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

[NickatNight1980]

FX8320 @ 4.16. 16gigs ddr3-1600. timings 9-9-9-27-36
Win10-pro

x265 [info]: consecutive B-frames: 9.6% 1.8% 7.6% 35.9% 45.1%
encoded 1806 frames in 1117.63s (1.62 fps), 11820.04 kb/s, Avg QP:29.09

using cpu capabilities: MMX2 SSE2Fast LZCNT SSSE3 SSE4.2 AVX XOP FMA4 FMA3 BMI1

 

[AnnoyedGrunt]

2700X @ Stock, 16 GB 3200, 14-14-14 RAM, Asrock B450M
Win10
Handbrake 1.2.0

encoded 1806 frames in 430.94s (4.19 fps), 11820.04 kb/s, Avg QP:29.09

 

[Dan057]

HandBrake 1.1.0 (2018040700) - 64bit
OS: Microsoft Windows
CPU: AMD Ryzen 7 2700 Eight-Core Processor
Ram: 16316 MB,
GPU Information:
NVIDIA GeForce GTX 980 Ti - 23.21.13.9135
Screen: 2560x1440

encoded 1806 frames in 410.08s (4.40 fps), 11820.04 kb/s, Avg QP:29.09

with other apps open, like Teamspeak, Chrome, Viber, Skype, Avast...etc

 

[Cata40]

I9 7980xe , ddr4 unbuferred 2133mhz, at stock, load 3.4 ghz
encoded 1806 frames in 211 second , avg fps 8.9

 

[Lexingtonian]

2012 tech (everything old is new again)

Setup:

• 4 AMD Opteron 6386 SE OC'd to 3.5Ghz all cores on a Supermicro Quad Core MoBo.
• No special cooling. All procs run between 24-31C during test.
• 1333Mhz ECC RDIMMs (all cores quad channel) - No special RAM tuning
• Archlinux 4.20.1-arch1-1-ARCH
• Handbrake 1.2.0 GTK compiled with ENABLE_LIBNUMA:BOOL=ON for 8/10/12 bit x265.
• Spectre/Meltdown neutering off at Kernel

encoded 1806 frames in 275.66s (6.55 fps), 10505.01 kb/s, Avg QP:28.59


For a config that's basically less than $600 on Ebay in 2019, it's still a relevant in some small historical way. My guess is I could OC a bit more and get 1ish more FPS out of this box. Maybe 2 (max). Waiting on some new thermal paste and will push them a little harder. Still happy with it, I basically use this box for encoding. It pulls so much power I just turn it on for encoding workloads (like a utility) then turn it back off via IPMI. Works great for that. Cheers all.

Lex

 

[JRzoid]

Should get my 4k monitor tomorrow but no I don't really have a 4k card...the 1060 is not a 4k card but will run some older stuff.

Id rather run the 2600x right now...has the multithreaded Powerrrr.

 

[Lexingtonian]

Should get my 4k monitor tomorrow but no I don't really have a 4k card...the 1060 is not a 4k card but will run some older stuff.

Id rather run the 2600x right now...has the multithreaded Powerrrr.

Wrong thread?

 

[CricketSP]

encoded 1806 frames in 281.80s (6.41 fps), 11820.04 kb/s, Avg QP:29.09

9900K @ 4.9ghz all cores, mem@ 3200.

 

[rvborgh]

hi Lex,

since we have similar setups (aside from processors)... have you tried running the benchmark with

1) in the BIOS: node interleaving off, and SRAT table disabled
2) in the BIOS: node interleaveing on (auto), and SRAT table enabled?

The 6386SE should have better bandwidth than the older K10. i think the K10s had more powerful general purpose FPUs though (aside from the newer instructions that the construction core cpus had).

[edit] i think i recall that the Bulldozer based cpus had a single FPU shared between two integer cores. K10 has one FPU per core. So that would account for the difference i think (32 actual fpu vs 48)... not counting for clock speed.

i'm running ddr3 1600 as well... but its running at only 1333 (i haven't yet upgraded to the OCNG BIOS).

Since i don't run across many other quad socket setups like mine... here's some photos from back in 2014 when i was putting it together. I guess it just brute forces its way through this benchmark

2012 tech (everything old is new again)

Setup:

• 4 AMD Opteron 6386 SE OC'd to 3.5Ghz all cores on a Supermicro Quad Core MoBo.
• No special cooling. All procs run between 24-31C during test.
• 1333Mhz ECC RDIMMs (all cores quad channel) - No special RAM tuning
• Archlinux 4.20.1-arch1-1-ARCH
• Handbrake 1.2.0 GTK compiled with ENABLE_LIBNUMA:BOOL=ON for 8/10/12 bit x265.
• Spectre/Meltdown neutering off at Kernel

encoded 1806 frames in 275.66s (6.55 fps), 10505.01 kb/s, Avg QP:28.59


For a config that's basically less than $600 on Ebay in 2019, it's still a relevant in some small historical way. My guess is I could OC a bit more and get 1ish more FPS out of this box. Maybe 2 (max). Waiting on some new thermal paste and will push them a little harder. Still happy with it, I basically use this box for encoding. It pulls so much power I just turn it on for encoding workloads (like a utility) then turn it back off via IPMI. Works great for that. Cheers all.

