User avatar
yv1hx
Posts: 380
Joined: Sat Jul 21, 2012 10:09 pm
Location: Now an expatriate, originally from Zulia, Venezuela
Contact: Website Skype

Insane RPi overclocking...

Sat Aug 10, 2013 8:21 pm

Hi RPi enthusiasts:

This guy claims being overclocked his RPi to a 5GHz level (!!!?) using a commercial refrigerant liquid.

This is possible without melting the RPi, BTW, I don't want to blame the author post or create a angry discussion about morals on voiding the warranty ;)

http://misapuntesde.com/post.php?id=225
Marco-Luis
Telecom Specialist (Now Available for Hire!)

http://www.meteoven.org
http://twitter.com/yv1hx

User avatar
Jim Manley
Posts: 1600
Joined: Thu Feb 23, 2012 8:41 pm
Location: SillyCon Valley, California, and Powell, Wyoming, USA, plus The Universe
Contact: Website

Re: Insane RPi overclocking...

Sat Aug 10, 2013 10:09 pm

yv1hx wrote:This guy claims being overclocked his RPi to a 5GHz level (!!!?) using a commercial refrigerant liquid. This is possible without melting the RPi, BTW, I don't want to blame the author post or create a angry discussion about morals on voiding the warranty ;)
http://misapuntesde.com/post.php?id=225
5 GHz for 10 seconds, then the SoC apparently goes into thermal shutdown protection mode. He also claims 2 GHz 24/7, which sounds feasible using refrigeration.

OK, who has ready access to some liquid helium? If you use hydrogen and you get it cold enough, it becomes a metal, which would become a problem for the circuitry, not to mention the thermal issues concerning potentially shattering the SoC, the PCB, other components, etc., if they're not very slowly cooled down symmetrically in all dimensions (otherwise, it's Thermal Cracking Time) :lol:
The best things in life aren't things ... but, a Pi comes pretty darned close! :D
"Education is not the filling of a pail, but the lighting of a fire." -- W.B. Yeats
In theory, theory & practice are the same - in practice, they aren't!!!

gyeben
Posts: 150
Joined: Tue Jun 19, 2012 6:38 pm
Location: Hungary, Eger

Re: Insane RPi overclocking...

Sat Aug 10, 2013 10:24 pm

yv1hx wrote:This is possible without melting the RPi
Yeah, the RPi didn't melt, but the "SD card has burnt". :|

User avatar
yv1hx
Posts: 380
Joined: Sat Jul 21, 2012 10:09 pm
Location: Now an expatriate, originally from Zulia, Venezuela
Contact: Website Skype

Re: Insane RPi overclocking...

Sat Aug 10, 2013 10:30 pm

Jim Manley wrote:
yv1hx wrote:This guy claims being overclocked his RPi to a 5GHz level (!!!?) using a commercial refrigerant liquid. This is possible without melting the RPi, BTW, I don't want to blame the author post or create a angry discussion about morals on voiding the warranty ;)
http://misapuntesde.com/post.php?id=225
5 GHz for 10 seconds, then the SoC apparently goes into thermal shutdown protection mode. He also claims 2 GHz 24/7, which sounds feasible using refrigeration.

OK, who has ready access to some liquid helium? If you use hydrogen and you get it cold enough, it becomes a metal, which would become a problem for the circuitry, not to mention the thermal issues concerning potentially shattering the SoC, the PCB, other components, etc., if they're not very slowly cooled down symmetrically in all dimensions (otherwise, it's Thermal Cracking Time) :lol:
I would guess that there is a "optimal" temperature were the SoC can have the best performance in terms of speed?
Marco-Luis
Telecom Specialist (Now Available for Hire!)

http://www.meteoven.org
http://twitter.com/yv1hx

User avatar
Jim Manley
Posts: 1600
Joined: Thu Feb 23, 2012 8:41 pm
Location: SillyCon Valley, California, and Powell, Wyoming, USA, plus The Universe
Contact: Website

Re: Insane RPi overclocking...

Sun Aug 11, 2013 2:38 am

yv1hx wrote:I would guess that there is a "optimal" temperature were the SoC can have the best performance in terms of speed?
We have to distinguish between the temperature the SoC is operating at and the ambient (surrounding) temperature. The SoC will perform best at the highest temperature where damage doesn't occur. Using refrigeration can help remove heat at least as fast as it's being generated, but the more the difference in temperature between the SoC and the cooling medium (the refrigerant), the higher the risk of thermal cracking due to uneven heat transfer in all dimensions.

