Benchmark your computer @4K with Handbrake 1.1 and H265! NOW WITH ZEN3!

[Hail The Brain Slug]

5950x
4.4 GHz all core
DDR 4000 CL16 1:1
Average effective clockspeed 3647
Total CPU usage 82.4%

Handbrake 1.3.3

encoded 1806 frames in 128.51s (14.05 fps), 11820.04 kb/s, Avg QP:29.09

 

[JoeRambo]

10900K @5.1 10C10T
DDR4 3900C15 handtuned
average effective clockspeed 4957
Total CPU usage 97.2%

Handbrake 1.3.3

encoded 1806 frames in 234.71s (7.69 fps), 11820.04 kb/s, Avg QP:29.09

 

[Hulk]

Zen 3 - What?!

Rank	User	CPU	Handbrake Version Verified	Total CPU Usage (Hwinfo)	Average Effective Clockspeed (Hwinfo)	seconds	fps/core/GHz (fps per available GHz)	fps	CPU Core	Cores
1	XabanakFanatik	5950X 16/32	1.3.3	82.4%	3647	129	0.241	14.05	Zen 3 (Vermeer)	16
2	JoeRambo	10900K	1.3.3	97.2%	4957	235	0.155	7.69	Comet Lake (Skylake)	10
3	EvilRick	Ryzen 7 2700x 8/16	1.3.3	95.2%	3910	392	0.147	4.61	Zen+ (Summit Ridge)	8
4	Hulk	4770K 4/8	1.3.3	98.1%	3620	815	0.153	2.22	Haswell	4
5	Hulk	8250u 4/8 (Surface 2 laptop)	1.3.3	98.3%	2400	1065	0.177	1.70	Kaby Lake R (Skylake)	4

 

[Hail The Brain Slug]

Zen 3 - What?!

I'm not even getting started yet..... 4.4 is just a temporary low power overclock because I'm using a U12A. The Optimus AM4 block is on the way

Based on the voltages I'm able to run 4.6 ghz for short term testing I wouldn't be surprised if 4.7-4.8 all core will be doable on water. I'm 100% thermally bound at the moment.

 

[Hitman928]

Here's my results on Linux (Ubuntu 20.10). I used turbostat to get CPU utilization and frequency since HWinfo is Windows only.

encoded 1806 frames in 379.67s (4.77 fps), 10567.33 kb/s, Avg QP:28.58

AMD 2700 (8c/16t) @ 4.25 GHz
Handbrake 1.3.3
DDR4@ 3200 MHz, tuned RAM timings.
3970 MHz average frequency
90.6% average utilization

Edit: Also, @Hulk , you should probably edit the first post in the thread as new people may read the first post and then add their results without getting the updated info.

Edit2: Sorry, just noticed the OP is @ElFenix. Maybe we should just start a new thread?

 

[Hulk]

Here's my results on Linux (Ubuntu 20.10). I used turbostat to get CPU utilization and frequency since HWinfo is Windows only.

encoded 1806 frames in 379.67s (4.77 fps), 10567.33 kb/s, Avg QP:28.58

AMD 2700 (8c/16t) @ 4.25 GHz
Handbrake 1.3.3
DDR4@ 3200 MHz, tuned RAM timings.
3970 MHz average frequency
90.6% average utilization

Edit: Also, @Hulk , you should probably edit the first post in the thread as new people may read the first post and then add their results without getting the updated info.

Edit2: Sorry, just noticed the OP is @ElFenix. Maybe we should just start a new thread?

I would like to start a new thread but I won't do it without the OP's permission.

 

[Hulk]

I'm not even getting started yet..... 4.4 is just a temporary low power overclock because I'm using a U12A. The Optimus AM4 block is on the way

Based on the voltages I'm able to run 4.6 ghz for short term testing I wouldn't be surprised if 4.7-4.8 all core will be doable on water. I'm 100% thermally bound at the moment.

While the overall score is fantastic what really amazes me is what appears to be a tremendous IPC increase for Zen 3 over everything else. We'll know more when (if) we get some more results with the new more rigorous testing format.

 

[LightningZ71]

Of note in the Zen3 bench result is the low CPU usage. If my guess is correct, in this one test, the core is do efficient that it's outstripping the memory subsystem's ability to keep it fed with data, and even with SMT, it's still spending a good bit of time just waiting for more data.

Perhaps more of a memory overclock?

