From what I see, the Vin doesn't affect the final Varm. I see my Vin dropping to 4.7V now, but the Varm still measures 1.4V. Seems like the Varm voltage will be fixed regardless of what the Vin is.
I also read on the official RPi forums that the limit of 1.4V is hardware related, not firmware.
I think I found the external bits of the internal VRM. If I did, it's EPower time.
First thing though is getting an install on there that will actually overclock.
I'll post a few pictures in a bit here, need to take 'em and do some classy MSPaint work.
EDIT:
Here we go:
Your mileage may vary, void where prohibited by law, not responsible for bad descisions, etc.
I have a hell of a time making SD cards, so it may be a bit till I can test this.
Putting some appropriate voltage caps across those two MLC jobs may well help, there's about 10mV of switching ripple on the VARM line, plus any transient ripple at higher loads. This is all at idle (no display attached, no Ethernet attached, booting to a command prompt I'd assume.), load ripple will be higher. Of course, more caps may blow things to hell, who knows.
If you do and it is Varm and you make WRs, if you could credit me that'd be great
you can find some schematics of the board online. I was just looking at them to notice that one possibility is indeed to do the same thing as mentioned above:
Remove L5 and connect some external source to the point where you read the core voltage.
Maybe L5 can stay in place and you just connect some external source to the voltage readout point, but it would not be the proper way of doing things. Don't know how things have changed or got better the last few years in overclocking
The only thing that bothers me is the voltage feedback which could be monitoring the output voltage for low and high overvoltage. I need my Raspberry for programming stuff so I won't be testing it anyway
Grabbed some time to redo the Raspbian(or whatever it is) install in the hopes that OCing will work. If it does it's external VRM time.
I'm torn between something simple like a heavily heatsinked adjustable linear regulator (it's only 3.5w stock max draw for the whole board, that's not bad linear reg wise) or going whole hog and EPowering it. The linear reg would be nice from the standpoint of starting up along with the board for proper(ish) timing, but the EPower is just so much more me.
It all depends on whether the damn thing will OC. If it won't OC adding voltage isn't exactly useful.
Pi was free, if it dies I'll be annoyed, but I don't use it much. Maybe I can get a sample from someone to replace it if I give 'em some press.
I tried the hardware modification with a linear voltage regulator but with no help. It does work, but I could not get higher clocks on the pi, in contrary my SD cards got corrupted as soon as I tried +1200MHz... L5 is back where it belongs now, I'm done with overclocking the pi
no pics, didn't draw any schematics but the mod is explained on the previous page I think: remove L5 on the downside of the board, take a variable linear voltage regulator, feed it from another USB post, hook up both (Vregulator and R-pi) GND lines and attach the regulators Vout to the Raspberry-pi Vcore readout point.
Linear is not the best choice though, even with this low current dev-board I notice a lot of voltage swings between idle and load.
The swings might have been the cause of the SD-Card corruption. No voltage scaling doesn't make sense. Maybe you were just limited by the instable voltages?
This got me about 40 MHz more on the core @ 1.4v and also about 40-50ish on the memory. The actual temp was around -24 according to temp gun and raspbian:P
It's just a stock prometia btw. I think I need to do some modding on the raspberry.
Let's say I use E-Power, I just remove L5 and connect E-Power to the V readout and a ground?
Since the Raspberry uses no actual bios, you need an OS that has one, like Raspbian or Pidora, which are both meant for the Raspberry pi. Personally I would use Raspbian, because I think it works and handles well. If you are doing it for Hwbot Prime, don't forget to install Java JDK 8. It can be a pain to install in some cases tho.
Let me know if you have any issues!
nice guide M
I've been testing Arch Linux with Java 8 preview, no real gains there.
Closing some processes might help you a little bit, but 'ps aux' (linux task manager) doesn't show any process that is taking a lot of cpu time
We could use things like http://www.dhgate.com/product/lm317-assembly-board-dc-dc-ac-dc-voltage/159646661.html to control the Vin.
From what I see, the Vin doesn't affect the final Varm. I see my Vin dropping to 4.7V now, but the Varm still measures 1.4V. Seems like the Varm voltage will be fixed regardless of what the Vin is.
I also read on the official RPi forums that the limit of 1.4V is hardware related, not firmware.
I think I found the external bits of the internal VRM. If I did, it's EPower time.
First thing though is getting an install on there that will actually overclock.
I'll post a few pictures in a bit here, need to take 'em and do some classy MSPaint work.
EDIT:
Here we go:
Your mileage may vary, void where prohibited by law, not responsible for bad descisions, etc.
I have a hell of a time making SD cards, so it may be a bit till I can test this.
Putting some appropriate voltage caps across those two MLC jobs may well help, there's about 10mV of switching ripple on the VARM line, plus any transient ripple at higher loads. This is all at idle (no display attached, no Ethernet attached, booting to a command prompt I'd assume.), load ripple will be higher. Of course, more caps may blow things to hell, who knows.
If you do and it is Varm and you make WRs, if you could credit me that'd be great
you can find some schematics of the board online. I was just looking at them to notice that one possibility is indeed to do the same thing as mentioned above:
Remove L5 and connect some external source to the point where you read the core voltage.
