RipBot264 v1.18.3 - Simple and easy to use GUI -> IPOD . PSP . CONSOLES . BLURAY

[TDS]

Originally Posted by PetBel

Hi all
is there a command line option to set the number of encoding threads for EncodingServer?

Do you mean something like these options :-



Click on the "spanner" icon to get to this menu

[rlev11]

Originally Posted by PetBel

Hi all
is there a command line option to set the number of encoding threads for EncodingServer?

To directly answer your question, YES
encodingserver /restart-if-no-progress /priority normal /affinity 3FFC3FFC

The affinity uses a hex number to enable/disable specific cores. You can use a site like (bitsum.com/tools/cpu-affinity-calculator) to get the hex number you desire.

If you are asking this to knock down say a 16 core ryzen that falls off a cliff when doing 4k encoding, you can use the above affinity switch or I have found that the following command lines works actually a tiny bit better. This also doesn't change regular 1080 and 720 encoding speeds.

encodingserver /priority normal /restart-if-no-progress /avisynth-prefetch-threads 12 /x264-threads 16 /x265-threads 16

The avisynth-prefetch-threads is what actually kills the 16 core ryzens when doing 4k encoding for whatever reason if you do not knock them down to 12

This works best regardless of what you are encoding and is what I use on all my 16 core ryzens

[Ryushin]

Originally Posted by apol847

Yes, that's true I could setup a dual boot system. But this did not solve the initial problem!
Now it is going slightly off-topic:

[off-topic] People complain: I can't use Linux! There is no e.g. Photoshop available. I still need to use Windows. If there was a Linux Photoshop version I could switch to Linux. Adobe says: There is no need of a Photoshop Linux version because only 1 or 2% of the users are using Linux, development for Linux is not worth it. It's like the hen and egg problem. I switched to Linux 3 years ago privately. First with dual boot option Windows/Linux and since last year to Linux only as main OS. I am happy with it and I found replacement apps for my used Windows applications. But there are a few apps where no Linux version is available. There is a little video toolset I want to use with Linux. Most binaries in this toolset are also available for Linux. The toolset uses DetectBorders.exe and some Windows Batch Scripts to call the Windows binaries. I need a native Linux binary so I can use DetectBorders from the RipBot project with some shell scripts on Linux. So I will take the chance and try to port the DetectBorders Windows Tool to Linux. It's a start and more Windows only applications will be available for additional platforms Some people here in the forums said earlier that they can't use the toolset on their Linux systems because it's Windows only at the moment. I think they would also be happy if there is a Linux version available later. [/off-topic]

Sorry, for the long explanation and off-topic but I had a reason to ask for a native Linux version... If not, I wouldn't have asked for it...

Kudos for attempting to port it. I've been driving Linux as my main desktop since '97. Been using the same Debian install from Potato until today on my Thinkpad P17 Gen2. My main RB install is running in a VM on my Linux Server. I use Linux for just about everything and for the most part there is *nix software for everything I want to do. I also have a soft spot for FreeBSD. I also know Solaris quite well, though I don't care for it I has to dump half of GNU on it to make it usable. LOL

All my physical servers run ZFS as a file system including my laptop (has three M.2 NVMEs). I even take care of large installs of ZFS over 2PB with 232 hard drives and 100TB of NVME Cache (L2ARC). ZFS is almost unbelievable in what it can do.

So yea, I belong to the cult of the Penguin.

[TDS]

I stumbled across something today, and for those of you that are lucky enough to own one on the AWESOME CPU's.

Add this command to your x265 profile:-

--asm avx512

Too bad for you 13th Gen Intel owners

[JThomasG]

Hi,

quick question, im using an intel i9-12900KF,

will i see a performance benefit if i use the prefetch option below.

looking at cpu performance all cores seems to works at mostly 100%

Originally Posted by rlev11

To directly answer your question, YES
encodingserver /restart-if-no-progress /priority normal /affinity 3FFC3FFC

The affinity uses a hex number to enable/disable specific cores. You can use a site like (bitsum.com/tools/cpu-affinity-calculator) to get the hex number you desire.