However, as the SoC temperature increases, its lifespan will start to drop, and the higher the temperature, the shorter the lifespan at an exponential rate. In other words, for each degree of SoC temperature increase, the lifespan can drop first by minutes, then hours, then days, weeks, months, and ultimately, years from the normally-expected duration of decades. If the temperature gets high enough, fast enough, the SoC can be destroyed by its semiconductor junctions vaporizing before the SoC's thermal protection mechanism kicks in and shuts it down. Even if the SoC is protected by the thermal shutdown mode though, the lifespan can still be reduced, so be aware that any amount of overclocking and/or overvoltage operation can reduce the expected time before failure.

Note that it's not possible to predict the lifespan of any particular SoC - some will last longer than others even though they were all operated at the same clock speeds, overvoltage amount, temperature, etc. That's the nature of manufacturing yields of semiconductor devices - ya rolls yer dice and ya takes yer chances! :lol:
The best things in life aren't things ... but, a Pi comes pretty darned close! :D
"Education is not the filling of a pail, but the lighting of a fire." -- W.B. Yeats
In theory, theory & practice are the same - in practice, they aren't!!!

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5537
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Insane RPi overclocking...

Sun Aug 11, 2013 10:28 am

It's fake. The PLLs don't go that high, and extreme cooling doesn't really help overclock (the SoC will run faster at room temperature than below zero).

User avatar
yv1hx
Posts: 380
Joined: Sat Jul 21, 2012 10:09 pm
Location: Now an expatriate, originally from Zulia, Venezuela
Contact: Website Skype

Re: Insane RPi overclocking...

Sun Aug 11, 2013 5:03 pm

dom wrote:It's fake. The PLLs don't go that high, and extreme cooling doesn't really help overclock (the SoC will run faster at room temperature than below zero).
Glup :shock:
Marco-Luis
Telecom Specialist (Now Available for Hire!)

http://www.meteoven.org
http://twitter.com/yv1hx

User avatar
ulysess
Posts: 310
Joined: Thu Aug 02, 2012 6:35 am
Location: Spain
Contact: Website

Re: Insane RPi overclocking...

Tue Aug 13, 2013 7:00 am

dom wrote:It's fake. The PLLs don't go that high, and extreme cooling doesn't really help overclock (the SoC will run faster at room temperature than below zero).
RTFM dom, on the post from overclock.net has too many clues to believe. I don't think is fake. It's insane to pretend 3 Ghz+ was stable, but this guy reach the goal.

Anyway, too many people has trying overclock their PIs to the limit, so it's not a surprise. If you need some technical information, I refer you to his author.
  • PiKISS for Raspberry Pi: https://github.com/jmcerrejon/PiKISS
  • Blog: https://misapuntesde.com/
  • Patreon: https://www.patreon.com/cerrejon?fan_landing=true
  • Twitter: https://twitter.com/ulysess10
  • Discord: https://discord.gg/Y7WFeC5

User avatar
mahjongg
Forum Moderator
Forum Moderator
Posts: 13101
Joined: Sun Mar 11, 2012 12:19 am
Location: South Holland, The Netherlands

Re: Insane RPi overclocking...

Tue Aug 13, 2013 7:47 am

no dom is right. his comments are factual.

that said, the cooling is not done to directly make the chip faster, instead its done to prevent the energy from overvolting the PI to heat is up so much that its lifespan is reduced to seconds, but if you actually would succeed in cooling the chip down it will become slower, not faster. so cooling it like this is a balancing act! You want the chip as hot as possible to support the speeds you need, but not so hot that it will fail.

But if we may believe dom, who has deep technical knowledge of the PI so should be believed, there is simply a limit to the degree you can change settings that affect the degree of over clocking/overvolting that is possible, and these settings do not allow for over clocking of up-to 5GHz, as the OP claimed.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 26665
Joined: Sat Jul 30, 2011 7:41 pm

Re: Insane RPi overclocking...

Tue Aug 13, 2013 8:28 am

mahjongg wrote:no dom is right. his comments are factual.

that said, the cooling is not done to directly make the chip faster, instead its done to prevent the energy from overvolting the PI to heat is up so much that its lifespan is reduced to seconds, but if you actually would succeed in cooling the chip down it will become slower, not faster. so cooling it like this is a balancing act! You want the chip as hot as possible to support the speeds you need, but not so hot that it will fail.

