Overclocking

[Tavalin]

Since previously (and reproducibly) getting the corrupt storage partition with openelec with overclocking, I've since downloaded the most recent image (r12006) and changed my config.txt to the following and so far I'm yet to experience the corruption.

Code: Select all

arm_freq=1000
disable_overscan=1
over_voltage=6
initial_turbo=30

The only difference this time is that I haven't overclocked the gpu_freq or core_freq. Maybe it's the combination of these that cause the issue?

[dom]

Since previously (and reproducibly) getting the corrupt storage partition with openelec with overclocking, I've since downloaded the most recent image (r12006) and changed my config.txt to the following and so far I'm yet to experience the corruption.

Code: Select all

arm_freq=1000
disable_overscan=1
over_voltage=6
initial_turbo=30

The only difference this time is that I haven't overclocked the gpu_freq or core_freq. Maybe it's the combination of these that cause the issue?

Thanks, that's useful information. When you are sure the corruption is not happening it would be useful if you could start creeping up the core_freq (Ieave gpu_freq out of it).
I'd be ineterested if core_freq=300 was still reliable. If so, then core_freq=400 and core_freq=500.

Can you run
vcgencmd version
to confirm what firmware version you have. Also:
cat /proc/cpuinfo

[dom]

If anyone is interested I seem to be getting config.txt corruption when I attempt overclocking via raspi-config. My SD card is a Samsung Class 10. I'm happy to carry out some testing, but need very step-by-step instructions as I'm a near total n00b in Linux.

Can you follow instructions here:
http://www.raspberrypi.org/phpBB3/viewt ... 75#p182639
Let me know if you get stuck.

[sam nazarko]

Thanks Sam and Dom. Can't wget from the above mentioned folder. wget returns:

ERROR: The certificate of `github.com' is not trusted.
ERROR: The certificate of `github.com' hasn't got a known issuer.

Anyway, I just want to tell that my system has been streaming from the Plex Media Server successfully and I did not have any more SD card corruption. I'm using a Sandisk Extreme Pro Class 10 UHS-I. Dom, FWIW, I've tried lowering and also raising the init_emmc_clock to 35000000 and 70000000, with no SD corruption. I'm using raspberrypi-firmware-b5898de on top of Sam's Raspbcm RC5 with XBMC updated to xbmc-rbp-20120922. I've successfully tried Normal, Fast and Super profiles. Didn't try Ultimate, as I got corruption with that setting before.

wget --no-check-certificate

[Tavalin]

Can you run
vcgencmd version
to confirm what firmware version you have. Also:
cat /proc/cpuinfo

Code: Select all

root /flash # vcgencmd version
Sep 25 2012 00:18:47
Copyright (c) 2012 Broadcom
version 339137 (release)

Code: Select all

root /flash # cat /proc/cpuinfo
Processor       : ARMv6-compatible processor rev 7 (v6l)
BogoMIPS        : 666.41
Features        : swp half thumb fastmult vfp edsp java tls
CPU implementer : 0x41
CPU architecture: 7
CPU variant     : 0x0
CPU part        : 0xb76
CPU revision    : 7

Hardware        : BCM2708
Revision        : 0002

I'll try upping the core_freq tomorrow, too tired to play trial and error tonight!

[Dilligaf]

Tried 800 then 900 then 100 at stock core and sdram, overvoltage 6 with no problems, bumped core to 350 and corrupted partition. Now reimaging card (made a backup first) and will try at 300.

[Dilligaf]

300 killed the partition, reimaged and set core to 275, booted and deleted database, went to reboot and it has already corrupted the partition. So it looks like any core overclock corrupts partitions. Is the core dynamic or static? Wondering what would happen with a static core overclock. Is it the overclock or the changing core that is doing it?

[dom]

300 killed the partition, reimaged and set core to 275, booted and deleted database, went to reboot and it has already corrupted the partition. So it looks like any core overclock corrupts partitions. Is the core dynamic or static? Wondering what would happen with a static core overclock. Is it the overclock or the changing core that is doing it?