If you are asking this to knock down say a 16 core ryzen that falls off a cliff when doing 4k encoding, you can use the above affinity switch or I have found that the following command lines works actually a tiny bit better. This also doesn't change regular 1080 and 720 encoding speeds.

encodingserver /priority normal /restart-if-no-progress /avisynth-prefetch-threads 12 /x264-threads 16 /x265-threads 16

The avisynth-prefetch-threads is what actually kills the 16 core ryzens when doing 4k encoding for whatever reason if you do not knock them down to 12

This works best regardless of what you are encoding and is
what I use on all my 16 core ryzens

[Ryushin]

Originally Posted by TDS

I stumbled across something today, and for those of you that are lucky enough to own one on the AWESOME CPU's.

Add this command to your x265 profile:-

--asm avx512

Too bad for you 13th Gen Intel owners

Now I'll have to do more testing.
Found this:
https://www.hwcooling.net/en/intel-a...-does-it-help/

Not sure if the power consumption is going to be worth slightly increased performance. AMDs implementation is double pumped compared to Intels 12th gen. I'm running a test now against a entire 4K movie (Puss in Boots) so I'll report back with the times to encode both with or without.

Unfortunately, I won't be able to use it for my set up as one of my two servers does not have AVX-512 and the X265 options is added to the main x265 settings. But perhaps x265 will just ignore the option if it's not available to use. I'll test that too.

[Ryushin]

Originally Posted by JThomasG

Hi,

quick question, im using an intel i9-12900KF,

will i see a performance benefit if i use the prefetch option below.

looking at cpu performance all cores seems to works at mostly 100%

Those are the results of the tests I ran for the 7950x which has 16 cores. Even though you will see 100% usage, it does not mean it is encoding as it's best efficiency. You might have to make a sample and run some tests. What I found is that you will want to set your avisynth-prefetch-threads to about 3/4 of that your x265 threads would be. Otherwise what I noticed is you will have threads waiting for other threads to be finished before they can continue.

What I've discovered is that you will want to see is your CPU usage bouncing between 85-100%.

So leave off the "/x264-threads 16 /x265-threads 16" and see what your Encoding Server shows as the default threads. Then just set the avisynth-prefetch-threads to about 3/4 of what it shows.

[rlev11]

The only time in my experience that you have to worry about knocking down cpu cores or avisynth threads is when you start getting into doing 4k encoding and specifically 4k encodes that are full screen aspect ratio (1.78:1 or even 1.85:1). There is something with encoding all those extra bits that I have found that the 16 core Ryzens really have an issue with those. I can't speak to the latest Intel i-9's as I don't have any of those. Not sure how the whole P and E cores effect encoding 4k, you may have to just do some testing.

When doing those 4k encodes with everything set full out on a 3950/5950/7950, The encoding speed takes a nose-dive to point where they are about 1/2 the speed of a 12 core 3900x or 5900x. Something gets overloaded with all those threads and they just can't handle it. This is where using either affinity mask or avi-synth thread setting knocking them down to only use 12 cores or threads is necessary. Even with cores or threads turned off, they will still out-perform their 12 core brothers.

I posted some quick tests a while ago that showed me that once I set the avisynth prefetch to 12, that it did not slow down or speed up doing regular 1080 or 720 encoding, but made the 4k stuff run as they should which is why I just use that setting now for everything.

[TDS]

Originally Posted by Ryushin

Now I'll have to do more testing.
Found this:
https://www.hwcooling.net/en/intel-a...-does-it-help/

I found that site, yesterday :

Unfortunately, this has old info, and it only shows 11th Gen Intel & 5000 Series Ryzen's....it clearly shows that the Ryzen's DO NOT support avx 512 but the the Intel chips do, with a small increase in speed.

However, 12th & 13th Gen Intel chips with their "E" & "P" cores do not. although it has been suggested that with some motherboards, there is an option to enable it (I haven't got a motherboard that has that), also it has been suggested that turning off the "E" cores with enable avx 512 support, again, I haven't got that option, either

Not sure if the power consumption is going to be worth slightly increased performance. AMDs implementation is double pumped compared to Intels 12th gen. I'm running a test now against a entire 4K movie (Puss in Boots) so I'll report back with the times to encode both with or without.

Unfortunately, I won't be able to use it for my set up as one of my two servers does not have AVX-512 and the X265 options is added to the main x265 settings. But perhaps x265 will just ignore the option if it's not available to use. I'll test that too.

On CPU's that don't support avx 512, the x265 encoder will simply show a message...[warning]: AVX512 is not supported, and then continue's.

So I did a lot of frustrating testing yesterday, and there doesn't appear to be that much benefit, but it's nice be able to enable it, and it's just another feather in AMD's cap.

