Jump to content
HWBOT Community Forums
havli

HWBOT x265 Benchmark

Recommended Posts

Update:

Version 1.1.1 is online http://downloads.hwbot.org/downloads/temp/HWBOT_x265_Benchmark_final_portable.rar The alpha stage has been successful, it is time to move on. I've enabled Beta status, the benchmark is open for submissions. Plenty of free gold cups to take. :D

 

At the moment we have no dedicated video encoding benchmark on HWBOT. So I thought it would be nice to create one. There are plenty of video encoders available on the internet and benchmark applications for most of them already exists. I don't like reinventing the wheel... :) Therefore I picked the most modern encoder - H265/HEVC.

Only one benchmark exists (that I know of) and it isn't really suitable for our needs. http://x265.ru/en/x265-hd-benchmark/'>http://x265.ru/en/x265-hd-benchmark/ My benchmark is working the same way, but GUI is used instead of command line interface. I also added some tweaks and options to get the best possible score on most computers, even very extreme ones. :)

 

HWBOT x265 Benchmark is based on the open source x265 encoder (http://x265.ru/en/). It can take advantage of modern CPUs instructions set and multithread support is also very good. However this benchmarks is also capable of running even on as old processors as Athlon XP (maybe Pentium III as well). Of course on the legacy hardware the encoding time is rather long. There are two presets available - 1080p and 4k. The main goal of both of them is to convert H264 source video to H265/HEVC and measure average fps.

 

x265_1.0.0iixxp.png

 

Now to describe the new Benchmark options:

1. Benchmark type - 32bit or 64bit encoder. Use 64bit if possible, it is faster, on some platforms by quite a big margin.

2. Priority - priority of the encoder process, not much to say here.

3. Pmode - enables better thread utilization, improves performance on some platforms. Also can slow things down a bit, depends on CPU type.

4. Overkill mode:

 

For even better multithreading support it is possible to activate the overkill mode. Two or more (up to 8 in the current version) instances of the encoder will run simultaneously and when all of them are finished, the final score is sum of all sub-scores minus small compensation to avoid score gain by uneven compute time. If the sub-scores time variability is bigger than 5%, the overkill run is considered invalid and no score is generated.

 

The HWBOT x265 Benchmark implements security features which should block any attempt to replace external components of the benchmark (source videos, ffmpeg, x265 encoder) or score manipulation. There are two ways to upload score to HWBOT. Save the data file which contains screenshot and all necessary information. Or direct online submission from the benchmark (currently in development, will be ready in the final version).

Also this benchmark should be safe to run using Windows 8(+) - when HPET is active.

 

Minimum system requirements:

Athlon XP / Pentium 4 (maybe PIII)

1 GB RAM (1080p) / 2 GB (4K)

1 GB free HDD space

Windows XP SP3

Java SE 7 or later

 

Recommended system requirements:

AMD FX / Core 2 Quad 45nm (with SSE 4.1)

4 GB RAM

1 GB free HDD space

Windows 7 x64 (SP1 for AVX support)

Java SE 7 or later

 

Beta testing of v1.0.0 has been running on local PC-related forums for a week or so and no serious bugs were found. So I think the benchmark should be ready soon. There are still some features to finish. When all is done I will enable the public status.

 

In the meantime if you like to test the current v1.0.0 (1080p only), link to download: http://hw-museum.cz/data/hwbot/HWBOT_x265_benchmark.rar

Warning - this version (1.0.0) is meant only for testing, saved data files will not be accepted when the submit option is enabled in the future.

 

Known bugs / things to be improved in the final version:

- wrong Overkill mode formula is displayed. Time of each instance is shown instead of fps.

- final score will be moved below the "1080p benchmark complete" message to be more clear

- online submission feature is work in progress

- portable version of the benchmark (java included) - No need to install java, can be useful on computers where you don't have permission to install stuff

- perhaps enhance the Overkill mode to use more than 8 instances of x265. Could be useful for very large servers (more than 30 CPU cores).