But if we may believe dom, who has deep technical knowledge of the PI so should be believed, there is simply a limit to the degree you can change settings that affect the degree of over clocking/overvolting that is possible, and these settings do not allow for over clocking of up-to 5GHz, as the OP claimed.
What he said. Dom is one of (if not THE) world leading experts on the 2835. Who would you believe, one of the people who helped designed and write the very core software for the chip, or someone on the internet.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

SirLagz
Posts: 1705
Joined: Mon Feb 20, 2012 8:53 am
Location: Perth, Australia
Contact: Website

Re: Insane RPi overclocking...

Tue Aug 13, 2013 8:41 am

So what *is* the highest possible overclock then ?
My Blog - http://www.sirlagz.net
Visit my blog for Tips, Tricks, Guides and More !
WiFi Issues ? Have a look at this post ! http://www.raspberrypi.org/phpBB3/viewtopic.php?f=28&t=44044

User avatar
tonyhughes
Posts: 951
Joined: Wed Dec 26, 2012 3:46 am

Re: Insane RPi overclocking...

Tue Aug 13, 2013 8:49 am

jamesh wrote:What he said. Dom is one of (if not THE) world leading experts on the 2835. Who would you believe, one of the people who helped designed and write the very core software for the chip, or someone on the internet.
Dom is on the internet...???

$ sudo ./run
$ sudo ./hide
$ exit

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5537
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Insane RPi overclocking...

Tue Aug 13, 2013 10:37 am

SirLagz wrote:So what *is* the highest possible overclock then ?
The highest I've believed is around 1250MHz.

Basically you need to run:

Code: Select all

vcgencmd measure_clock arm
If that reports the frequency claimed then I *might* believe it. They can back it up with suitably increased nbench results.

Most likely when choosing a ridiculously high frequency, what happens is you overflow a multiplier register in the PLLs and you're actually getting a half or quarter of what you think you are getting.

User avatar
mahjongg
Forum Moderator
Forum Moderator
Posts: 13101
Joined: Sun Mar 11, 2012 12:19 am
Location: South Holland, The Netherlands

Re: Insane RPi overclocking...

Thu Aug 15, 2013 11:20 am

Also note that the site misapuntesde.com is now on a list of "untrusted and possible malicious sites", and is now blocked as such by my browser!

also see:
http://www.scamadviser.com/is-misapunte ... -site.html

User avatar
CWRoos
Posts: 17
Joined: Thu Jul 18, 2013 4:43 am

Re: Insane RPi overclocking...

Thu Aug 15, 2013 1:37 pm

Of course its fake, but even so, what really is the point of overclocking?
Its the software that makes a computer go slow.
Programmers get lazy, linking enormous libraries for maybe just one simple function .
Want any computer to go fast.. program in asm and don't use libraries.

Where are the days that space quest and monkey island fit on a diskette and never needed overclocking :)

User avatar
RaTTuS
Posts: 10559
Joined: Tue Nov 29, 2011 11:12 am
Location: North West UK
Contact: Twitter YouTube

Re: Insane RPi overclocking...

Thu Aug 15, 2013 1:51 pm

CWRoos wrote:Of course its fake, but even so, what really is the point of overclocking?
Its the software that makes a computer go slow.
Programmers get lazy, linking enormous libraries for maybe just one simple function .
Want any computer to go fast.. program in asm and don't use libraries.

Where are the days that space quest and monkey island fit on a diskette and never needed overclocking :)
No - some problems cannot be speeded up by going to asm
infact a lot of projects will go a lot faster using a compiled high level language
you then ID the problem areas and optimize those if it is possible [even by doing hand crafted assembler yourself]
or re-code everything in an alternative way.
How To ask Questions :- http://www.catb.org/esr/faqs/smart-questions.html
WARNING - some parts of this post may be erroneous YMMV

1QC43qbL5FySu2Pi51vGqKqxy3UiJgukSX
Covfefe

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 26665
Joined: Sat Jul 30, 2011 7:41 pm

Re: Insane RPi overclocking...

Thu Aug 15, 2013 2:39 pm

ASM is useless except when you really REALLY need it (and those times are few and far between). It's difficult to write, difficult to maintain, is non-portable, and doesn't give much benefit if any over a decent compiler.

But that said, the way you write your high level language has a lots to do with efficiency - there is a lot of wasteful coding out there. Both in raw processing and memory usage. Although linking in a big library for one function doesn't affect program speed, just size. But that's not lazy. There's no point in reinventing the wheel.

By as said above, so algorithms do need lots of horsepower. Look up H264, and tell me how to do that at 1080p on a BBC micro...
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

User avatar
mahjongg
Forum Moderator
Forum Moderator
Posts: 13101
Joined: Sun Mar 11, 2012 12:19 am
Location: South Holland, The Netherlands

Re: Insane RPi overclocking...