 

[Hail The Brain Slug]

Of note in the Zen3 bench result is the low CPU usage. If my guess is correct, in this one test, the core is do efficient that it's outstripping the memory subsystem's ability to keep it fed with data, and even with SMT, it's still spending a good bit of time just waiting for more data.

Perhaps more of a memory overclock?

Can't overclock it really past 4000 realistically. I can run a couple times at lower memory settings to see if it hampers performance or decreases average effective clock. Probably once at 3200c14 and once at 3600c16.

I can also disable a CCD and see if with only 8C/16T it will achieve higher average effective clock.

 

[Kenmitch]

HandBrake 1.3.3 (2020061300)
OS: Microsoft Windows NT 10.0.19042.0
CPU: AMD Ryzen 9 5900X 12-Core Processor @ default 4224 MHz Average.
Ram: 16309 MB, (3800_C14)

encoded 1806 frames in 172.38s (10.48 fps), 11820.04 kb/s, Avg QP:29.09

I'll try again with PBO enabled and update.

HandBrake 1.3.3 (2020061300)
OS: Microsoft Windows NT 10.0.19042.0
CPU: AMD Ryzen 9 5900X 12-Core Processor with PBO 4427 MHz Average
Ram: 16309 MB, (3800_C14)

encoded 1806 frames in 163.07s (11.07 fps), 11820.04 kb/s, Avg QP:29.09

Usage fluctuated between 94-99%

 

[Hulk]

Of note in the Zen3 bench result is the low CPU usage. If my guess is correct, in this one test, the core is do efficient that it's outstripping the memory subsystem's ability to keep it fed with data, and even with SMT, it's still spending a good bit of time just waiting for more data.

Perhaps more of a memory overclock?

I have a feeling it's Handbrake's inability to continue linearly scaling with more cores. 10 cores is scaling pretty well but there is obviously a drop off to 16. We'll know more when we get some 12/14 core results. From what I've seen in the past most video operations require lots of compute, not so much memory bandwidth. It's interesting. We'll see.

 

[moinmoin]

I have a feeling it's Handbrake's inability to continue linearly scaling with more cores. 10 cores is scaling pretty well but there is obviously a drop off to 16. We'll know more when we get some 12/14 core results. From what I've seen in the past most video operations require lots of compute, not so much memory bandwidth. It's interesting. We'll see.

If that were the case, shouldn't turning off SMT help higher cores CPUs? After all if scaling is getting worse with a higher thread count, the threads getting whole cores might start to outweigh the numerical advantage of more threads sharing cores through SMT. Anybody with a 3950X or 5950X willing to test for this hypothesis?

 

[Hulk]

If that were the case, shouldn't turning off SMT help higher cores CPUs? After all if scaling is getting worse with a higher thread count, the threads getting whole cores might start to outweigh the numerical advantage of more threads sharing cores through SMT. Anybody with a 3950X or 5950X willing to test for this hypothesis?

That's a good theory. But if 16 cores are at 82% usage overall I think that's still effectively utilizing all of the cores. I think it's going to take 24 or 32 physical cores to test your theory. Seems like when CPU usage gets closer to 50% perhaps it'd so better with HT off.

It also depends on how Handbrake is allocating resources. It might be assigning a certain block of work to a core, meaning that if HT is off performance will suffer.

 

[.vodka]

Here's my 3900x @ per CCX OC (4350, 4350, 4300, 4300, 1.25v) with the new rules.

Had to drop clocks down a bit from my last entry (4375, 4375, 4325, 4300) as this is now hitting 90°C and handbrake was crashing. Thanks early summer and your ~30°C ambient temperatures! My air cooler hates you

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

HandBrake 1.3.3 (2020061300)
OS: Microsoft Windows NT 10.0.19042.0
CPU: AMD Ryzen 9 3900X 12-Core Processor
RAM: 4x16GB DDR4-3800 16-20-16-38 (tuned secondaries + teritiares), fclk 1900MHz

encoded 1806 frames in 191.07s (9.45 fps), 11820.04 kb/s, Avg QP:29.09

Hwinfo runtime: 3m 10s (~190s)
Average effective clockspeed: 3488MHz
Total CPU usage (average): 80.2%

 

[Hulk]

Here's my 3900x @ per CCX OC (4350, 4350, 4300, 4300, 1.25v) with the new rules.