 

 

Combined score table from ongoing beta testing at http://forum.cnews.cz/viewtopic.php?f=22&t=27656 and http://pctforum.tyden.cz/viewtopic.php?f=97&t=229511 You can get some inspiration here - what to expect from your CPU. :)

 

Score------------CPU--------------------Cores/Threads----------Clock-------------OS.............Tester

85,58........2x Haswell-EP ES Xeon........24/48..............3,2GHz.........Win 7 x64--------DOC-Zenith

30,03........Core i7 5820K...................6/12..............4,2GHz.........Win 10 x64-------le1tho

29,68........4x Xeon X7550................32/64.............2,0GHz.....Win Srv 2008 R2 x64--skipped1

21,81........Core i7 4770K...................4/8...............4,7GHz........Win 10 x64--------l.zdvorak

19,92........Core i5 4690K...................4/4...............4,9GHz.........Win 10 x64-------iOioo

19,44........Core i7 4790....................4/8...............4,0GHz.........Win 10 x64-------skipped1

16,05........Core i5 4670K...................4/4...............4,3GHz.........Win 10 x64-------Darth Daron

14,26........Core i5 4670K...................4/4...............3,6GHz.........Win 7 x64--------JXP

14,03........Core i7 2600K...................4/8...............4,4GHz.........Win 7 x64--------Caderom

13,39........FX-8350.........................8/8...............4,4GHz.........Win 7 x64--------husbja

12,92........Core i5 3570K...................4/4...............4,6GHz.........Win 7 x64--------NoNeStaciTi

12,51........FX-8350.........................8/8...............4,0GHz.........Win 10 x64-------berazde

12,20........Core i7 3770K...................4/8...............3,7GHz.........Win 7 x64--------Killing joke

11,85........Core i5 2500K...................4/4...............4,6GHz.........Win 10 x64-------havli

11,21........FX-8300.........................8/8...............3,6GHz.........Win 7 x64--------Mani2

10,94........Core i5 3470....................4/4...............3,8GHz.........Win 7 x64--------pohodar

10,79........Core i5 3570K...................4/4...............3,8GHz.........Win 10 x64-------Profi-Lama

6,75.........FX-6300.........................6/6...............3,6GHz.........Win 10 x64-------kolecko

6,72.........Core i7 5500U...................2/4...............2,7GHz.........Win 8.1 x64------Tomix

5,81.........A10-5700........................4/4...............3,7GHz.........Win 10 x64-------hob

4,83.........Core i5 3230M...................2/4...............3,0GHz.........Win 10 x64-------RayEndCZ

4,83.........Core i3 2100....................2/4...............3,1GHz.........Win 7 x64--------Caderom

4,45.........Xeon L5410......................4/4...............2,33GHz........Win 7 x64--------havli

4,37.........Core i5 2410M...................2/4...............2,7GHz.........Win 10 x64-------cpt.America97

3,68.........Core i5 520M....................2/4...............2,66GHz........Win 10 x64-------hob

3,05.........Phenom II X4 965..............4/4...............3,4GHz.........Win 10 x64-------cpt.America97

2,9..........Athlon 5350.....................4/4...............2,05GHz........Win 7 x64--------Tralalák

2,49.........A8-3850.........................4/4...............2,9GHz.........Win 8.1 x64------Jan Olšan

2,22.........Core i3 2367M...................2/4...............1,4GHz.........Win 7 x64--------Tralalák

2,17.........2x Xeon 5110....................4/4...............2,0GHz.........Win 7 x64--------havli

2,12.........Phenom II X3 720..............3/3...............3,2GHz.........Win 10 x64-------siddhi

1,42.........Atom Z3740......................4/4...............1,86GHz........Win 8.1 x64------Jan Olšan

1,16.........VIA U4650E......................4/4...............1,0GHz.........Win 7 x64--------Tralalák

0,71.........Core 2 Duo T5500..............2/2...............1,66GHz........Win XP-----------melkor unlimited

0,38.........Pentium 4 640...................1/2...............3,2GHz.........Win XP-----------skipped1

Edited by havli
update

Share this post


Link to post
Share on other sites

Oh, one more thing - I am really curious how good Skylake is in this benchmark.

If you have one, please post a screenshot. Preferably using Windows 8 / 8.1 / 10 with HPET enabled and disabled (so I can check whether timer detection works on socket 1151 platform).

Share this post


Link to post
Share on other sites

Yeah, if it works, then I will update the minimum system requirements. :)

 

Also the final version is almost ready. All the mentioned "Known bugs / things to be improved in the final version" things are fixed / implemented.

