Bacan
Posts: 347
Joined: Sun Sep 25, 2011 10:03 pm

Re: overclocking

Sat Oct 15, 2011 1:36 am

Ok, let me check the logic here.
1) Want computer to go faster. Apply more power, needs better cooling.
2) Cryo or submerged liquid coolant may work, required special effort and costly.
3) Cost of second R-Pi, minor compared to improved cooling.
Added software effort to split application. (remember, software is Always viewed as free )
4) I'm more software/firmware, than hardware. So add another R-Pi, and divide the workload.
5) Whoops, now I've taken the hardware fun away from the hackers.

So for 2013, do we want a Multi-Core R-Pi model D or one dressed in a "wet suit" for the tank?

User avatar
Lob0426
Posts: 2198
Joined: Fri Aug 05, 2011 4:30 pm
Location: Susanville CA.
Contact: Website

Re: overclocking

Sat Oct 15, 2011 4:21 am

Distributing the load. OMG, no, you have spoken sacrilege, Off with your head! lol.
I have brought it up several times but it just blows on by. A mild overclock plus some division of tasks. will equal a lot more power than a single RasPi.
512MB version 2.0 as WordPress Server
Motorola Lapdock with Pi2B
Modded Rev 1.0 with pin headers at USB

http://rich1.dyndns.tv/
(RS)Allied ships old stock to reward its Customers for long wait!

User avatar
Gert van Loo
Posts: 2485
Joined: Tue Aug 02, 2011 7:27 am
Contact: Website

Re: overclocking

Sat Oct 15, 2011 4:04 pm

The main reasons you cryo sub-micron stuff is to keep it from cooking itself when you overclock and overvolt it.
I know that is what a lot of big processors do but electronics does not always work that way. Think of it as a reverse-diode. No matter how far you cool it, it will blow up at a certain voltage.
Same with frequency. I can turn the frequency up for a circuit and at a certain point it will stop working. Cooling it down might help a bit but not that you can 'double the speed'. Energy efficient devices like the BCM2835 are the same. You can get 20-80% more out of them by over-voltage, then another 5-10% by cooling, but eventually the over-voltage blows them up, not the overheating.

User avatar
Lob0426
Posts: 2198
Joined: Fri Aug 05, 2011 4:30 pm
Location: Susanville CA.
Contact: Website

Re: overclocking

Sun Oct 16, 2011 12:14 am

Cooling or Supercooling? As was dicussed earlier in this thread the BCM2835 is a PoP package. The memory is mounted on top of the processor. There is a gap between them. Even though it is small, this air gap will still act as a insulator. Cooling the PCB underneath the SoC will help some, just as placing a cooler on the Memory package above the processor will help. Even supercooling may be rendered ineffective due to these air gaps. If you do not get the actual processor supercooled in will not become a superconductor.

Supercooling could render great gains in overall performance but it is not something that could be readily sustained "at home". Somewhere out there they are still working on room temperature superconductors. If it ever comes to be we will see a revolution in electronics and their size.

P.S. does anyone know why my statement above is could be proved false?
512MB version 2.0 as WordPress Server
Motorola Lapdock with Pi2B
Modded Rev 1.0 with pin headers at USB

http://rich1.dyndns.tv/
(RS)Allied ships old stock to reward its Customers for long wait!

langlo94
Posts: 29
Joined: Sun Aug 28, 2011 12:51 am

Re: overclocking

Tue Nov 01, 2011 9:38 am

Could it be because electronics and their size is already changing at a rapid pace?

Ampix0
Posts: 94
Joined: Sat Oct 22, 2011 3:03 am

Re: overclocking

Tue Nov 01, 2011 1:52 pm

Quote from liz on August 2, 2011, 18:02
We haven't tried. Suggest you go and buy some helium. :)

(In all seriousness, there's 10-20% headroom for overclocking. But as I say, we haven't tried it yet - I expect a full and detailed report from you if you give it a go!)
http://blogs.discovermagazine......of-helium/

Dave Duchesneau
Posts: 16
Joined: Mon Aug 29, 2011 7:48 pm
Contact: Website

Re: overclocking

Tue Nov 01, 2011 4:33 pm

Quote from Svartalf on October 13, 2011, 20:35
Quote from Bacan on October 13, 2011, 15:41
I would think that the Flash Card connector pads would get compromised by the liquid. Unless of course you cleaned those contact pads of fluid every Flash Card removal & insertion cycle.