Can you try
arm_freq=750
core_freq=300
sdram_freq=400
over_voltage=0

and see if that corrupts.

If it does, then try adding
force_turbo=1
(this won't set warranty bit as long as over_voltage=0)
which will disable the dynamic overclock, and leave it fixed high.
Is that any better?

[cesarvog]

wget --no-check-certificate

Thanks Sam, will try again tomorrow and let you know of the outcome. Let me tell you upfront that so far it has been working flawlessly. Watched a few TV shows tonight with my wife and everything (menus, thumbnail creation, folder enumeration) was almost as smooth as on my AppleTV with Crystalbuntu. File/folder operations are still a tad slower, but everything feels way faster than on RC4. Also, keep in mind that I have no local media on my RPi. Everything is streamed from the Plex Media Server on my Mac Mini Server. In order to acces it, I have installed Hippojay's excellent PleXBMC 2.0b4 add-on for XBMC.

[milhouse]

What I really need is for a number of people who are suffering the problem, to gradually wind down the overclock settings, and hopefully tell me that, e.g. core_freq > 800 causes corruption.
If I got a consensus of what the clock (or voltage) it is that is critical, I can possibly do something about it, but I've had very few useful bug reports.

I assume you weren't able to reproduce problems with 750/255/+3 on a Transcend Class 10 (do you have this card to test)? Note that I also get the same SD card corruption with a Transcend Class 2.

Can anybody else?

Do you have any thoughts on the network activity causing a hang? Again, is anyone else able to reproduce?

SD corruption (which I can avoid by using USB for /root) and hangs due to heavy network usage (which I can't avoid) are now my only overclock-related issues!

[Dilligaf]

300 killed the partition, reimaged and set core to 275, booted and deleted database, went to reboot and it has already corrupted the partition. So it looks like any core overclock corrupts partitions. Is the core dynamic or static? Wondering what would happen with a static core overclock. Is it the overclock or the changing core that is doing it?

Can you try
arm_freq=750
core_freq=300
sdram_freq=400
over_voltage=0

and see if that corrupts.

If it does, then try adding
force_turbo=1
(this won't set warranty bit as long as over_voltage=0)
which will disable the dynamic overclock, and leave it fixed high.
Is that any better?

Couldn't get 750/300 to act up, but at 840/375 with force_turbo=1 it was rock solid, without force_turbo=1 it corrupted the root partition. I'm not worried about the warranty bit, I blew that one a long time ago.

Mike

[milhouse]

All tests performed with latest OpenELEC r12014 build:

Code: Select all

root ~ # vcgencmd version
Sep 25 2012 00:18:47
Copyright (c) 2012 Broadcom
version 339137 (release)

root ~ # uname -a
Linux rpi 3.2.30 #1 PREEMPT Thu Sep 27 03:53:34 CEST 2012 armv6l GNU/Linux

root ~ # cat /proc/cpuinfo (force_turbo=1 in this case)
Processor       : ARMv6-compatible processor rev 7 (v6l)
BogoMIPS        : 749.56
Features        : swp half thumb fastmult vfp edsp java tls
CPU implementer : 0x41
CPU architecture: 7
CPU variant     : 0x0
CPU part        : 0xb76
CPU revision    : 7

Hardware        : BCM2708
Revision        : 1000002

At 750 (arm)/255 (core)/+3 (over_volt) I'm getting Transcend Class 2 AND Class 10 corruption with OR without force_turbo=1.

I didn't need to do anything: booted fresh OpenELEC image, waited about 2-3 minutes (some add-ons updated), e2fsck appeared clean, rebooted -> corrupt ext4 partition each time.

Tried again, this time swapping force_turbo=1 for temp_limit=1, fresh image, and no SD corruption on either Class 2 or Class 10.

WIth temp_limit=1 I'm able to write 500MB to either SD card without any corruption (dd if=/dev/zero of=./test.dat bs=1M count=500).