Share this post


Link to post
Share on other sites

Well, it work :) It mention that it use CPU features: MMX2, SSE and Cache32 (whatever that means) and 3% progress suggest that it works. I would wait for the final score, but seems that there is not any problem, even on highly nonstandard machine like the Dell OptiPlex GX110 is. (it use 24bit resolution depths - cannot have 32bit, so... even GPU-Z is buggy there, lol)

(pay no attention to the not detected ram frequency, that don't even latest CPU-Z detect well, as you see on the HWbot prime screenshot)

 

I downloaded a latest release 7 Java from there:

https://edelivery.oracle.com/akam/otn/java/jdk/7u80-b15/jre-7u80-windows-i586.exe

(for those, who did not have a Oracle account and not feeling like one, there are alternatives by me:

http://depositfiles.com/files/ulsoty1uf

http://www.mediafire.com/?vvg5k5nonuqwoip

http://rapidgator.net/file/190288e6602d80677c1490d15c35743d/JavaRuntime_build_7.80.exe.html )

 

...and at first, I give a HWbot prime a try. Worked:

HWbot_Prime_685_96.jpg

(see the GPU-Z bug? Low level programming sometimes fail, expecting 32bit, where 24bits happen :D )

 

So I try the x265 and it is running now. Albeit slowly, but that is somewhat expected on 512MB SDRAM running at "whooping" 100MHz. And even I gathered good 2-2-2-5 SDRAM chips, they all come to vain, as the FSB bus 133MHz is divided in the chipset to 100MHz for rams and that kill any chance of speed...

 

HWbot_x265_running.jpg

 

There is no swapping (witch is good, because the old 80G clunker would have a hard time and the PATA to SATA interface refused to work with 30G Corsair Nova2 SSD and better SSDs are currently busy with other tests...), so a 512MB of fresh WinXP SP3 install (just disabled the Security center service).

 

I understand that I slow down a bit the run when going for the screenshot and copying it over network to another PC, but all in all I doubt it will have any serious impact, as we all can predict that this is going to be the slowest run ever made yet :) And you deserve to see that it does work. Hopefully the menu frontend can be done without dreaded Java in the future, either by using other programming language or at least including few need libs and that will be it.

Benchmarks should be standalone, IMHO.

 

Pretty good work, mate!

Edited by trodas

Share this post


Link to post
Share on other sites

Cool, so it works on PIII after all. And even with 512 MB RAM. :cool:

 

Unfortunately Java is the only suitable language for me. My c++ or c# skills are way too low for complicated project like this. The final version however will be a portable benchmark with built-in java... :)

 

Thank you for the testing.

 

btw - little reminder - no need to waste more power and time continuing this run (unless you are curious what the final score will be :D). The result file is not valid for (future) submission.

Share this post


Link to post
Share on other sites

Yep, I'm curious and I know that more people are curious too. And it did not even swap. 18% is done, so I wait for the final - overnight it will do the trick. The only time it start swapping was, when screensaver kick it. I have to go and disable it, lol.

 

Given the number of Java security holes I think that you should learn C/C++ :) I know, easier to say that done, but maybe for next version? :) Nevermind. I will report by the morning what is the result. Currently 0.06fps and 19.33% :) Not bad, we are almost in 1/5 done!

Share this post


Link to post
Share on other sites
My c++ or c# skills are way too low for complicated project like this. The final version however will be a portable benchmark with built-in java... :)

 

If it's open source I can look into porting it.

Share this post


Link to post
Share on other sites

Hopefully it will be made Open Source... or at least closed source between you two?

 

So, my result is - 0.08fps (it get faster when I did not take screens/screensaver don't run):

 

HWbot_x265_0_08fps.jpg

 

The resulting file is rather big, 512k in size, so there it is:

http://depositfiles.com/files/zpjnjf80g

 

(backups: http://www.mediafire.com/?fvuar520p60dukc

http://rapidgator.net/file/4d0330c7bdd04c0cbf3ca79120217cfe/_HWBOT_x265-1080p_0.08_fps.hwbot.html

https://mega.co.nz/#!LFMBkZrK!0_aCBZJDmvloKG4wqDMrsnCqSyFMVynq2Mtw1igBFzo )

 

Hopefully that helps ;)

Share this post


Link to post
Share on other sites
Oh, one more thing - I am really curious how good Skylake is in this benchmark.

