: so4compoundplug ON. UVD ON. GPU is ASUS ROG Strix RX VEGA64, driver 20.Q1.1. Windows 10 version 1511.

[From above 3 tests, note that so4compoundplug reduces my playback smoothness, and AMD UVD reduces it more (also confirmed with my Panasonic GH4 AVC projects). Neither affected rendering time.]

All my MAGIX HEVC tests took exactly the same time as the AVC renders, so I won't bother listing them here.

Reyfox wrote on 3/10/2020, 6:24 AM

@JN- yes, my averages early were for the looped region only. Sorry about the mistake.

JN- wrote on 3/10/2020, 6:43 AM

@Reyfox AOK Reyfox. Great to have the results. Doing @NickHope's now, will update when done.

1 2 3 4 … 15

of 15