Lex

 

[ywasmd]

i5 8400 (Mid 2018) stock vs i5 3210M stock (Macbook Pro Mid 2012) both Win10pro



i5 3210M (Macbook Pro Mid 2012)

CPU: Intel(R) Core(TM) i5-3210M CPU @ 2.50GHz
- Intel microarchitecture Ivy Bridge
Ram: 10146 MB DDR3 1666,
GPU Information:
Intel(R) HD Graphics 4000 - 10.18.10.3345

Max. Temp. 92°

encoded 1806 frames in 3144.79s (0.57 fps), 11820.04 kb/s, Avg QP:29.09


i5 8400 (Mid 2018)

CPU: Intel(R) Core(TM) i5-8400 CPU @ 2.80GHz
Ram: 16036 MB DDR4 2660,
GPU Information:
Intel(R) UHD Graphics 630 - 24.20.100.6194

Max Temp. 61°

encoded 1806 frames in 562.68s (3.21 fps), 11820.04 kb/s, Avg QP:29.09

 

[DaPoets]

encoded 1806 frames in 123.66s (14.60 fps), 11820.04 kb/s, Avg QP:29.09

Threadripper 3960X stock settings. 64GB DDR4 3600Mhz 16-19-19-39-1T

 

[Junky228]

FX8320 @ 4.16. 16gigs ddr3-1600. timings 9-9-9-27-36
Win10-pro

x265 [info]: consecutive B-frames: 9.6% 1.8% 7.6% 35.9% 45.1%
encoded 1806 frames in 1117.63s (1.62 fps), 11820.04 kb/s, Avg QP:29.09

using cpu capabilities: MMX2 SSE2Fast LZCNT SSSE3 SSE4.2 AVX XOP FMA4 FMA3 BMI1

oh man! I upgraded from an 8320 I had running at 5GHz some years ago, got an older xeon as it had better performance for me, even at the 3.34GHz stock speed.

Xeon w3670 @ 4.0GHz 24GB ddr3-1600 9-9-9-24-33-2T Win 8.1-Pro

encoded 1806 frames in 886.67s (2.04 fps), 11820.04 kb/s, Avg QP:29.09

I also had a bunch of stuff open in the background so I don't know how much that would affect it

 

[DrMrLordX]

3900x (default) with DDR4-3733 14-16-14-28:

encoded 1806 frames in 198.75s (9.09 fps), 11820.04 kb/s, Avg QP:29.09

 

[Junky228]

wow the new AMD CPUs are absolutely crushing it

 

[Markfw]

wow the new AMD CPUs are absolutely crushing it

Instead of 12 of those cores, imagine 128 of them in one box !!! It would look exactly like this (but mine are generation one of EPYC, not gen 2/7002 series, and twice the cores that I have)

 

[sdifox]

Ryzen 3700X
16 GB DDR4-3600
Win 10 1909

encoded 1806 frames in 306.52s (5.89 fps), 11820.04 kb/s, Avg QP:29.09

 

[Arkaign]

Instead of 12 of those cores, imagine 128 of them in one box !!! It would look exactly like this (but mine are generation one of EPYC, not gen 2/7002 series, and twice the cores that I have)

Beautiful work sir, I dig those Noctuas on there as well. Running pretty cool?

 

[Markfw]

Beautiful work sir, I dig those Noctuas on there as well. Running pretty cool?

Its running linux, so harder to tell, but I think it was like 46c under full load in windows when I was testing.

 

[Hulk]

4770k running stock @ 3.9GHz
average encoding speed for job is 2.297762 fps

 

[Makaveli]

3800X @ stock

encoded 1806 frames in 313.61s (5.76 fps), 11820.04 kb/s, Avg QP:29.09

 

[sdifox]

3800X @ stock

encoded 1806 frames in 313.61s (5.76 fps), 11820.04 kb/s, Avg QP:29.09

Eh? What speed is your memory?

 

[Makaveli]

Eh? What speed is your memory?

in the sig

DDR4-3200 14-14-14-34 1T

 

[sdifox]

in the sig

DDR4-3200 14-14-14-34 1T

Surprised cpu speed made less diff than memory speed sicne your timing looks better than mine. Mine is 3600 16-18-18 and got 5.89fps

 

[Makaveli]

Surprised cpu speed made less diff than memory speed. Mine is 3600 16-18-18 and got 5.89fps

Not surprised actually encoding has always been senistive to memory bandwidth and the amount of available cores.

I can clock my memory up to that speed and run it again.

But what I will try first is using the Ryzen Dram calculator @3200 speed first to see if that will improve results.

And i'm using handbrake 1.31 not the requested version in the first post which is older

 

[sdifox]

Not surprised actually encoding has always been senistive to memory bandwidth and the amount of available cores.

I can clock my memory up to that speed and run it again.

But what I will try first is using the Ryzen Dram calculator @3200 speed first to see if that will improve results.

And i'm using handbrake 1.31 not the requested version in the first post which is older

I thoguht the newer one would return better results.