dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5397
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Overclocking

Mon Oct 29, 2012 12:16 am

Dilligaf wrote:Only overvolt of 6 or less.
Actually, over_voltage=8 doesn't set the warranty but it will get limited to 6.
force_turbo=1 does set the warranty bit, and will not limit the over_voltage setting.

User avatar
wallarug
Posts: 460
Joined: Mon May 14, 2012 8:21 am
Location: Sydney, Australia

Re: Overclocking

Mon Oct 29, 2012 1:46 am

dom wrote:
Dilligaf wrote:Only overvolt of 6 or less.
Actually, over_voltage=8 doesn't set the warranty but it will get limited to 8.
force_turbo=1 does set the warranty bit, and will not limit the over_voltage setting.
Does this hold true for the RAM?

shalo
Posts: 74
Joined: Tue May 08, 2012 7:25 pm

Re: Overclocking

Mon Oct 29, 2012 11:20 am

dom wrote:
Dilligaf wrote:Only overvolt of 6 or less.
Actually, over_voltage=8 doesn't set the warranty but it will get limited to 8.
force_turbo=1 does set the warranty bit, and will not limit the over_voltage setting.
I think that's a typo and should read, "but will get limited to 6".
dom wrote:Yes, without force_turbo (or current_limit_override) over_volt is limited to 6. You can specificy 7 or 8, but you will get 6.

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5397
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Overclocking

Mon Oct 29, 2012 11:23 am

shalo wrote:I think that's a typo and should read, "but will get limited to 6".
Thanks. I've corrected my post.

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5397
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Overclocking

Mon Oct 29, 2012 11:25 am

wallarug wrote:
dom wrote:
Dilligaf wrote:Only overvolt of 6 or less.
Actually, over_voltage=8 doesn't set the warranty but it will get limited to 8.
force_turbo=1 does set the warranty bit, and will not limit the over_voltage setting.
Does this hold true for the RAM?
over_voltage_sdram doesn't get limited, and won't set warranty bit unless force_turbo is set.

User avatar
wallarug
Posts: 460
Joined: Mon May 14, 2012 8:21 am
Location: Sydney, Australia

Re: Overclocking

Mon Oct 29, 2012 11:33 am

dom wrote: Does this hold true for the RAM?
over_voltage_sdram doesn't get limited, and won't set warranty bit unless force_turbo is set.[/quote]

thanks Dom. Will experiment with these settings minus force_turbo later.

User avatar
wallarug
Posts: 460
Joined: Mon May 14, 2012 8:21 am
Location: Sydney, Australia

Re: Overclocking

Wed Oct 31, 2012 1:20 am

The wiki says:
Warning: Warranty will be voided if:
(force_turbo || current_limit_override || temp_limit>85) && over_voltage>0
What is the over_voltage in reference to? The RAM, CPU or GPU? Which voids warranty?

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5397
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Overclocking

Wed Oct 31, 2012 11:48 am

wallarug wrote:What is the over_voltage in reference to? The RAM, CPU or GPU? Which voids warranty?
Either can set warranty bit, if used in conjunction with force_turbo or current_limit_override.

User avatar
wallarug
Posts: 460
Joined: Mon May 14, 2012 8:21 am
Location: Sydney, Australia

Re: Overclocking

Wed Oct 31, 2012 9:19 pm

dom wrote:
wallarug wrote:What is the over_voltage in reference to? The RAM, CPU or GPU? Which voids warranty?
Either can set warranty bit, if used in conjunction with force_turbo or current_limit_override.
overvolting settings by them selves will not set the warranty bit then? (unless you want to go outside the range >6.)

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5397
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Overclocking

Wed Oct 31, 2012 9:23 pm

wallarug wrote:overvolting settings by them selves will not set the warranty bit then? (unless you want to go outside the range >6.)
overvolting on its own won't set warranty bit (even if you try to go up to 8).
Warning: Warranty will be voided if:
(force_turbo || current_limit_override || temp_limit>85) && over_voltage>0

User avatar
wallarug
Posts: 460
Joined: Mon May 14, 2012 8:21 am
Location: Sydney, Australia

Re: Overclocking

Wed Oct 31, 2012 9:45 pm

dom wrote:
wallarug wrote:overvolting settings by them selves will not set the warranty bit then? (unless you want to go outside the range >6.)
overvolting on its own won't set warranty bit (even if you try to go up to 8).
Warning: Warranty will be voided if:
(force_turbo || current_limit_override || temp_limit>85) && over_voltage>0
Thanks dom. I understand now.

thogue
Posts: 131
Joined: Wed Sep 19, 2012 2:16 am

Overclocking causes SD card to slow down.

