• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Question Handbrake 1.3.3 - Benchmark your System - New benchmark criteria

Hulk

Diamond Member
Oct 9, 1999
3,108
554
126
Back in 2018 ElFenix posted the original Handbrake Benchmark Thread. I thought it was cool so I started updating the thread with a results table. With his permission I'm posting an update to that thread. If you want to participate we're going to tighten up the testing methodology (see below) so the results provide data better for analysis.

I'll update the second post of this thread periodically with results.

The test file:
sdifox said:
et voila. we can all host the file and round robin it :p

LG New York HDR UHD 4K Demo.zip

drive.google.com
drive.google.com


1. Use the following version of Handbrake with the built-in h.265 mkv 2160p60 preset
HandBrake-1.3.3-x86_64-Win_GUI.exe
Don't forget to turn on logging in Handbrake so you can retrieve your time. Tools>Preferences>Advanced>Logging
Once this current version is replaced you'll be able to access this version from the following link.
HandBrake: Nightly Builds
Nightly builds of HandBrake
handbrake.fr

2. Report average effective clockspeed using HWinfo - https://www.hwinfo.com/
I'm sure there are better ways to do this using the logging feature but here's how I did it. I downloaded the portable version, which runs from the exe file.

Run the exe and select "sensors only." You only need this window open.
Minimize the sensors window so you can have it and Handbrake visible at the same time. Select the "average effective clockspeed" line so you can spot it easily. It will highlight.
Note the icon for "reset min/max/average values and timer" in the bottom right area of the window, it's the clock icon.

Start the handbrake but don't start/reset the timer until handbrake is actually encoding video. Record the average effective clockspeed right before handbrake finishes the encode. It should settle in pretty quickly after the encode starts and not change during the encode.

More info on "effective" vs "instant" clock: https://www.hwinfo.com/forum/threads/effective-clock-vs-instant-discrete-clock.5958/