Depends on the fluid. Novec wouldn't do that, for example.

Only problem with Novec is that it's thermal transport characteristics are such that it's only slightly poorer than air conduction- compelling reason for it is that you don't have to worry about condensation with it when you drag it deep below ambient.

I'd think you'd be "okay" with Mineral Oil, really- the only problem is that it wicks up through stranded wire bundles and it's messy overall.

Really, though... I doubt that this will NEED any of those sorts of cooling. It's TDP's not much more than a Watt, if that much.

This not-so-brief note is addressed to several here who may be thinking about liquid-cooling a RasPi, etc., perhaps just for the fun of it.

It's true that 3M's Novec fluids would not cause the problems and messiness you'd encounter with oils (e.g., mineral oil), but they have their own set of challenges. Novec is actually a family of engineered fluids with different chemistries and boiling points, and they're all solvents for *something* (like any liquid, including water). In the case of Novec fluids, such as variants of hydrofluorinated ether (HFE), they're solvents for plasticizer (think soft plastic), like what you'd find in the insulation on wires and cables. As a solvent, Novec fluids don't "eat away" the insulation (i.e., it doesn't act like acid), but rather, they make make it swell and get brittle (i.e., not soft). The plasticizer dissolves into the Novec fluid, and contaminates it (creating a need for a filtering mechanism). Also, the Novec fluids I know about are generally not compatible with the silicone that you might apply to seal capacitors if you were using oil, but fortunately, no silicone is needed for Novec fluids (with most types of capacitors). Novec fluids aren't compatible with thermal grease either, not that you'd need thermal grease on a RasPi.

On the other hand, I'd have no hesitation whatsoever about immersing a RasPi in Novec HFE-7000, which has a boiling point of 34°C (94°F) at atmospheric pressure, but that's more easily said than done. You would need to use teflon-coated wires (or coat the plastic insulation with epoxy), and use a very well sealed container. HFE-7000 is pumpable below -121°C (-186°F), and circulates at that temperature about as well as water does at room temperature. HFE-7000 is slippery indeed, so a water-tight joint that would easily hold back pressurized water at room temperature would likely not hold pressurized HFE-7000 at all; it could very well spray out like a lawn sprinkler. At more than $230 to $250 per gallon, that would be an expensive lesson. The boiling point is low enough that your skin (37°C or 98.6°F) will boil it on contact, so it would be quite the negative experience if you drank some (it's inert and not particularly toxic, but would expand more than 100:1).

That said, HFE-7000 is an excellent coolant for electronics, but the packaging needs to be designed as a pressure vessel (because the liquid expands by more than 100x as it boils). The liquid would easily flow through the gap between the RasPi's processor and piggybacked memory chip(s), and would boil on the surfaces. Since liquid at its boiling point does not change temperature, all the chips would be held to the boiling point temperature plus a small delta (say, 5°C to 10°C, to account for thermal resistance, although this could be reduced somewhat with a suitable "boiling enhancement" coating), thereby achieving isothermal (single-temperature) operation with no hot spots. This has all been done before, and it works well in an engineered system (I have lots of data).

Water can be a good, cheap, phase-change thermal working fluid. Although it should be intuitive that water cannot be used in this way for direct immersion of electronics, there are four primary reasons why this is so:
1) While *pure* water may be relatively dielectric (electrically non-conductive), contaminated water is not (think of being electrocuted in water).
2) While pure water may seem inert, it can be quite corrosive to various materials, and likewise contaminated water.
3) The normal boiling point of water (100°C) is well beyond the operating temperature of most electronics.
4) When boiled, water expands by a factor of 1700x, which is so dangerous that firefighters are specially trained to deal with it, and why special licenses are required to build and operate steam-based facilities.