If you have one, please post a screenshot. Preferably using Windows 8 / 8.1 / 10 with HPET enabled and disabled (so I can check whether timer detection works on socket 1151 platform).

 

On it!

Share this post


Link to post
Share on other sites

Well, opensource... I don't know, the code is quite a mess. :D And also rather long - over 10000 lines of code. Maybe in future.

 

The screenshot inside data file is in PNG format, I dont like the artifacts usually produced by JPEG.

 

0.58fps @ 4K is good score af a ULV Laptop. My workstation based on two Core2 Xeons (4C/4T in total) only scores 0.4fps.

 

I'm looking forward to see the Skylake score. Hopefully the Timer detection will be working reliably, as it does on older platforms.

Share this post


Link to post
Share on other sites

useplatformclock = no

 

attachment.php?attachmentid=3173&stc=1&d=1439892667

 

attachment.php?attachmentid=3174&stc=1&d=1439892667

 

useplatformclock = yes

 

attachment.php?attachmentid=3175&stc=1&d=1439892942

Edited by Massman

Share this post


Link to post
Share on other sites

You are using windows 10, right? Weird thing is cpu-z detects it as Vista. Therefore the benchmark runs even without HPET.... which it shouldn't.

 

Can you please upload hardware.txt located in x265 folder (only available when the benchmark is running)?

Share this post


Link to post
Share on other sites

Hey Havli, if this would ever make it to HWBot as a for-points benchmark, I think it would be helpful to add another decimal to FPS. Easy to do? Difficult?

Share this post


Link to post
Share on other sites
Hey Havli, if this would ever make it to HWBot as a for-points benchmark, I think it would be helpful to add another decimal to FPS. Easy to do? Difficult?

 

Or set the score as MBit/s rather than FPS.

Share this post


Link to post
Share on other sites
Hey Havli, if this would ever make it to HWBot as a for-points benchmark, I think it would be helpful to add another decimal to FPS. Easy to do? Difficult?

Yes, this is possible and relatively easy. X265.exe only provides two decimal digits, but I can calculate the final score with more precision easily as: total frames / elapsed time. It would be best to implement this change before the benchmark goes public.

 

MBit/s sounds kinda weird. FPS is a common unit for video encoding speed, I want to keep it that way.

Share this post


Link to post
Share on other sites
You are using windows 10, right? Weird thing is cpu-z detects it as Vista. Therefore the benchmark runs even without HPET.... which it shouldn't.

 

Can you please upload hardware.txt located in x265 folder (only available when the benchmark is running)?

 

http://forum.hwbot.org/attachment.php?attachmentid=3178&stc=1&d=1439948039

 

About the granularity of the score: note that HWBOT will only show granularity up to 2 digits.

Share this post


Link to post
Share on other sites

Well, no need to change the score precision then.

The Windows Vista - 10 mixup seems to be cpu-z error, this is beyond my power to fix. And Skylake is not affected by the RTC bug, so no problem there. http://forum.hwbot.org/showthread.php?p=402927#post402927

 

If the OS is detected wrong on older platforms too, please use older cpu-z version 1.72.1 (copy it to the x265 older) - there was no problem with win10 during my testing.

 

http://hwbot.org/benchmark/hwbot_x265_benchmark_-_1080p/

http://hwbot.org/benchmark/hwbot_x265_benchmark_-_4k/

 

Enjoy. :)

Share this post


Link to post
Share on other sites
Hopefully it will be made Open Source... or at least closed source between you two?

 

So, my result is - 0.08fps (it get faster when I did not take screens/screensaver don't run):

 

HWbot_x265_0_08fps.jpg

 

The resulting file is rather big, 512k in size, so there it is:

http://depositfiles.com/files/zpjnjf80g

 

(backups: http://www.mediafire.com/?fvuar520p60dukc

http://rapidgator.net/file/4d0330c7bdd04c0cbf3ca79120217cfe/_HWBOT_x265-1080p_0.08_fps.hwbot.html

https://mega.co.nz/#!LFMBkZrK!0_aCBZJDmvloKG4wqDMrsnCqSyFMVynq2Mtw1igBFzo )

 

Hopefully that helps ;)

 

how many time needed with pIII ?!

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×