Thu Aug 15, 2013 3:36 pm

Well, yes, its not so much the language you are using that has major effects on the speed of which a task is executed (at least not for compiled languages) but the efficiency of the algorithm you are using. If your whole application is riddled with poorly conceived algorithms then the whole thing could end up being a hundred times slower than it has to be. That has in parts to do with lazy programming. Not putting in enough effort to find the most efficient way to solve a task, and perhaps poor overall planning.

That said, with more memory to "play with", meaning that things like much larger video frame buffers and such exist now than the similar structures of the past things naturally take more time. Software now simply needs much more data to move around and search through etc. than was needed in the past.

XBMCantgetenough
Posts: 31
Joined: Wed Jan 09, 2013 2:05 am

Re: Insane RPi overclocking...

Fri Aug 23, 2013 5:33 pm

I wrote a response to that articel but it was considered to harsh so It was taken of. I've modified it here it is

"1/you have not disable dynamic clocking with force_turbo=1 . This means you cannot overclock past turbo settings i.e 1000 arm

2/the over_voltage is ARM/GPU dependent you have set it to 16 but the arm soc has a HARDWARE LIMIT of 8 no matter what you do you cannot exceed that. Also without force_turbo=1 you cannot even pass over_voltage=6

3/core_freq has a limit of 600 (this is done with using "vcgencmd measure_clock core" command) set it at 599 and check the output and set it at 601 and you would get an output of 500! anything over 600 just defaults to 500! so your not running 1600

4/you believe you have sdram_freq=1333 but failed to add any further voltage to the sdram namely "over_voltage_sdram=x". Default sdram_freq is 400 and you've trippled this without adding any extra power? yet again there is also a hardware limit of 8 (by the way this comes from the Dom from the Raspberry Pi foundation I can give links if you don't believe but you can check the output of the voltage with this command
"vcgencmd measure_volts sdram_c" set it at 8 and you would get the max over_voltage of 1.4v (default os 1.2 and each over voltage gives an addtitional 0.025 v giving maximum 1.4) try it past over_voltage_sdram=8 and you would still get 1.4.

5/without running force_turbo=1 your running dynamic overclocking meaning that YOU'VE ALWAYS BEEN RUNNING AT 700 ARM. It only increases when the Pi is pushed hence dynamic overclocking and since you only had it running for 10 seconds theirs no way it was even pushed so yet again all this time you have been running at arm_freq=700 (the default)

So basically all your settings are completely rubbish. The arm, core, sdram, over voltage. You're even missing stuff that without, your config.txt don't even make sense. Sorry you don't even show that you understand that basics of the basics. I really don't mean to be this harsh but the absurdness of this and the fact that every single person believed this is just mind boggling. The Op needs to be ejected from the club ASAP sorry but this whole project, after I've give factual evidence shows to me that your a kid that cooled his Pi alot and did nothing else , worse you are most proberbly on default settings. Infact why did no one mention that this commands tells you the tempreture of the Pi
"

He doesn't have a clue what he is doing. These are my settings that I'm currently running that so far seem to be stable. (I mess around with it alot)

arm_freq=1150
core_freq=580
avoid_pwm_pll=1
current_limit_override=0x5A000020
sdram_freq=700
over_voltage_sdram=8
over_voltage=8
gpu_mem_512=128
initial_turbo=60
hdmi_ignore_cec=1
disable_pvt=1

I can reach 1170 - 1175 but it will reboot xbmc after I stress it slightly (move around the menus). I beleive thats because the psu can't keep up with the power my pi needs at these config.txt (I also have a wifi in one port, a USB 3.0 in another and a 5v tsop running of 3v) I've got a nexus 4 charger.

I believe the only other way I can reach higher clock speeds is to use a faster USB 3.0

I would love see proof of the Pi running at 1250. and passing a stress test. not just sat on idle.

XBMCantgetenough
Posts: 31
Joined: Wed Jan 09, 2013 2:05 am

Re: Insane RPi overclocking...

Fri Aug 23, 2013 5:34 pm

also you can't reach my settings without disspating the heat (I use about 12 heatsinks stacked ontop of eeach other while at 45 degrees to horizontal)

If anyone wants to benchmark my Pi, I won't mind doing that. But I run XBMC.

Heater
Posts: 15952
Joined: Tue Jul 17, 2012 3:02 pm

Re: Insane RPi overclocking...