TBBH, I'm not all that impressed with the 13900KF, it run's as hot as a furnace, and is slower than the 7950X !!!!! (which is also a "cooler" furnace)
I am running a "budget" B660 MSI motherboard, but the "grunt" of the 13900KF should still be fully enabled, I think spending another couple of $100's wouldn't make much difference, and overclocking would just create a LOT more unwanted heat !!!

I will be interested in your test results

[TDS]

Originally Posted by JThomasG

Hi,

quick question, im using an intel i9-12900KF,

will i see a performance benefit if i use the prefetch option below.

looking at cpu performance all cores seems to works at mostly 100%

Hi JThomasG,

In my limited experience with my new 13900KF, this prefetch option doesn't seem to be as "important" with the Intel chips, as it is with the 16 core Ryzen's.

I did some tests using 3 different command line's, and there was little difference,

encodingserver /priority normal /restart-if-no-progress /avisynth-prefetch-threads 12 /x264-threads 16 /x265-threads 16

encodingserver /priority normal /restart-if-no-progress /avisynth-prefetch-threads 14 /x264-threads 16 /x265-threads 16

encodingserver /priority normal /restart-if-no-progress /avisynth-prefetch-threads 16 /x264-threads 16 /x265-threads 16

However, I did notice that if these commands are not included, the speed drops noticeably

As rlev11 mentioned, it does also depend on the "size" of the video being encoded, and of course what filters are being used.

specifically 4k encodes that are full screen aspect ratio (1.78:1 or even 1.85:1)

Good luck..

[LigH]

Originally Posted by TDS

Add this command to your x265 profile:-

--asm avx512

Guys, there is a good reason why this option has to be enabled manually and is not enabled when its CPU support is autodetected:

There is no certain advantage in it. Not as much between AVX2 and AVX512 as there is between AVX and AVX2.

It may execute some calculations faster in net CPU time. But gross CPU time can still be worse, because switching between a more basic CPU mode and the AVX512 context requires clearing more CPU registers which must be loaded again, and the power consumption is so much higher that the temperature rises much faster so that the CPU must be cooled with more efforts, up to throttling down the CPU clock. This may also worsen the durability of the whole CPU.

I don't want to discourage you from the start. Yes, try it. But take all the side effects into account.

[TDS]

Originally Posted by LigH

Guys, there is a good reason why this option has to be enabled manually and is not enabled when its CPU support is autodetected:

There is no certain advantage in it. Not as much between AVX2 and AVX512 as there is between AVX and AVX2.

It may execute some calculations faster in net CPU time. But gross CPU time can still be worse, because switching between a more basic CPU mode and the AVX512 context requires clearing more CPU registers which must be loaded again, and the power consumption is so much higher that the temperature rises much faster so that the CPU must be cooled with more efforts, up to throttling down the CPU clock. This may also worsen the durability of the whole CPU.

I don't want to discourage you from the start. Yes, try it. But take all the side effects into account.

https://forum.doom9.org/showthread.p...75#post1981275

[Ryushin]

Originally Posted by Ryushin

Now I'll have to do more testing.

Not sure if the power consumption is going to be worth slightly increased performance. AMDs implementation is double pumped compared to Intels 12th gen. I'm running a test now against a entire 4K movie (Puss in Boots) so I'll report back with the times to encode both with or without.

So I finished testing the AVX-512 testing and I also tested against underclocking the 7950x.

What I found, as others have found, there was no improvement in enabling AVX-512. In fact it was ever so slightly slower when using AVX-512.

Underclocking the 7950x resulted in a nice discovery. I have my 7950x underlocked to all cores at 4900MHz using 1.08 volts.
So the 7950x at the default frequencies took 1:24m:28s to run the encode while averaging 210 watts of energy. The 7950x underclocked took 1h:26m:58s while averaging 140 watts of energy. So underclocked it was 2.3% slower than the default frequencies and the 7950x uses 50% more electricity at the default frequencies. Pretty amazing how well the 7950x performs when underclocked slightly. So yea, the 7950x is not so much a furnace if you underclock it.