Maybe L5 can stay in place and you just connect some external source to the voltage readout point, but it would not be the proper way of doing things. Don't know how things have changed or got better the last few years in overclocking
The only thing that bothers me is the voltage feedback which could be monitoring the output voltage for low and high overvoltage. I need my Raspberry for programming stuff so I won't be testing it anyway
Grabbed some time to redo the Raspbian(or whatever it is) install in the hopes that OCing will work. If it does it's external VRM time.
I'm torn between something simple like a heavily heatsinked adjustable linear regulator (it's only 3.5w stock max draw for the whole board, that's not bad linear reg wise) or going whole hog and EPowering it. The linear reg would be nice from the standpoint of starting up along with the board for proper(ish) timing, but the EPower is just so much more me.
It all depends on whether the damn thing will OC. If it won't OC adding voltage isn't exactly useful.
Pi was free, if it dies I'll be annoyed, but I don't use it much. Maybe I can get a sample from someone to replace it if I give 'em some press.
You're correct on the read points. This is what I meaured (tested with changing the voltages)
http://www.raspberrypi.org/wp-content/uploads/2012/10/Raspberry-Pi-R2.0-Schematics-Issue2.2_027.pdf
it has the caps/resistors/... mentioned in the schematic, so easy to find the correct spot on the pcb
http://www.decryptedtech.com/news/element14-website-for-raspberry-pi-enthusiasts
Hey Pieter, in
it should say openjdk-7-jre (or 6)OR sudo apt-get purge openjdk*
Btw, I see there is a new build b102, here's the donwload link for ARM http://www.java.net/download/jdk8/archive/b102/binaries/jdk-8-ea-b102-linux-arm-vfp-hflt-07_aug_2013.tar.gz
Changelog: http://download.java.net/jdk8/changes/jdk8-b102.html?q=download/jdk8/changes/jdk8-b102.html
Thanks, updated.
I tried the hardware modification with a linear voltage regulator but with no help. It does work, but I could not get higher clocks on the pi, in contrary my SD cards got corrupted as soon as I tried +1200MHz... L5 is back where it belongs now, I'm done with overclocking the pi
Pics? Schematics?
no pics, didn't draw any schematics but the mod is explained on the previous page I think: remove L5 on the downside of the board, take a variable linear voltage regulator, feed it from another USB post, hook up both (Vregulator and R-pi) GND lines and attach the regulators Vout to the Raspberry-pi Vcore readout point.
Linear is not the best choice though, even with this low current dev-board I notice a lot of voltage swings between idle and load.
The swings might have been the cause of the SD-Card corruption. No voltage scaling doesn't make sense. Maybe you were just limited by the instable voltages?
nah, swings are really typical for linear voltage regulators
Interesting ... I go with the competition, looks funny
Hmmmmmmm doesn't fit
This got me about 40 MHz more on the core @ 1.4v and also about 40-50ish on the memory. The actual temp was around -24 according to temp gun and raspbian:P
It's just a stock prometia btw. I think I need to do some modding on the raspberry.
Let's say I use E-Power, I just remove L5 and connect E-Power to the V readout and a ground?
Another friday with the Raspberry Pi
This time: E-Power!
This got me to about 1450MHz with around 1.8v real(DMM). Next time I will make it cold to go even further. For now just a score with 1400MHz.
http://hwbot.org/submission/2500624_
Cool!
Do you have schematics?
It's exactly like they said in this topic. Remove L5 and put power from Epower on Varm readout and I used the other side of the resistor as ground.
is there any known cold bug ?
ln2 ?
No problems till -24
Should try it on my cascade soon.
of courser you should try it with phase change
would be n1 to see
Will do
It will show up here eventually 
This is great, but the other guy deserves recognition too.
He fed it with 1.8V from his vga's memory circuitry which is also awesome idea, without buying evga untouchables.
Both results are awesome, keep us up to date. Interesting to see where the limits are
.
He sure does
Also one of the sources to make the Epower work properly(not solder to wrong points).
I tried first with a voltage divider with a pack of resistors, which also worked, but couldn't regulate the voltage, it went from 5v to 1.66v.
Hello,
I'm a french student in IT and I'm just beginning in OC.
I have a question: Is the OS important in the overclicking of the RPI?
If yes, what is the best OS?
Regards,
Hopetech
PS: Sorry for my english.
Since the Raspberry uses no actual bios, you need an OS that has one, like Raspbian or Pidora, which are both meant for the Raspberry pi. Personally I would use Raspbian, because I think it works and handles well. If you are doing it for Hwbot Prime, don't forget to install Java JDK 8. It can be a pain to install in some cases tho.
Thanks for the answer.
I thought a lightweight OS such as archlinux or gentoo used fewer resources and thus improving performance.
Last question: Did you tweak raspbian?
My knowledge of linux in general is very slim, so I didn't do more than what is told in this topic:http://forum.hwbot.org/showthread.php?t=88888
Apearently, someone was able to bypass the 2GHz wall for me in the Raspbian core.
Tested last night:
how do u guys get such high results?
i only got 478 pts @1150 mhz cpu/450 mhz sdram with java8. what am i doing wrong to get such weak results? need help!
what clock is possible for sdram without cooling?
http://forum.hwbot.org/showthread.php?t=88888
That's all you need to know.
i´ve found the thread, but it doesn´t work on my pi
on runlevel 1 it won´t run prime...
anyway, team-cup is over for me now
Please log in or register to comment.