Wed Oct 31, 2012 11:12 pm

I searched around and didn't see anything about this.

Overclocking causes SD card to slow down. I tested with 3 different SD cards. One of which was a brand new 8GB class 10, Once I pass 900/2overvolt setting I get dramatically slower sc card speeds with a DD test. (read and write slower) I tested 2 more sd cards, 1 of which also had the same problem, another class 10 16GB card (Wintec). However, my 16GB class4 has the same read/write speeds after overclock, as would be expected.

Don't know if anyone else has witnessed this or not. but it seems to help me dictate which cards are going to have a problem with the overclock.

captn
Posts: 8
Joined: Thu Nov 01, 2012 11:07 am

Re: Overclocking causes SD card to slow down.

Thu Nov 01, 2012 2:48 pm

thogue wrote:I searched around and didn't see anything about this.

Overclocking causes SD card to slow down. I tested with 3 different SD cards. One of which was a brand new 8GB class 10, Once I pass 900/2overvolt setting I get dramatically slower sc card speeds with a DD test. (read and write slower) I tested 2 more sd cards, 1 of which also had the same problem, another class 10 16GB card (Wintec). However, my 16GB class4 has the same read/write speeds after overclock, as would be expected.

Don't know if anyone else has witnessed this or not. but it seems to help me dictate which cards are going to have a problem with the overclock.
Are you sure, you have the latest kernel modules and /boot blobs?
$ uname -a
Linux raspberrypi 3.2.27+ #250 PREEMPT Thu Oct 18 19:03:02 BST 2012 armv6l GNU/Linux
$ vcgencmd version
Oct 29 2012 23:14:38
Copyright (c) 2012 Broadcom
version 346980 (release)

Have you checked that the TP1 - TP2 voltage is more than 4.7 V ?

I tested several SD cards and RPis and I don't see any significant throughput decrease.
Can you try with my (100% stable) values?

cat /boot/config.txt
arm_freq=950
core_freq=500
sdram_freq=500
over_voltage=2
gpu_mem_256=64
gpu_mem_512=128

/dev/mmcblk0 is a ScanDisk ultra cl10 8GB (30 MB/s)
/dev/sda is a 4GB USB stick

Code: Select all

$ ./iozone -e -I -a -s 50M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2 -f tmpfile.bin
                                                            random  random
              KB  reclen   write rewrite    read    reread    read   write
           51200       4    1119    1257     4241     4174    4057     714
           51200     512   17503   19361    21578    21530   21607    3513
           51200   16384   18971   20470    22275    22294   22275   20750
$ sudo hdparm -tT --direct /dev/mmcblk0
/dev/mmcblk0:
Timing O_DIRECT cached reads: 44 MB in 2.08 seconds = 21.11 MB/sec
Timing O_DIRECT disk reads: 64 MB in 3.03 seconds = 21.11 MB/sec

$ sudo hdparm -tT --direct /dev/sda
/dev/sda:
Timing O_DIRECT cached reads: 24 MB in 2.10 seconds = 11.45 MB/sec
Timing O_DIRECT disk reads: 38 MB in 3.08 seconds = 12.33 MB/sec

[/color]----
$ cat /boot/config.txt
#arm_freq=950
#core_freq=500
#sdram_freq=500
#over_voltage=2
gpu_mem_256=64
gpu_mem_512=128

Code: Select all

$ ./iozone -e -I -a -s 50M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2 -f tmpfile.bin
                                                            random  random
              KB  reclen   write rewrite    read    reread    read   write
           51200       4    1128    1197     4334     4312    3738     741
           51200     512   17920   19959    21586    21692   21583    3645
           51200   16384   20356   20693    22281    22262   22269   20842
$ sudo hdparm -tT --direct /dev/mmcblk0
/dev/mmcblk0:
Timing O_DIRECT cached reads: 44 MB in 2.07 seconds = 21.29 MB/sec
Timing O_DIRECT disk reads: 64 MB in 3.01 seconds = 21.26 MB/sec