Testing Detail:
AVX-512 X265 Instruction set test. AMD 7950x with default frequencies.
Source: Puss In Boots 4K - 1h:30m:35s
AIO 420mm Cooler
CQ16

Without AVX-512:
Time to encode: 1:24m:28s
Avg Temp: 85 Celsius
Avg Power Package Watts: 210
Avg Power Cores Watts: 145

With AVX-512:
Tiem to Encode: 1h:25m:44s
Avg Temp: 89 Celsius

Underclocked to 4.9GHz (1.08v)
Without AVX-512
Time to encode: 1h:26m:58s
Avg Temp: 60 Celsius
Avg Power Package Watts: 140
Avg Power Cores Watts: 85

[rlev11]

Those are great numbers on under-clocking the 7950x. 2.3% performance drop is basically nothing for the huge power savings. This is something I am going to have to seriously consider.

I will probably also look into and see if doing the same with the 5950x results in a similar small performance hit with a big power savings as well.

Just did a quick test with my 7950x. All I did was go into Ryzen Master and in the Basic View changed the profile from Default to Eco Mode. This knocks the 7950x down to 65watt maximum. My CPU temp went from 85c to 63c. CPU Power via Ryzen Master showed a drop from 140w down to appx 55watt. Clock Speeds dropped about a 1,000 Mhz. FPS showed in the Encoding Client in default mode was about 26-27 fps on a 1080p encode. After ECO Mode the fps only dropped 1.5 to 3fps. next up will put a Kill-a-watt meter on at the wall and see what real world power drop is.

Looks like on the 7950x you can change the mode on the fly in Ryzen Master. No reboot required, just click ECO Mode and hit apply.

[TDS]

Originally Posted by Ryushin

So I finished testing the AVX-512 testing and I also tested against underclocking the 7950x.

Well, first of all, maybe I should offer a bit of an apology for introducing the AVX512 option, I guess I was "sucked in" by the hype that AVX2 & AVX512 was a bit of a game changer, but it's clearly hyped BS !!

I have also tried supposed AVX2 optimised x265's, and there's no significant improvement.

It might be advantageous in some scenarios, but not this.

It's also a shame that some of us spend our hard-earned money on "top of the line" hardware (eg:- 7950X's, and 13900K's), only to find they use SO much power, and produce SO much heat, that "we" go to great lengths to "hobble" them, to make them more acceptable...it's almost pointless, really.

But having said that, is underclocking the way to go, or is undervolting & overclocking the new thing ??

I was watching a YT clip of JayzTwoCents where he undervolted an Intel CPU, and was also able to overclock it, so more speed, less volts, less power use, less heat.

https://www.youtube.com/watch?v=4jjopjkJzxA

And a couple of other clips on undervolting 7950X's..

https://www.youtube.com/watch?v=1pizvaYiVbk

https://www.youtube.com/watch?v=7JiYAwKIHRY

And now back to your tests:-

Those encoding time's for "Puss In Boot's", that's encoding the movie in less time that it takes to watch it !!!

Are you using any filters, or is it only CQ 16 ??

And as for power usage, that fortunately isn't that much of an issue for me, as I only encode when the Solar Panels are generating enough, and I don't do any "overnight" encoding, been there, done that, was very disappointing, most of the time

[TDS]

Originally Posted by rlev11

Those are great numbers on under-clocking the 7950x. 2.3% performance drop is basically nothing for the huge power savings. This is something I am going to have to seriously consider.

I will probably also look into and see if doing the same with the 5950x results in a similar small performance hit with a big power savings as well.

Just did a quick test with my 7950x. All I did was go into Ryzen Master and in the Basic View changed the profile from Default to Eco Mode. This knocks the 7950x down to 65watt maximum. My CPU temp went from 85c to 63c. CPU Power via Ryzen Master showed a drop from 140w down to appx 55watt. Clock Speeds dropped about a 1,000 Mhz. FPS showed in the Encoding Client in default mode was about 26-27 fps on a 1080p encode. After ECO Mode the fps only dropped 1.5 to 3fps. next up will put a Kill-a-watt meter on at the wall and see what real world power drop is.

Looks like on the 7950x you can change the mode on the fly in Ryzen Master. No reboot required, just click ECO Mode and hit apply.

That's all very interesting as well, I think I've tried a few of these settings, but it's just killing the powerhouse that the 7950X is.

You'll probably see a much bigger decrease when encoding 4K, with strong filters

I've gone back to some pretty basic BIOS settings, a small negative curse optimiser, and left almost everything else as default, it's running at approx 95- 96°C, on a hot day (but it doesn't seem to matter what the ambient is), and it's around 5100 - 5400MHz, when encoding 4K.