[Tavalin]

All tests performed with latest OpenELEC r12014 build:

Code: Select all

root ~ # vcgencmd version
Sep 25 2012 00:18:47
Copyright (c) 2012 Broadcom
version 339137 (release)

root ~ # uname -a
Linux rpi 3.2.30 #1 PREEMPT Thu Sep 27 03:53:34 CEST 2012 armv6l GNU/Linux

root ~ # cat /proc/cpuinfo (force_turbo=1 in this case)
Processor       : ARMv6-compatible processor rev 7 (v6l)
BogoMIPS        : 749.56
Features        : swp half thumb fastmult vfp edsp java tls
CPU implementer : 0x41
CPU architecture: 7
CPU variant     : 0x0
CPU part        : 0xb76
CPU revision    : 7

Hardware        : BCM2708
Revision        : 1000002

At 750 (arm)/255 (core)/+3 (over_volt) I'm getting Transcend Class 2 AND Class 10 corruption with OR without force_turbo=1.

I didn't need to do anything: booted fresh OpenELEC image, waited about 2-3 minutes (some add-ons updated), e2fsck appeared clean, rebooted -> corrupt ext4 partition each time.

Tried again, this time swapping force_turbo=1 for temp_limit=1, fresh image, and no SD corruption on either Class 2 or Class 10.

WIth temp_limit=1 I'm able to write 500MB to either SD card without any corruption (dd if=/dev/zero of=./test.dat bs=1M count=500).

If I understand what force_turbo and temp_limit are doing correctly, it would seem that it's the fact that the core_freq has been overclocked, rather than the fact core_freq is dynamic, that is causing corruption?

[kevs3d]

I don't think it's anything todo with dynamic overclocking - i've been using force_turbo since it was added - and i can corrupt my /boot partition on the class 10 card effortlessly when overclocked.

I will try just upping core and then just arm to see if I can isolate it tonight.

Kev

[dom]

I assume you weren't able to reproduce problems with 750/255/+3 on a Transcend Class 10 (do you have this card to test)? Note that I also get the same SD card corruption with a Transcend Class 2.

Does 750/250/+3 corrupt?
Does 750/250/0 corrupt?
Have you seen corruption in a stock raspbian image?

Yesterday I installed latest raspbian on Transcend class 6 8G card. Selected top overclock setting. Rebooted. Exandanded root partition.
Launched X. In one window did apt-get upgrade whilst web browsing. No problem. Installed libreoffice whilst web browsing. No problem.
Serveral reboots. Numerous apps launched. No problem.

Possibly some Pi boards (specifically the 2835 SoC) are not succeptable to this problem.
Possibly some sdcards are not succeptable to this problem.

We need someone who's hitting this problem to either swap Pi boards, or swap sdcards to try and determine which of these is true.

millhouse - I think yor experiences are atypical. Everyone else seems to need a more significant overclock to see problems, so I've a suspicion yours is a different problem.
Do you have any other scdards you can try? Do you know anyone else with a Pi you can test on?

[elatllat]

For me the "moderate" setting on a fresh 2012-09-18-wheezy-raspbian.img causes a
sudo apt-get update; sudo apt-get upgrade to corrupt the file system.

[cesarvog]

Sam and Dom,

I was finally able to run vcgencmd measure_clock emmc from my RC5 and here are the results:
a) with default init_emmc_clock and Super profile: frequency(47)=50000000
b) with default init_dmmc_clock and Ultimate profile: frequency(47)=50000000
c) with init_emmc_clock=70000000 and Super profile: frequency(47)=69984000
d) with init_emmc_clock=70000000 and Ultimate profile: frequency(47)=69984000
e) with init_emmc_clock=35000000 and Super profile: frequency(47)=frequency(47)=34992000
f) with init_emmc_clock=35000000 and Ultimate profile: frequency(47)=34992000

All tests were run on a rev 00002 board, with the latest firmware and initial_turbo=30 set (don't really know if this is still needed with latest fw), on a Kingston 8 GB Class 10 SDHC.

I'm going out of town and will only be able to do further testing next Monday.

[dom]

For me the "moderate" setting on a fresh 2012-09-18-wheezy-raspbian.img causes a
sudo apt-get update; sudo apt-get upgrade to corrupt the file system.