Using HFE-7000 as an example working fluid (since it is inert and dielectric, and poses no problem for electronics immersed in it), and by operating at the normal boiling point of the working fluid (34°C), the fluid’s capacity to absorb heat is much better than water at the same temperature, by a factor of 47x. It's also better than cold air (0°C) by a factor of 150,000x, and better than warm air (40°C) by a factor of 175,000x. The key is to use the fluid at its boiling point, where phase-change effects dominate. At that temperature, what really matters is the LHV, or latent heat of vaporization. In our example, HFE-7000 has more than 100x as much heat-absorbing capacity at its boiling point as it does at lower temperatures, and similar characteristics may be observed with other phase-change fluids. The converse is also true, in that, once the fluid has acquired the heat, it is also very efficient at rejecting it back to the ambient environment (i.e., to any temperature lower than itself).

If I were going to use an engineered fluid with a low-cost device like the RasPi, I would also consider a fluorinated ketone (FK) like Novec 1230 fire protection fluid, which has a little higher boiling point (49°C), and is thus somewhat easier to contain (due to a lower vapor pressure at ambient temperatures). Accordingly, the chips would operate at a higher temperature (say, between 55°C and 60°C), but that would typically be just fine for embedded chips. The key would be to construct the chassis from aluminum or another thermally conductive material, and hermetically seal it (which is a bit of a magic trick) with electronics and fluid inside (adding the fluid later instead requires more tricks). Simplistically, the cooling cycle would be like this: The RasPi chips boil some of the liquid, creating local vapor bubbles which condense in the fluid or on the inner sides of the container (as long as the container is below the boiling point of the fluid). Given a low-heat-flux CPU like the RasPi's ARM11, and a container with sufficient surface area, no pumps or circulation systems would be necessary to reject the heat. To minimize fluid volume (cost), the electronics should occupy most of the space within the container. The key difficulty is with achieving the hermetic sealing, which gets more difficult as the ambient temperature goes up. If the container itself cannot be kept below the normal boiling point (i.e., below 49°C, which is higher than the ambient temperature anywhere on earth), then the internal pressure will continue to rise, thereby raising the boiling point, until one of four things happens: 1) the boiling point exceeds the container's temperature, allowing the vapor to condense, or 2) the boiling point temperature exceed the electronics limits such that that devices fail and cause shutdown that reduces the temperature, or 3) the container springs a link due to the increased pressure, or 4) the container bursts due to high pressure, having no leaks to relieve the pressure (which indicates that a good design should have a pressure relief valve, like a water heater). A good pressure relief valve might cost as much as a RasPi, so such designs may be difficult to do cost effectively.

More could be said, but I've already been too verbose (more than you probably wanted to read), and I'm out of time.

Dave

radu
Posts: 110
Joined: Mon Nov 21, 2011 8:19 pm

Re: overclocking

Tue Nov 22, 2011 6:46 am

I don't think it should need any cooling (maybe a small heatsink) up to ~1ghz or so. We are talking low power ARM here, not a 125W desktop CPU.
My N900 is clocked at 650 mhz, I keep mine at 850 and it barely gets warm. Some people pushed it as far as 1.1ghz, and no one reported problems. That is, no fried CPU reports.
Anyway, with an aluminium case that touches the CPU in some area should provide more than enough cooling.

BTW, does anyone know what kind of technology the CPU is? .45nm? .32nm?

Asgo
Posts: 20
Joined: Sun Oct 30, 2011 6:10 pm

Re: overclocking

Tue Nov 22, 2011 12:40 pm

The pi is so small one could think of alternative ways of cooling.
How about putting it on the tip of a rotor for a small tabletop windmill. Cooling plus nice visual effect. ;)

User avatar
reiuyi
Posts: 165
Joined: Sun Oct 09, 2011 4:59 pm
Contact: Website

Re: overclocking

Tue Nov 22, 2011 2:01 pm

Awesome, I remember back when I had a Motorola MPx200 with a 133mhz arm processor (Texas Instruments OMAP 710), there were some tools available to overclock the CPU. I think I got it up til around 180mhz before the device crashed due to instability. The fun thing with it was that you couldn't break it no matter what you do. It'd just crash, shut down and boot up again at the stock frequency.