3. Report "Total CPU Usage" (also from HWinfo)

4. Report your encoding time, CPU Model, and RAM specs (amount, speed, latency)
 
Last edited:
  • Like
Reactions: ElFenix

Hulk

Diamond Member
Oct 9, 1999
3,108
554
126
RankUserCPU/MemoryHandbrake VersionCoressecondsfpsfps/core/GHz (fps per available GHz)Total CPU Usage (Hwinfo)Average Effective Clockspeed (Hwinfo)CPU CorePercent of Top Score
1XabanakFanatik5950x 16/32 - DDR4-4000, CL16 1:11.3.316118.3215.260.24182.4%3960Zen 3 (Vermeer)
100.0%​
2Det0x5950x 16/32 - 32GB, DDR4-3800, C141.3.316118.6715.220.24082.5%3959Zen 3 (Vermeer)
99.7%​
3Noid5900x - 12/24 - 2x16 32GB 3800 Cas181.3.312157.9211.440.24085.4%3969Zen 3 (Vermeer)
74.9%​
4hate1875900x 12/24 pbo -25 all core auto - 2x16 CL16 3600mhz1.3.312169.1110.680.23686.7%3763Zen 3 (Vermeer)
70.0%​
5Kenmitch5900x - 12/24 - 16GB DDR4-3800, C141.3.312172.3810.480.23886.2%3664Zen 3 (Vermeer)
68.6%​
6CakeMonster5900X 12/24 - 2x16Gb 3600MHz @ CR1 16-16-16-36 (44-320)1.3.312186.149.700.23192.3%3502Zen 3 (Vermeer)
63.6%​
7.vodka3900x 12/24 - 4x16GB DDR4-3800 16-20-16-381.3.312191.079.450.22680.2%3488Zen 2 (Rome)
61.9%​
8Arni9011900K 8/16 - 32GB DDR4-36001.3.38219.268.240.21492.3%4815Rocket Lake (Cypress Cove)
54.0%​
9JoeRambo10900k 10/20 - DDR4-3900, C15 (hand tuned)1.3.310234.717.690.15597.2%4957Comet Lake (Skylake)
50.4%​
10Makaveli5800x 8/16 - 16GB DDR @ 3200 CL14 + Dram Fast profile1.3.38239.837.530.22991.4%4115Zen 3 (Vermeer)
49.3%​
11brinstar117Xeon E5-2673v4 20/40 - DDR4-2133 ECC 15-15-15-36 (quad)1.3.320271.476.650.18674.5%1791Broadwell
43.6%​
12Kenmitch5600x 6/12 - 3600 CL14 memory1.3.36280.566.440.23996.7%4486Zen 3 (Vermeer)
42.2%​
13richierich12125600x 6/12 - 16GB @ 3600 CL 16-18-18-18-38-84-1T1.3.36314.845.740.23995.6%3994Zen 3 (Vermeer)
37.6%​
14dr13375600x 6/12 - 16GB DDR4 at 32001.3.36330.505.460.24096.8%3798Zen 3 (Vermeer)
36%​
15ElFenixRyzen 3600x 6/12 - 16GB DDR4-32001.3.36367.914.910.20697.4%3970Zen 2 (Rome)
32.2%​
16Hitman9282700 8/16 - DDR4-3200, 14-17-17-18-341.3.38379.674.760.15090.6%3970Zen+ (Summit Ridge)
31.2%​
17EvilRickRyzen 7 2700 8/16 - DDR4-3200, C161.3.38391.984.610.14795.2%3910Zen+ (Summit Ridge)
30.2%​
18burninatortech42700X 8/16 - 16GB DDR4@3200 CL141.3.38414.954.350.13798.9%3970Zen+ (Summit Ridge)
28.5%​
19Mir76TA3500x 6/12 - G.Skillz AGES 3000 16 18 18 38 56 1T1.3.36477.403.780.12699.0%5007Zen 2 (Rome)
24.8%​
20Mir76TA2700 8/16 - 32GB DDR4-24001.3.38500.823.610.14199.0%3193Zen+ (Summit Ridge)
23.6%​
21mikk1165G7 Dell Inspiron 15 7506 - 2x16GB DDR4-3200 22-22-22-521.3.34775.012.330.21599.0%2712Tiger Lake (Willow Cove)
15.3%​
22Hulk4770k 4/8 - 8GB DDR2-800, C111.3.34813.322.220.15398.1%3620Haswell
14.5%​
23tamz_msc1135G7 - HP 14s-dr2006tu 4/8 - DDR4, CL151.3.34850.772.120.21799.2%2447Tiger Lake (Willow Cove)
13.9%​
24Hulk7500T - Dell Optiplex Micro 7050 -8GB - DDR41.3.34965.001.870.15299.2%3070Kaby Lake (Skylake)
12.3%​
25Hulk8250u 4/8 (Surface 2 laptop) - 8GB DDR3-1066, C101.3.341065.001.700.17798.3%2400Kaby Lake R (Skylake)
11.1%​
26Hulki3-4130 2/4 - 8GB DDR2-800, C111.3.321682.561.070.15999.3%3370Haswell
7.0%​
27Hulk4130T 2/4 - HP All-In-One q116 - 8GB DDR4 16001.3.321813.391.000.15799.5%3175Haswell
6.5%​
28ElFenixi3-4130 2/4 - 8GB DDR31.3.321842.970.980.14698.7%3349Haswell
6.4%​
29Hulk5200u 2/4 - 12GB DDR3-1600 (Lenovo t450s)1.3.322225.230.810.16399.8%2490Broadwell
5.3%​
SCORES WITH AVX512 ENABLED - ASM512
RankUserCPU/MemoryHandbrake VersionCoressecondsfpsfps/core/GHz (fps per available GHz)Total CPU Usage (Hwinfo)Average Effective Clockspeed (Hwinfo)CPU Core
1Arni9011900K 8/16 - 32GB DDR4-36001.3.38215.488.380.23191.5%4543Rocket Lake (Cypress Cove)
2mikk1165G7 Dell Inspiron 15 7506 - 2x16GB DDR4-3200 22-22-22-521.3.34722.672.500.23196.2%2700Tiger Lake (Willow Cove)
3tamz_msc1135G7 - HP 14s-dr2006tu 4/8 - DDR4, CL151.3.34899.812.010.25589.0%1970Tiger Lake (Willow Cove)

1617934541736.png
 

Attachments

Last edited:

ElFenix

Elite Member
Super Moderator
Mar 20, 2000
101,624
5,919
126
Ryzen 3600X, 16GB DDR4@3200, HB 1.3.3, windows 1909, 97.4% usage, 3970.4 clock, encoded 1806 frames in 367.91s (4.91 fps), 11820.04 kb/s, Avg QP:29.09
 

Kenmitch

Diamond Member
Oct 10, 1999
8,502
2,244
136
My 5900x at stock clocks. I forgot to snip a photo of the activity log, but the overall score is there anyways. Less clocks and load then
I originally had thought.

Handbrake.png
 

Det0x

Senior member
Sep 11, 2014
545
759
136
Baby 5950x testrun with PBO enabled, static OC would score much better in this bench.

14.89 fps done.png

Kinda strange the benchmark says my average fps is 15.1 in the last 3 seconds in the screenshot, but the text log says 14.89 fps
encoded 1806 frames in 121.28s (14.89 fps), 11820.04 kb/s, Avg QP:29.09
3807mhz average effective allcore clock
Total cpu usage average = 80.9% / Max cpu/thread usage average = 90.8%
237 max watt usage / 200 watt average usage (!)
 
Last edited:

Kenmitch

Diamond Member
Oct 10, 1999
8,502
2,244
136
Puzzles me why 5900x is at 80.2% average effective CPU while 5950x with more cores is at 82.5%? Memory seems virtually identical? What does the "1:1" mean in XabanakFanatik's results mean. I'm not well versed on AM4 memory architecture.
I think your mixing up the 3900x and 5900x results is why. They're correct in the chart. The 1:1 means the infinity fabric and memory clocks are in sync.
 