[rlev11]

Yea I'm not completely sold on ECO Mode. It was just a "Let's see what happens" and I was really surprised that the performance was as good as it was. It something I might enable on the fly if I que up a bunch of stuff to do overnight and don't really care when it gets done.

[Ryushin]

Originally Posted by TDS

It's also a shame that some of us spend our hard-earned money on "top of the line" hardware (eg:- 7950X's, and 13900K's), only to find they use SO much power, and produce SO much heat, that "we" go to great lengths to "hobble" them, to make them more acceptable...it's almost pointless, really.

Those encoding time's for "Puss In Boot's", that's encoding the movie in less time that it takes to watch it !!!

Are you using any filters, or is it only CQ 16 ??

And as for power usage, that fortunately isn't that much of an issue for me, as I only encode when the Solar Panels are generating enough, and I don't do any "overnight" encoding, been there, done that, was very disappointing, most of the time

Not using any filters for those tests, just CQ16. I can run the same tests but I think the results will be similar.

I think both AMD and Intel are chasing every last bit of performance they can to one up each other. The amount of heat/watts consumed by the new Intel 13900K is staggering (yea, that is a furnace):
https://www.youtube.com/watch?v=P40gp_DJk5E&t=364s
The 13900K seems very bursty in the way it was designed and it throttles back in only a few seconds. The 7950x seems to go to it's limit and hold there. I'm curious, since you have both CPUs, you should run a test and let us know which CPU wins (regardless of power used). I think you'll need to disable the GPU portion though so it runs all CPU.

So even though the chips can eek out more performance by pushing them to their max, I don't think for real world use it's necessary. Even for us that torture the CPUs for encoding, underlocking them ever so slightly produces great performance. And if I want to get 2.3% more performance for 50% more power I can.

[TDS]

Originally Posted by Ryushin

Not using any filters for those tests, just CQ16. I can run the same tests but I think the results will be similar.

I think both AMD and Intel are chasing every last bit of performance they can to one up each other. The amount of heat/watts consumed by the new Intel 13900K is staggering (yea, that is a furnace):
https://www.youtube.com/watch?v=P40gp_DJk5E&t=364s
The 13900K seems very bursty in the way it was designed and it throttles back in only a few seconds. The 7950x seems to go to it's limit and hold there. I'm curious, since you have both CPUs, you should run a test and let us know which CPU wins (regardless of power used). I think you'll need to disable the GPU portion though so it runs all CPU.

So even though the chips can eek out more performance by pushing them to their max, I don't think for real world use it's necessary. Even for us that torture the CPUs for encoding, underlocking them ever so slightly produces great performance. And if I want to get 2.3% more performance for 50% more power I can.

OK, well those times make more sense, now that you've confirmed your settings.

I wonder if when using a filter, you might get different results ??

The 13900KF, is the non GPU variant

I'm going to try an older x265 build that's supposed to be optimised for AMD & Intel chips...not sure if it's going to yield better results.

And you're correct, that the Ryzen seems to get to it's limit, and stay's there, and the 13900 does indeed throttle back and forth, sadly.

But the Ryzen has the advantage, most definitely.

[Atak_Snajpera]

Originally Posted by rlev11

The only time in my experience that you have to worry about knocking down cpu cores or avisynth threads is when you start getting into doing 4k encoding and specifically 4k encodes that are full screen aspect ratio (1.78:1 or even 1.85:1). There is something with encoding all those extra bits that I have found that the 16 core Ryzens really have an issue with those. I can't speak to the latest Intel i-9's as I don't have any of those. Not sure how the whole P and E cores effect encoding 4k, you may have to just do some testing.

When doing those 4k encodes with everything set full out on a 3950/5950/7950, The encoding speed takes a nose-dive to point where they are about 1/2 the speed of a 12 core 3900x or 5900x. Something gets overloaded with all those threads and they just can't handle it. This is where using either affinity mask or avi-synth thread setting knocking them down to only use 12 cores or threads is necessary. Even with cores or threads turned off, they will still out-perform their 12 core brothers.

I posted some quick tests a while ago that showed me that once I set the avisynth prefetch to 12, that it did not slow down or speed up doing regular 1080 or 720 encoding, but made the 4k stuff run as they should which is why I just use that setting now for everything.

I wonder what happens if you downgrade your 7950 to 7700 by disabling one CCD in BIOS. Will you still experience the same problem with 4k working terrible with 8 threads (or 16)?