$ sudo hdparm -tT --direct /dev/sda
/dev/sda:
Timing O_DIRECT cached reads: 24 MB in 2.04 seconds = 11.76 MB/sec
Timing O_DIRECT disk reads: 38 MB in 3.03 seconds = 12.56 MB/sec

thogue
Posts: 131
Joined: Wed Sep 19, 2012 2:16 am

Re: Overclocking

Thu Nov 01, 2012 8:41 pm

Sorry If I wasnt clear, but the slowdown occurs once I pass 900/overvolt2, Speeds are good at 900/overvolt 2, but 950 overvolt6 1000overvolt6, cause the slowness, im 99.999999 percent positive that I was on the same version as you. But I can always try again, id I would be happy to try your values. Ill report back as soon as I have some personal time to test.

captn
Posts: 8
Joined: Thu Nov 01, 2012 11:07 am

Re: Overclocking

Thu Nov 01, 2012 9:09 pm

thogue wrote:Sorry If I wasnt clear, but the slowdown occurs once I pass 900/overvolt2, Speeds are good at 900/overvolt 2, but 950 overvolt6 1000overvolt6, cause the slowness, im 99.999999 percent positive that I was on the same version as you. But I can always try again, id I would be happy to try your values. Ill report back as soon as I have some personal time to test.
We've got a couple of 500MB RPis for a school project, so I could run tests with them.
Looks like:
for an ARM CPU > 950Mhz the quake3 crashes,
for a SDRAM or core > 500Mhz a 1080p video sometimes freezes the RPi and
for a core voltage > 1.25V copying a large data (2GB mpeg4) over network produces sometimes a garbage.

Below those values, everything seems to work perfect (except those known USB bugs).
So, actually, I don't see any reason to set the overvolt to more than 2 :geek:

thogue
Posts: 131
Joined: Wed Sep 19, 2012 2:16 am

Re: Overclocking

Thu Nov 01, 2012 11:59 pm

/dev/mmcblk0:
Timing O_DIRECT cached reads: 40 MB in 2.05 seconds = 19.50 MB/sec
Timing O_DIRECT disk reads: 62 MB in 3.01 seconds = 20.61 MB/sec

Linux raspberrypi 3.2.27+ #250 PREEMPT Thu Oct 18 19:03:02 BST 2012 armv6l GNU/Linux

Running your stable settings. :ugeek: Voltage has always been good.

thogue
Posts: 131
Joined: Wed Sep 19, 2012 2:16 am

Re: Overclocking

Fri Nov 02, 2012 12:15 am

Just formatted another card, and running another test.

root@raspberrypi:~# hdparm -tT --direct /dev/mmcblk0

/dev/mmcblk0:
Timing O_DIRECT cached reads: 44 MB in 2.07 seconds = 21.29 MB/sec
Timing O_DIRECT disk reads: 64 MB in 3.01 seconds = 21.27 MB/sec
root@raspberrypi:~#

^ no overclock,

/dev/mmcblk0:
Timing O_DIRECT cached reads: 44 MB in 2.06 seconds = 21.31 MB/sec
Timing O_DIRECT disk reads: 64 MB in 3.01 seconds = 21.27 MB/sec

^ stable settings provided
/dev/mmcblk0:
Timing O_DIRECT cached reads: 44 MB in 2.06 seconds = 21.31 MB/sec
Timing O_DIRECT disk reads: 64 MB in 3.01 seconds = 21.27 MB/sec
26112000 bytes (26 MB) copied, 0.533387 s, 49.0 MB/s
root@raspberrypi:~# /home/pi/diskbench.sh
1024+0 records in
1024+0 records out
524288000 bytes (524 MB) copied, 23.1451 s, 22.7 MB/s
1024+0 records in
1024+0 records out
524288000 bytes (524 MB) copied, 26.2869 s, 19.9 MB/s
^TURBO! (Overvolt 6 etc)

After posting last night, I noticed some corrupted files on the 8GB class10 card, so I wiped it. Seemed to of cleared up that issue :-/ Maybe I spoke too soon.

Jon Severinsson
Posts: 6
Joined: Mon Nov 05, 2012 7:50 am

Re: Overclocking

Mon Nov 05, 2012 8:06 am

I'm currently running my Pi with the following settings:

Code: Select all

over_voltage_min=-1
over_voltage=2
arm_freq_min=400
arm_freq=1000
core_freq_min=250
core_freq=500
gpu_freq=250 
over_voltage_sdram=0
sdram_freq_min=400 
sdram_freq=500
This seems stable, both when idling (stress tested by untaring kdelibs) and when busy (stress tested by building kdelibs and simultaneously untaring kde-workspace). I have yet to stress test the GPU.