Hulk

Diamond Member
Oct 9, 1999
3,108
554
126
I think your mixing up the 3900x and 5900x results is why. They're correct in the chart. The 1:1 means the infinity fabric and memory clocks are in sync.
Yes you are right. But now I'm wondering why the 5900x has higher average effective CPU usage than 3900? Perhaps the better Out-of-Order scheduler/branch predictor is doing a better job?
 

Kenmitch

Diamond Member
Oct 10, 1999
8,502
2,244
136
Yes you are right. But now I'm wondering why the 5900x has higher average effective CPU usage than 3900? Perhaps the better Out-of-Order scheduler/branch predictor is doing a better job?
Less cores doing more work?
 

ElFenix

Elite Member
Super Moderator
Mar 20, 2000
101,624
5,919
126
i3-4130/8GB DDR3something/windows 2004/3.4GHz avg/99.9% usage

encoded 1806 frames in 2236.56s (0.81 fps), 11820.04 kb/s, Avg QP:29.09


closed a bunch of programs/turned off folding et walla:

i3-4130 / 8GB DDR3 1600/11/11/28 / windows 2004 / 3.35GHz avg / 98.7% usage (though handbrake itself was usually a couple % below that and dropped a bit a few times when i had to open acrobat)

encoded 1806 frames in 1842.97s (0.98 fps), 11820.04 kb/s, Avg QP:29.09

explorer_2020-12-15_16-19-49.png
 
Last edited:

Det0x

Senior member
Sep 11, 2014
545
759
136
New run from me, still with PBO enabled.

encoded 1806 frames in 119.34s (15.13 fps), 11769.09 kb/s, Avg QP:29.15
handbrake.png

3901mhz average effective allcore clock
Total cpu usage average = 81.8% / Max cpu/thread usage average = 92.1%
259 max watt usage / 214 watt average usage (!)
 

Det0x

Senior member
Sep 11, 2014
545
759
136
This is interesting. Too bad we don't have more data but let's try to analyze this.

We have two 16 core Zen 3 parts, which both have average effective CPU of about 82%.

12 core Zen 3 at 86% and 12 core Zen 2 at 80%.

10 core Comet Lake at 97%

The rest of the scores are from 8 or less core parts and except for 1 result they all are 95% or greater.

The most obvious conclusion is that Handbrake isn't able to use 16 cores/32 threads very effectively. But then we have the 80% usage for the 12 core Zen 3 and the 90% for the 10 core Comet Lake.

Without more scores to analyze I'm stumped?

I noticed that the three results you provided with nearly 100% CPU usage were benchmarks, which are specifically designed to stress all available cores.

Do you have any programs that you use day-to-day that stress the CPU higher than Handbrake?
Think its better if we take this discussion here in this thread instead :)

And the answer to your question relays soley on what preset we use.. with the "production max" i have 100% usage.
prod max preset.png

Or if i modify the h.265 preset to higher quality setting i can also use all 32 threads:
h.265 higher quality.png
 
Last edited:

burninatortech4

Senior member
Jan 29, 2014
381
107
116
Ryzen 2700X (stock), 16GB DDR4@3200 CL14, HB 1.3.3, windows 20H2, 98.9% usage, 3970.4 clock, encoded 1806 frames in 414.95s (4.35 fps), 11820.04 kb/s, Avg QP:29.09
 

Hulk

Diamond Member
Oct 9, 1999
3,108
554
126
@Det0x,

Thanks for moving the discussion to the proper forum.

I'm going to take a crack at analyzing this and hope one of the many people a lot smarter than me will step in.
Why does my 5950x use 80% CPU on this test at QP 24 but 100% CPU at QP 40?
Handbrakes scans the video to identify a GOP (group of frame) to be encoded
Each frame is converted to a compressed image use quantization selected with the QP slider (like jpeg)
Regardless of the QP slider value or preset the DCT (discrete cosine transform) above uses the same amount of compute
But, motion/vector searches require more compute for more detailed (lower QP values)
Motion/vector search is not highly threaded and is the most compute intensive part of the encoding process
If the motion search is still progressing after the DCT is finished then all available thread for 12+ core CPUs are not utilized
When the QP slider is set to higher values or when using faster presets the motion/vector search "keeps up" with the DCT and other encoding work and CPU usage remains high.
If this analysis is correct then at a QP of 24 a highly threaded CPU (16 or more physical core) should use more CPU as the presets are moved to faster options
 
Last edited:
  • Haha
Reactions: richierich1212

Det0x

Senior member
Sep 11, 2014
545
759
136
Veryfast = 82.1%
veryfast.png

Faster = 82.6%
faster.png

Fast = 82.3%
fast.png

Medium = 79.3%
medium.png

Slow = baseline

Slower = 82.5%
slower.png

Very slow = 84.7%
veryslow.png

Placebo = 94.4% (run took 22min)
placebo.png
 

ASK THE COMMUNITY