Had to drop clocks down a bit from my last entry (4375, 4375, 4325, 4300) as this is now hitting 90°C and handbrake was crashing. Thanks early summer and your ~30°C ambient temperatures! My air cooler hates you

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

HandBrake 1.3.3 (2020061300)
OS: Microsoft Windows NT 10.0.19042.0
CPU: AMD Ryzen 9 3900X 12-Core Processor
RAM: 4x16GB DDR4-3800 16-20-16-38 (tuned secondaries + teritiares), fclk 1900MHz

encoded 1806 frames in 191.07s (9.45 fps), 11820.04 kb/s, Avg QP:29.09
Hwinfo runtime: 3m 10s (~190s)
Average effective clockspeed: 3488MHz
Total CPU usage (average): 80.2%

Can you tell us your average effective clock and CPU usage percentage from HWinfo please?

 

[.vodka]

Can you tell us your average effective clock and CPU usage percentage from HWinfo please?

Click the linked text. There's all the relevant hwinfo data you need.

 

[Hulk]

Click the linked text. There's all the relevant hwinfo data you need.

Sorry! Don't know how I missed that!

 

[ElFenix]

I would like to start a new thread but I won't do it without the OP's permission.

go right on ahead, OP is now a mess trying to put the new info into it.


it's kinda weird that i didn't get @Hitman928 's summon.

 

[Hulk]

HandBrake 1.3.3 (2020061300)
OS: Microsoft Windows NT 10.0.19042.0
CPU: AMD Ryzen 9 5900X 12-Core Processor @ default 4224 MHz Average.
Ram: 16309 MB, (3800_C14)

encoded 1806 frames in 172.38s (10.48 fps), 11820.04 kb/s, Avg QP:29.09

I'll try again with PBO enabled and update.

HandBrake 1.3.3 (2020061300)
OS: Microsoft Windows NT 10.0.19042.0
CPU: AMD Ryzen 9 5900X 12-Core Processor with PBO 4427 MHz Average
Ram: 16309 MB, (3800_C14)

encoded 1806 frames in 163.07s (11.07 fps), 11820.04 kb/s, Avg QP:29.09

Usage fluctuated between 94-99%

If you use HWinfo as indicated in my post you'll be able to get values for average effective clock and CPU usage. We need those number from HWinfo to make an apples-to-apples comparison.
Thanks - Mark

 

[Hulk]

go right on ahead, OP is now a mess trying to put the new info into it.


it's kinda weird that i didn't get @Hitman928 's summon.

Okay thanks. I didn't want to steal your thunder!

 

[LightningZ71]

This seems to

HandBrake 1.3.3 (2020061300)
OS: Microsoft Windows NT 10.0.19042.0
CPU: AMD Ryzen 9 5900X 12-Core Processor @ default 4224 MHz Average.
Ram: 16309 MB, (3800_C14)

encoded 1806 frames in 172.38s (10.48 fps), 11820.04 kb/s, Avg QP:29.09

I'll try again with PBO enabled and update.

HandBrake 1.3.3 (2020061300)
OS: Microsoft Windows NT 10.0.19042.0
CPU: AMD Ryzen 9 5900X 12-Core Processor with PBO 4427 MHz Average
Ram: 16309 MB, (3800_C14)

encoded 1806 frames in 163.07s (11.07 fps), 11820.04 kb/s, Avg QP:29.09

Usage fluctuated between 94-99%

This seems to support my theory. We're at 75% of the threads with similar clocks and memory throughput and we have a much higher usage percentage. It still doesn't rule out some sort of algorithm or thread sharing limitation on scaling.

 

[Mir96TA]

HandBrake 1.3.3 (2020061300)
OS: Microsoft Windows NT 10.0.19042.0
CPU: AMD Ryzen 5 3500X 6-Core Processor
Ram: 32721 MB,
GPU Information:
Radeon (TM) RX 480 Graphics - 27.20.14501.24001

encoded 1806 frames in 480.11s (3.76 fps), 11820.04 kb/s, Avg QP:29.09

 

[Det0x]

HandBrake 1.3.3 (2020061300)
OS: Microsoft Windows NT 10.0.19042.0
CPU: AMD Ryzen 5 3500X 6-Core Processor
Ram: 32721 MB,
GPU Information:
Radeon (TM) RX 480 Graphics - 27.20.14501.24001

encoded 1806 frames in 480.11s (3.76 fps), 11820.04 kb/s, Avg QP:29.09

We have a new updated thread here