Fri Aug 23, 2013 7:02 pm

CWRoos,
Of course its fake, but even so, what really is the point of overclocking? Its the software that makes a computer go slow. Programmers get lazy, linking enormous libraries for maybe just one simple function .
Want any computer to go fast.. program in asm and don't use libraries.
When on earth is this myth going to die?

I would like CWroos to consider the following:

0) Yes, it's the software that makes a computer go slow. But not in the way you might think. You can spend forever tuning your code in assembler for that extra 2 or 3 times speed up over C and miss the point that a smarter algorithm will speed things up by thousands or millions of times. Check out a Direct Fourier Transform vs a Fast Fourier Transform for a classic example. Or the various sorting algorithms. Or the various approaches to B-Trees in databases.

1) If your program relies on a lot of interaction over a network it is going to be spending most of it's time waiting for remote servers and such to respond. Fetching data over a network connection can be millions, or thousands
of millions of time slower than reading it from local memory. You might as well write it in the slowest interpreted language you can find it won't make any noticeable difference.

2) If your program relies on reading and writing data to disk, even a solid state disk, it will be spending most of it's time waiting for the disk to respond. Accessing data on disk connection can be millions, or thousands of millions of time slower than reading it from local memory. You might as well write it in the slowest interpreted language you can find it won't make any noticeable difference.

3) Any big interesting program relies on reading and writing data from main memory, it will be spending most of it's time waiting for the processor caches to fill and empty to main memory. Accessing data in main memory can be hundreds or thousands of times slower than getting it from CPU registers.

4) If your program needs something like a Fast Fourier Transform (FFT), for example, you can:
a) Write an FFT for yourself, in assembler for speed, it will take you a while to figure out how.
b) Use a ready made FFT library that will no doubt outperform yours even if it is written in C.

At the end of the day writing your program in assembler will be 2 or 4 times faster than writing it in C in the best possible case which is rare. That speed up will mostly be lost in the noise. You will have spent a life time writing a web browser, for example, with no noticeable speed up. Not only that but your lifetimes work is totally useless on a different machine. What a waste of life.

As an example. I have server processes written in JavaScript. You know, interpreted and slow and horrible and they use a lot of libraries written in C. Guess what? They consume only twice the CPU power of the same thing I have written in C++. I could do it in assembler but then it would take forever to write and debug and it would not run on my PC and my Pi and many other places.

Assembler has it's place. A wise programmer knows where that is.
Where are the days that space quest and monkey island fit on a diskette and never needed overclocking
Those were wise programmers:)
Memory in C++ is a leaky abstraction .

User avatar
Jim Manley
Posts: 1600
Joined: Thu Feb 23, 2012 8:41 pm
Location: SillyCon Valley, California, and Powell, Wyoming, USA, plus The Universe
Contact: Website

Re: Insane RPi overclocking...

Sun Aug 25, 2013 4:02 am

Heater - your points are good ones, but you need to be more precise in some of your numbers. You're conflating access speeds and data transfer rates for the most part, and you're spanning multiple technology eras over time, so the ratios of access speeds aren't quite as bad as you indicate. Clock speed is not necessarily an indication of actual performance, which has long been something marketing types have intentionally used to bamboozle prospective customers. For example, register access speed is not millions of times faster than RAM access speed even at multi-GHz clock speeds, unless you routinely access your RAM over a dial-up modem.

Likewise, disk access speeds aren't millions of times slower than RAM access speeds, more like about five orders of magnitude (~10,000:1). The foregoing is all especially true on the Pi as its clock speed is much closer to the access speeds for RAM used with the SoC than high-end CISC systems. I don't know about the BCM2835's ARM CPU, but RISC systems tend to have less in the way of parallel pipelines and interleaved fetch-execute RAM access that modern CISC CPUs generally have, which reduce the differences between RAM, cache, and register access speeds.

BTW, "it's" is the contraction form of "it is" - "its" is the possessive form, which is generally what you should be using where you use "it's". I only bring this up because I find it troubling how many professed software developers can't make the effort to spell correctly, and then wonder why their code has so many latent bugs that are devilishly difficult to find (and "latent" means "hidden", not "delayed", as is so often incorrectly used). I cringe every time I see "definately" and "definatly" in so many posts (not yours, but I figured while I was up on the soapbox, I might as well get that one out into the cockroach-scattering spotlight).

Otherwise, keep up the good point-making!
The best things in life aren't things ... but, a Pi comes pretty darned close! :D
"Education is not the filling of a pail, but the lighting of a fire." -- W.B. Yeats
In theory, theory & practice are the same - in practice, they aren't!!!

Return to “General discussion”