HWBOT x265 Benchmark

[cbjaust]

cant hwbot host this? 40KB/s is pretty slow going for a 500GB file!

[MaddMutt]

I would normally be okay with what is voted by the community. I have a problem with x265 2.2 in scoring. This my be classified as a BUG or a CHEAT it's all how you wish to look at it

We have the x265 4K 2.2 in this years TC - 2019. I pulled out my "OLD i3-6320" to compete in the Dual Core section. I noticed that SWEET - (I've talked to and also looked at other scores) using the same Motherboard and CPU combo  - BEAT - -> Destroyed <- my x265 submission. I had over 1.3GHz higher CPU&IMC speed but he scored 2x better than me??? https://hwbot.org/submission/3955708_maddmutt_hwbot_x265_benchmark___1080p_core_i3_6320_8.845_fps

Is this problem reproducible YES... Here are screen shots of the the x265 1080p and 4K results. I do not have my system overclocked but running at BIOS default.

Is this limited to only this CPU, CPU/MB combo, Software, Ect, Ect???

Thank You

[Leeghoofd]

Quote

 

special non K bios being used? AVX coding goes down the drain over a certain Bclock (102.8ish) Perfectly could explain why your OCed submission scores way lower than the top scores (at way lower MHz)

Edited September 14, 2019 by Leeghoofd

[Mr.Scott]

Not to mention, Alex is about a thousand times better bencher than you are.

[Guest]

3 hours ago, Leeghoofd said:

special non K bios being used? AVX coding goes down the drain over a certain Bclock (102.8ish) Perfectly could explain why your OCed submission scores way lower than the top scores (at way lower MHz)

What he said ^. Bclk bouncing at/ over 103 will break it.

[nick_name]

When using Overkill does anyone else see the top or bottom progress bar run away from the other?  I've played with a bunch of different things and it feels like it's a Windows 10 scheduler bug.  Or is that normal behavior?  I've only seen one person bench it on YouTube and I myself only have about a day of experience playing with the benchmark.  

[Mr.Scott]

Seen it before. No idea what causes it.

I know overkill likes a lot of memory though, if that helps you out.

[jab383]

I see that big difference between the overkill bars when priority is above NORMAL - in either Win 7 or Win 10.  Try all the priorities and take a peak in task manager during the run to see what actual priority is being used for the workload processes.

[nick_name]

1 hour ago, jab383 said:

I see that big difference between the overkill bars when priority is above NORMAL - in either Win 7 or Win 10.  Try all the priorities and take a peak in task manager during the run to see what actual priority is being used for the workload processes.

I've played with all the priorities too, but that alone didn't cure it for me.  I did see favorable results combining that while playing with other Windows power plans though.  

 

What program is it in Task Manager?  Is it Javaw?

Edited August 23, 2020 by nickname2019

[Mr.Scott]

12 minutes ago, nickname2019 said:

I've played with all the priorities too, but that alone didn't cure it for me.  I did see favorable results combining that while playing with other Windows power plans though.  

 

What program is it in Task Manager?  Is it Javaw?

Yes

[nick_name]

59 minutes ago, Mr.Scott said:

Yes

That stays Normal when I change priority in the bench, but messing with that Javaw priority didn't yield different results.  

Edited August 23, 2020 by nickname2019

[havli]

Javaw.exe is just the benchmark GUI, this one should be left as it is. It has no effect on the performance. When you launch the benchmark run with overkill enabled, then 2 or more instances of the x265 encoder will show up in the device manager and these should have the selected priority set.

When running overkill with high(er) priority, the Windows scheduller might assing more system resources to one instance and less to the other - which results in one instance running faster and the other slower. If the score varies too much, then final score won't be valid.

[nick_name]

1 hour ago, havli said:

-snip-

When running overkill with high(er) priority, the Windows scheduller might assing more system resources to one instance and less to the other - which results in one instance running faster and the other slower. If the score varies too much, then final score won't be valid.

Is there anything that can be done about the scheduler assigning more resources to one instance over the other?  Or is that by design?  

In my experiences the gap between the two is usually 40% ~ 45%.  

[havli]

Perhaps some Windows tweaks, but that is just a wild guess. I never liked tweaking OS.

[Splave]

Try playing with pagefile, no pagefile, try playing background service priority and different win32priorityseparation values in the registry, try the new 20** build of windows 10 or depending on your hardware windows 8 or server 2012. Sometimes it is what it is and fails due to variance. Run it again. 

[Fragmeister]

I had issues with this today, tried many diff mem timings and mem speed but the only thing that solved it for me was to lower the cpu speed by 100 MHz, ymmv

[nick_name]

On 8/24/2020 at 7:32 PM, Splave said:

Try playing with pagefile, no pagefile, try playing background service priority and different win32priorityseparation values in the registry, try the new 20** build of windows 10 or depending on your hardware windows 8 or server 2012. Sometimes it is what it is and fails due to variance. Run it again. 

Ooh, I read your write up over at Tom's and am eager to try dual rank my next go around.  And will give your above suggestions a try too.  Many thanks.  

[nick_name]

On 8/24/2020 at 2:25 PM, havli said:

Perhaps some Windows tweaks, but that is just a wild guess. I never liked tweaking OS.

Roger that.  

[mirrormax]

should add support for more than 64cores/128threads if possible

capped at 64 when overkill is an option is

[havli]

There is no artificial limit to cores/threads. Perhaps the encoder doesn't scale with so many cores.... it is 4 years old after all.

[Leeghoofd]

Version 2.3 will be finally implemented early 2021 Havli due to the imposed HPET.
Finally the Bot will be getting somewhere! 

 

[flanker]

guys, any tip why I can not on fresh OS (Win10) run this benchamrk? I got only first screen and still loading...and program is not fully opened.

[Leeghoofd]

Did you use maxmem ? blocked by win smartscreen ?

[keeph8n]

W11. W790. Enabled hpet and system is crawling trying to open and run this benchmark. Any ideas?

[_mat_]

HPET system-wide enabled + many cores = no performance.

It's all bottlenecked by the slow timer calls.