Benchmark your computer with Handbrake 1.01 and x265

[superstition]

I remember when (some) took exception to naming a module as one core

Vishera does have 8 integer cores. It also has 4 FPU cores. AMD's architecture isn't a copy of Intel's so its cores aren't the same as Intel's. Labeling "cores" without differentiating between big architectural differences isn't very accurate as a result. There is nothing incorrect about referring to modules rather than cores. It's a lot more clarifying.

It is an objective fact that there are four modules with eight threads. What is less accurate is calling an 8350 4 cores or 8 cores. As far as I recall, Cinebench erroneously listed Piledriver as 4 cores. That's substantially more inaccurate than calling it 8 cores. That's also why the lawsuit had no merit. The guy suing tried to argue that the design makes it a 4 core part which isn't the case. A core doesn't have to have an FPU to be a core. However, for accuracy purposes, it's better to label the modules as modules. This is why I think AMD should have marketed construction cores in that manner.

Update: Here is another score for the list

i5 750 "Lynnfield" 4C/4T @3.8 16GB 1600 9-9-9-24-1T Win7, 425.33s (3.52 fps), 4036.02 kb/s, Avg QP:26.24

update 2: updated score. Third and final run gave tiny improvement.

update 3: edited crash's post and my response due to his request. I was not arguing. I just explained why it's better to use the module naming convention.

 

[crashtech]

EDIT:

HT OFF:

CPU: Intel(R) Core(TM) i7-5820K CPU @ 3.30GHz (4.5GHz actual)
encoded 1497 frames in 149.44s (10.02 fps), 4036.02 kb/s, Avg QP:26.24

 

[Hail The Brain Slug]

CPU: Intel(R) Core(TM) i7-5820K CPU @ 3.30GHz (4.5GHz actual)
encoded 1497 frames in 149.44s (10.02 fps), 4036.02 kb/s, Avg QP:26.24[/B]

Is the 4.5 GHz overclock accurate? Judging from my 5960x score and the 5930k score we have I would think you should score somewhere around 130-140, 10 or more seconds faster than that. What's your cache and memory at?

 

[crashtech]

Cache @ x36, my X99 Extreme 4 does not seem to like cache overclocks. Not exactly sure on memory. will edit.

Edit: The RAM did get set to 2133 somehow, but setting it back to 2666 made no difference to my score.

 

[superstition]

Here's another score from me to add:

i7 6700K, stock, 16GB 3200 16-18-18-38-2T, Win 10AE, 177.11s (8.45 fps), 4036.02 kbps, Avg QP:26.24

note: AsRock's "multi core enhancement*" is turned on in BIOS as well as power-saving states and turbo. Voltage is set to fixed 1.170 with level 2 LLC. The system is using an Intel 600p 512GB NVMe ssd.

*"Perform the highest frequency on all CPU core at the same time"

 

[Drazick]

Hello,
What would be the command line to run this test on Command Line Version?

It will also make the configuration in tact for all testers .

Update

Those who want to use HandBrake CLI (No installation needed) the command line (Windows) is (See Command Line Guide):

HandBrakeCLI.exe -Z "H.265 MKV 1080p30" -v 1 -i D:\sample-Elysium.2013.2160p.mkv -o D:\TryOut.mkv

In the case above the file was on D:\ Drive. One could change accordingly.

My results on Intel Core i7-6800K @ 3.4 GHz (Stock) , ASRock X99M Killer, Memory 4 x G.Skill F4-2800C16-8GRK @ 2133 MHZ, Windows 10 64 Bit:

encoded 1497 frames in 166.14s (9.01 fps)

What I found interesting is that the CPU Temperature never got above 45 Celsius.
I wonder why SpeedBoost didn't kicked in and the CPU were running at 3.4 GHz almost 95% of the time.
I'd expect it to bump speed to 3.6 GHz.

 

[ElFenix]

i'll get to work on spreadsheeting all the scores on google docs and making it shareable.



please don't argue in this thread.

edit: we're cool

 

[Hulk]

encoded 1497 frames in 220.59s (6.79 fps), 4036.02 kb/s, Avg QP:26.24

220.59s, Haswell 4770k stock, 8GB DDR-1600 RAM, Windows 7 - Hulk

Lenovo T450s
673.52s, i5 T5200 (Broadwell), 12GB DDR 1600, Windows 8.1

 

[crashtech]

please don't argue in this thread.

I apologize. I have removed my ill-advised comment, and encourage you and superstition to do the same. I actually have no idea why I did that.

 

[ElFenix]

i have no idea how to make pretty bar graphs

 

[Makaveli]

let edgemeal handle it if he has time

 

[SKORPI0]

Could be faster if system is OC'd to 4.3 GHz.
-------------------------------------------------------------------------
HandBrake 1.0.1 (2016122900) - 64bit
OS: Microsoft Windows NT 10.0.10586.0 - 64bit
CPU: Intel(R) Core(TM) i7-3930K CPU @ 3.20GHz
Ram: 32707 MB,
GPU Information:
AMD Radeon R9 200 Series - 15.301.1801.1001
AMD Radeon R9 200 Series - 15.301.1801.1001
Screen: 2560x1440

-------------------------------------------
encoded 1497 frames in 233.88s (6.40 fps), 4036.02 kb/s, Avg QP:26.24
# Encode Completed ...

Screenshot:

 

[rancherlee]

Didn't see a Kavari on the list so I ran my Daughters rig
A10-7850K @ 3.7ghz DDR2133 11-11-11-27-1T windows 10
encoded 1497 frames in 575.35s (2.60 fps)

 

[thigobr]

3570K @ 4200MHz / 4x4GB 1866MHz 9-9-9-28-1T Windows 10
encoded 1497 frames in 309.68s (4.83 fps), 4036.02 kb/s, Avg QP:26.24

 

[ElFenix]

updated!

