Wil Initial turbo speedup bootprocess when underclock RP

[MrHifiTunes]

I wonder if you underclock your RP, will initial-turbo setting in the confix.txt boost to max STOCK CPU speed...or will it be limited to the max speed defined in the confix.txt?
If its limited to the max speed defined in config.txt...will it have any effect on the boot time?

[Tim Curtis]

Here's the official documentation.
https://www.raspberrypi.org/documentatio...locking.md

[MrHifiTunes]

Here's the official documentation.
https://www.raspberrypi.org/documentatio...locking.md

Thanks for the link.... I read it before but is't confusing.  They talk about overclocking while I want to underclock my PI4 and reduce noise when I play music.   But I also dont want to loose boot up time because of the lower speedlimit of the CPU.

I dont know who has experience in this area...  

 Sometimes I think if its not better to increase the lowest speed of the  CPU.  This way there is enough power to handle music playing and the CPU doenst need to scale up.  The max speed is still there to boot up.   
Very confusing how to set up things.

Im new in these area...no experience...Its my first RP ever and still waiting for my allo boss  (I heard they has issues with the new batch after corona?)  TRUE?

All help is welcome 

[TheOldPresbyope]

So there are two assertions here:

while I want to underclock my PI4 and reduce noise when I play music.   But I also dont want to loose boot up time because of the lower speedlimit of the CPU.

You'll have to wait for your Boss to arrive to test whether underclocking has any effect on "noise" but you can test today whether underclocking increases boot time. Just grab a stop watch and time how long the RPi4 takes to boot with and without underclocking.


Regards,
Kent

[MrHifiTunes]

So there are two assertions here:

while I want to underclock my PI4 and reduce noise when I play music.   But I also dont want to loose boot up time because of the lower speedlimit of the CPU.

You'll have to wait for your Boss to arrive to test whether underclocking has any effect on "noise" but you can test today whether underclocking increases boot time. Just grab a stop watch and time how long the RPi4 takes to boot with and without underclocking.


Regards,
Kent

Hi Kent,

i did some testing
Reduce arm_freq to 800 didnt have much effect on boot time
adding arm_freq_min=400 and gpu_freq = 250 increased the boot time with 20%  
I assume the GPU changes made the boot time longer

I noticed that RPi4 doestnt have HW acceleration to decode FLAC.  Is this normal? 

[TheOldPresbyope]

I noticed that RPi4 doestnt have HW acceleration to decode FLAC.  Is this normal? 

I'm not sure what you are referring to. I wasn't aware that any CPU or SBC supported the FLAC codec algorithms directly. The RPi's ARM CPU does have a hardware floating point unit.

By design, the FLAC decoding algorithm is fast even on slow machines.

Regards,
Kent

[MrHifiTunes]

I noticed that RPi4 doestnt have HW acceleration to decode FLAC.  Is this normal? 

I'm not sure what you are referring to. I wasn't aware that any CPU or SBC supported the FLAC codec algorithms directly. The RPi's ARM CPU does have a hardware floating point unit.

By design, the FLAC decoding algorithm is fast even on slow machines.

Regards,
Kent

As a newby I have no clue myself but I came across 
https://www.raspberrypi.org/documentatio...cgencmd.md

codec_enabled [type]

Reports whether the specified CODEC type is enabled. Possible options for type are AGIF, FLAC, H263, H264, MJPA, MJPB, MJPG, MPG2, MPG4, MVC0, PCM, THRA, VORB, VP6, VP8, WMV9, WVC1. Those highlighted currently require a paid for licence (see the FAQ for more info), except on the Pi4, where these hardware codecs are disabled in preference to software decoding, which requires no licence. Note that because the H265 HW block on the Raspberry Pi4 is not part of the VideoCore GPU, its status is not accessed via this command.

[TheOldPresbyope]

Huh.

Didn't know that FLAC was included in the list of possibilities. I think the vcgencmd developers just threw in every possibility they could think of.

In any case, here's the output from an RPi3B+

Code:

pi@moode:~ $ vcgencmd codec_enabled flac
flac=disabled

so we haven't "lost" anything going to the RPi4B.

This is a non-issue for me. Every track I've ever ripped is FLAC encoded to compression level 5 and no model RPi I own (Zero, 2B, 3A+, 3B, 3B+, 4B) has had a problem decoding any of them. Like I said, the algorithm is tweaked to favor decoding efficiency.

Regards,
Kent

[MrHifiTunes]

Thanks Kent for verifying.  I agree with you...maybe for video usage the Pi4 need some tweaking

Back to the topic of speeding up the boot time.
I saw that DHCPd service is taking up most of the time when booting.  Do i need that service?  Don't know which services can be disabled.  

Any help is welcome 

pi@moode:~ $ systemd-analyze
Startup finished in 1.025s (kernel) + 19.220s (userspace) = 20.245s
multi-user.target reached after 19.176s in userspace
pi@moode:~ $ systemd-analyze blame
         12.438s dhcpcd.service
          2.136s dev-mmcblk0p2.device
          1.478s rc-local.service
          1.272s nmbd.service
          1.018s php7.3-fpm.service
           874ms nginx.service
           523ms keyboard-setup.service
           502ms rpi-eeprom-update.service
           452ms systemd-timesyncd.service
           424ms networking.service
           415ms systemd-udev-trigger.service
           395ms smbd.service
           390ms systemd-fsck@dev-disk-by\x2dpartuuid-738a4d67\x2d01.service
           382ms winbind.service
           364ms raspi-config.service
           286ms systemd-logind.service
           281ms avahi-daemon.service
           267ms systemd-journald.service
           228ms systemd-fsck-root.service
           219ms user@1000.service
           213ms sysstat.service
           212ms systemd-remount-fs.service
           210ms rsyslog.service

[TheOldPresbyope]

Why does it take seconds to obtain IP address via DHCP?

You can play games assigning static IP in various ways (e.g., client side vs server side) and see if it helps you. Since we can't predict each user's networking environment, moOde shoots for the general case.

I've been down the route of interpreting the systemd-analyze output before. Some services can start up in parallel, some must start sequentially, and it's the overall value which matters. Turned out to be harder to minimize boot time substantially than I had hoped.

Regards,
Kent