I'm currently wondering a couple of things:
1. Does anyone see anything wrong with these figures?
2. Does the cpu governor scale freely between 400 and 1000 MHz, or is it all-or-nothing? (I have yet to spot this config at anything but 400 or 1000, but I think I did spot 800 once with arm_freq_min=700 arm_freq=900)
3. If it does scale freely, what happens with the voltage when in between? Eg. will it try to run at 700 MHz with over_voltage=-1, because that wasn't stable when set explicitly :(

milhouse
Posts: 642
Joined: Mon Jan 16, 2012 12:59 pm

Re: Overclocking

Mon Nov 05, 2012 2:18 pm

1. What benefit is there to using the min frequencies/voltages? Are you running on batteries? If not, what is the point?

2. Frequencies will be either min or "max", there will be no in between.

Jon Severinsson
Posts: 6
Joined: Mon Nov 05, 2012 7:50 am

Re: Overclocking

Mon Nov 05, 2012 2:25 pm

Jon Severinsson wrote:This seems stable, both when idling (stress tested by untaring kdelibs) and when busy (stress tested by building kdelibs and simultaneously untaring kde-workspace).
That was unfortunately a bit premature, just got a corrupted SD-card after some additional compiling :cry:. Trying again without core_freq this time, hoping for a better result :roll:.
milhouse wrote:1. What benefit is there to using the min frequencies/voltages? Are you running on batteries? If not, what is the point?
Currently none, but if all work out I'm going to run it in hot ambient temperature (35-40°C) and I'm hoping the lower idle voltage will produce less heat, thus giving a larger headroom for when I need burst speed.
milhouse wrote:2. Frequencies will be either min or "max", there will be no in between.
Since I posted my last post I have spotted both 600 MHz and 900MHz being reported by atop, so either it or you are obviously in error, does anyone else know which? :twisted:.

Cael
Posts: 93
Joined: Sat Jul 07, 2012 7:08 am

Re: Overclocking

Sat Nov 10, 2012 11:04 am

my Pi has Hynix Ram as well seems to OC ok but 450 is absolute max before it plain ies (not kernel panic but plain out Whitescreen if i set to 600 i get a green screen instead of rainbow and it sits there (Crash)

Model is E2212RSV1.0B1.1
Hynix chip shows:

h91knnn2g0mp
(under that) lrnom 216a
(at bottom) cant make it out w/o a magnifying glass it hurts my eyes too much to strain -Focus on such lil text)

carlosfm
Posts: 132
Joined: Fri Oct 21, 2011 3:23 pm
Location: Lisbon, Portugal

Re: Overclocking

Sat Nov 10, 2012 1:09 pm

IME the turbo mode is too much for us with Hynix RAM chips.
600Mhz is way too much, 500Mhz would work for everybody.

Why not create two turbo modes: Turbo and BiTurbo? :shock: :mrgreen:
Do you Pi?

xnosek00
Posts: 27
Joined: Mon Nov 12, 2012 2:17 am

Re: Overclocking

Mon Nov 12, 2012 2:27 am

I have two pieces of Raspberry Pi - 512 MB Ram.

First of them is overclocked to 1GHz (through raspi-config). Everything works fine. No corrupted SD card (checked).

I move the SD card from first Raspberry Pi into the second Raspberry Pi. The second Raspberry Pi has corrupted this SD card.

I received both of RasPi in the same time (big package) :)

If no overvolting - the second Raspberry Pi works fine.

Maybe this information helps to you. I don't know, why is data corrupted on second Raspi and not corrupted on the first Raspi.

I tried "initial_turbo=30" - didn't help.

n1ko
Posts: 10
Joined: Mon Nov 12, 2012 5:17 am

Re: Overclocking

Mon Nov 19, 2012 4:53 am

I also am having trouble overclocking the core. Current (stable) settings are:

arm_freq = 1000
sdram_freq=450
gpu_freq=250
overvoltage=6
force_turbo=1

Any rise done on the core_freq and I'm hit with SD corruption. Using class 10 transcend mini SDHC card.

dero
Posts: 21
Joined: Wed Nov 07, 2012 3:21 pm

Re: Overclocking

Mon Nov 19, 2012 8:12 pm

Can you see the corruption in kern.log, i.e. does the SD driver (mmc0) tell you that something is going wrong?

Code: Select all

static int __trace( const char* file, const int line ) { pr_info( "#TRACE %s:%d\n", file, line ); return 1; }
#define return if( __trace( __FILE__, __LINE__) == 0 ) {} else return
#undef pr_debug
#define pr_debug pr_info

Return to “Advanced users”