 

[crashtech]

Is the 4.5 GHz overclock accurate? Judging from my 5960x score and the 5930k score we have I would think you should score somewhere around 130-140, 10 or more seconds faster than that. What's your cache and memory at?

I know why the score is low, it's because HT is off. I remoted in to verify this, but will need to be there to fix it.

 

[Hail The Brain Slug]

I know why the score is low, it's because HT is off. I remoted in to verify this, but will need to be there to fix it.

Ahhh, that makes sense. I was wondering if disabling HT on 6 or 8 core i7's would improve scores since it would help offset some of the poor scaling past 4C/8T, but this would help show it adds more than the scaling takes away.

 

[Iron Woode]

You can add me, too:

CPU: Intel(R) Core(TM) i5-4690K CPU @ 3.50GHz - Intel microarchitecture Haswell - logical processor count: 4
OpenCL device #1: NVIDIA Corporation GeForce GTX 1060 3GB - OpenCL version: 1.2 CUDA - driver version: 376.33 - device type: GPU - supported: no
Intel Quick Sync Video support: no

encoded 1497 frames in 269.03s (5.56 fps), 4036.02 kb/s, Avg QP:26.24

might be faster if I reboot win 10.

a reboot helped:

encoded 1497 frames in 267.35s (5.60 fps), 4036.02 kb/s, Avg QP:26.24

I have 8GB DDR3 1333 ram.

 

[crashtech]

Ahhh, that makes sense. I was wondering if disabling HT on 6 or 8 core i7's would improve scores since it would help offset some of the poor scaling past 4C/8T, but this would help show it adds more than the scaling takes away.

It's very dependent on the app, in this case the operations are highly parallel, which takes away from SMT's effectiveness, but the data set is large and being fetched from main memory all the time, which plays to SMT's strength.

 

[Jan Olšan]

the E5420 doesn't support AVX and it got the same as the AVX CPUs, it's something else.

Only AVX2 really matters, AVX ("AVX1") has negligible influence. The reason is that x265 like all video encoders works with integer datatypes (should be 8bit and 16bit math). AVX didn't have usable instructions for needed operations on these datatypes being floating-point focused. Those instructions only come in AVX2 and for that reason, AVX2 is what matters for video encoding, not AVX. AVX1 only has some very minor use in calculations of rate control decisions (mbtree in x264, cutree in x265). In x264, AVX is also used for 3-operand functionality, but I think that might be obsoleted by mov elimination in newer CPUs?

As a result, disabling/enabling AVX on Sandy/Ivy or AMD chips should have only small impact, but having AVX2 on Haswell/Broadwell and Skylake makes huge difference

with a "Core 2 Duo" Pentium(T4400):
encoded 1497 frames in 2798.17s (0.53 fps), 3978.85 kb/s, Avg QP:26.14

if you can disable 2 cores of a quad core to test if it changes the bitrate it could be interesting

x265 AFAIK requires SSE4 which is artificially disabled on this Pentium. For that reason, it suffers a big speed hit, because a lot of the hand-written assembly requires SSE4 and in its absence, scalar code is ran isntead. This particularly affects AMD K10 chips which run x265 with maybe 1/3 of their possible speed they would have should SSE2 versions of all assembly exist (like it exists in x264).

 

[crashtech]

My updated score with HT enabled, I am going to edit my original post accordingly. @ElFenix , feel free to either replace my old score with this one, or include them both as HT on and HT off...

HT ON:

CPU: Intel(R) Core(TM) i7-5820K CPU @ 3.30GHz (4.5GHz actual)
encoded 1497 frames in 130.23s (11.49 fps), 4036.02 kb/s, Avg QP:26.24

 

[superstition]

I apologize. I have removed my ill-advised comment, and encourage you and superstition to do the same. I actually have no idea why I did that.

In light of your request I have removed the contentious label from your post and my response. It's a legitimate question and so is the answer so I don't see why the post needs to be deleted.

ElFenix, you missed my Lynnfield score:

Update: Here is another score for the list

i5 750 "Lynnfield" 4C/4T @3.8 16GB 1600 9-9-9-24-1T Win7, 425.33s (3.52 fps), 4036.02 kb/s, Avg QP:26.24

 

[ElFenix]

In light of your request I have removed the contentious label from your post and my response. It's a legitimate question and so is the answer so I don't see why the post needs to be deleted.

ElFenix, you missed my Lynnfield score:

That weird because I remember typing that result. Probably lost it during a conversion to space delimited text

 

[VirtualLarry]

Kaby Lake G4600 3.6Ghz dual-core with HT, no AVX/AVX2, DDR4-2400 @ 2400 (single-channel), Windows 10 Pro 64-bit 1607

encoded 1497 frames in 580.38s (2.58 fps), 4036.02 kb/s, Avg QP:26.24

Just marginally faster than my G4400 @ 4.455Ghz, it looks like. Around the ballpark.

NOT FOR LEADERBOARD. Just for curiousity, I went and unchecked the "Disable QSV" and did another encode.

QSV: "encoded 1497 frames in 538.97s (2.78 fps), 4036.02 kb/s, Avg QP:26.24"

Not that much faster, really.

 

[superstition]

That weird because I remember typing that result. Probably lost it during a conversion to space delimited text

I just noticed the other OS X result (with the other Macbook Pro) disappeared, too.

Macbook Pro "Late 2013", i7-4850HQ, 2.3 GHz base, macOS 10.12.2

16 GB 1600L, Iris Pro OpenCL version 1.2

encoded 1497 frames in 283.30s (5.28 fps), 4036.54 kb/s, Avg QP:26.24