I assume you mean modest:
arm_freq=800
core_freq=300
sdram_freq=400

Can you try again, but first, add
force_turbo=1
to config.txt before rebooting. (that won't set warranty bit if over_voltage=0).

Also can you try (without force_turbo=1):
core_freq=250
before rebooting?

[pmk]

I am confused .........

Ran some tests today and got the following results:

arm_freq=1000, core_freq=500, sdram_freq=500, over_voltage=6 - no problem found
arm_freq=750, core_freq=275, sdram_freq=400, over_voltage=0 - corruption
arm_freq=750, core_freq=250, sdram_freq=400, over_voltage=6 - no problem found

Hope this helps track it down.

[dom]

I am confused .........

Ran some tests today and got the following results:

arm_freq=1000, core_freq=500, sdram_freq=500, over_voltage=6 - no problem found
arm_freq=750, core_freq=275, sdram_freq=400, over_voltage=0 - corruption
arm_freq=750, core_freq=250, sdram_freq=400, over_voltage=6 - no problem found

Hope this helps track it down.

Thanks.

Well the middle one is surprising. Probably there is a low probability of corruption and that test just happened to hit it.
Can you see if this setting reliably corrupts?
arm_freq=750, core_freq=275, sdram_freq=400, over_voltage=0 - corruption

Does it corrupt after doing some set operation, or does it just die eventually?

Is
arm_freq=750, core_freq=250, sdram_freq=400, over_voltage=0
reliable?

[milhouse]

Does 750/250/+3 corrupt?
Does 750/250/0 corrupt?
Have you seen corruption in a stock raspbian image?

Will re-test and let you know.

We need someone who's hitting this problem to either swap Pi boards, or swap sdcards to try and determine which of these is true.

millhouse - I think yor experiences are atypical. Everyone else seems to need a more significant overclock to see problems, so I've a suspicion yours is a different problem.
Do you have any other scdards you can try? Do you know anyone else with a Pi you can test on?

I've managed to locate a SanDisk 4GB SDHC Class 2 and will test it under various conditions and let you know how I get on. Unfortunately I don't know anyone else with a Pi (how sad) - I guess the next option would be to buy another but I hadn't planned on doing that for a while yet.

One thing I've learned from all this is that I need to vary my choice of memory cards in future...

[Tavalin]

I seem to have corrupted the SD card again by running dd with the following config.txt

Code: Select all

arm_freq=1000
disable_overscan=1
over_voltage=6
initial_turbo=30
core_freq=275

The only difference from when it was stable is that I added core_freq=275.

[Dilligaf]

I've corrupted the filesystem on sandisk ultras class 6 and class 10, a MicroCenter class 10 and some no name class 2, 4, and 6 cards. To me the card doesn't seem to matter. The above results were done with a Sandisk Ultra class 10

[elatllat]

For me the "moderate" setting on a fresh 2012-09-18-wheezy-raspbian.img causes a
sudo apt-get update; sudo apt-get upgrade to corrupt the file system.

I assume you mean modest:
arm_freq=800
core_freq=300
sdram_freq=400

Can you try again, but first, add
force_turbo=1
to config.txt before rebooting. (that won't set warranty bit if over_voltage=0).

Also can you try (without force_turbo=1):
core_freq=250
before rebooting?

Yes I did mean modest.
force_turbo=1 caused corruption
core_freq=250 seems to have worked

[dom]

I am confused .........

Ran some tests today and got the following results:

arm_freq=1000, core_freq=500, sdram_freq=500, over_voltage=6 - no problem found
arm_freq=750, core_freq=275, sdram_freq=400, over_voltage=0 - corruption
arm_freq=750, core_freq=250, sdram_freq=400, over_voltage=6 - no problem found

Hope this helps track it down.

core_freq=500 => PLL frequency of 2G
core_freq=275 => PLL frequency of 2.2G
core_freq=250 => PLL frequency of 2G

If it is related to high PLL frequency, then I'd expect
core_freq=300 => PLL frequency of 2.4G to be bad
core_freq=310 => PLL frequency of 1.86G to be good

Can anyone try that?