Back then, the best way to estimate the improved CPU efficiency was to use TCPMP (some program to play media files; it had a built-in benchmark tool). Hopefully there will be proper applications to estimate the maximum calculating capacity. As for practical uses, I doubt there are any at all! With just 1W of power consumption, you aren't saving on electricity by underclocking it, and even if you can get a 20% improved performance, you won't be able to get staggering better results from anything you can do on it. The GPU is already powerful enough to play 1088p (or 1080p, I forgot) video and some 3d games.

User avatar
abishur
Posts: 4477
Joined: Thu Jul 28, 2011 4:10 am
Location: USA
Contact: Website

Re: overclocking

Tue Nov 22, 2011 2:12 pm

All I can imagine is putting a custom board full of leds over the r-pi itself if you did that and then using the spinning motion and the LEDs to create fun images (like a raspberry)
Dear forum: Play nice ;-)

Lakes
Posts: 267
Joined: Wed Aug 24, 2011 2:17 pm

Re: overclocking

Tue Nov 22, 2011 2:13 pm

Quote from Asgo on November 22, 2011, 12:40
The pi is so small one could think of alternative ways of cooling.
How about putting it on the tip of a rotor for a small tabletop windmill. Cooling plus nice visual effect. ;)Add a row of LED`s for some POV effects. :)

Asgo
Posts: 20
Joined: Sun Oct 30, 2011 6:10 pm

Re: overclocking

Tue Nov 22, 2011 5:53 pm

Quote from abishur on November 22, 2011, 14:12
All I can imagine is putting a custom board full of leds over the r-pi itself if you did that and then using the spinning motion and the LEDs to create fun images (like a raspberry)
sounds like an interesting new output device.
does anyone has a hdmi to spinning wheel converter? :)

radu
Posts: 110
Joined: Mon Nov 21, 2011 8:19 pm

Re: overclocking

Tue Nov 22, 2011 6:39 pm

Seriously though, when you guys get some time, can you try some overclocking and tell us some real results? I am pretty sure most of the CPUs should be able to reach 1GHZ with some minor tweaks. Some heat readings would be nice as well, so people can decide for a case with passive cooling features and stuff.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 23876
Joined: Sat Jul 30, 2011 7:41 pm

Re: overclocking

Tue Nov 22, 2011 10:14 pm

We do run these things a bit faster at work sometimes, just to see when they die, but we don't do any additional cooling or even heatsinks. You don't use them in moble devices, so we don't test with them.

I think Dom or Eben in another thread said how fast we had got them to go.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
“I think it’s wrong that only one company makes the game Monopoly.” – Steven Wright

radu
Posts: 110
Joined: Mon Nov 21, 2011 8:19 pm

Re: overclocking

Tue Nov 22, 2011 10:28 pm

Did you have any die on you from OCing?

User avatar
Jessie
Posts: 1754
Joined: Fri Nov 04, 2011 7:40 pm
Location: C/S CO USA

Re: overclocking

Tue Nov 22, 2011 11:02 pm

Quote from radu on November 22, 2011, 18:39
Seriously though, when you guys get some time, can you try some overclocking and tell us some real results? I am pretty sure most of the CPUs should be able to reach 1GHZ with some minor tweaks. Some heat readings would be nice as well, so people can decide for a case with passive cooling features and stuff.

The CPU isn't the issue, it is all the the other things attached to the bus that start to fail first. Like the GPIO, USART, USB, SD IO, ect.

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5341
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: overclocking

Tue Nov 22, 2011 11:13 pm

No, I've never done any permanent damage to a chip through overclocking.
I've never experienced any failure at 800MHz (and I've ran several devices for many days at high and low temperatures).
My board seemed okay at 900MHz with limited testing.
My board crashed during boot up at 950MHz.
With 10,000 boards being sold, I wouldn't be surprised if there's a golden one that runs at 1GHz.

Note: these chips aren't like Intel processors. They don't fail due to temperature. Cooling has a minimal effect on the speed they can run at.

And obviously 700MHz is all we are promising. If it runs faster consider yourself lucky.

radu
Posts: 110
Joined: Mon Nov 21, 2011 8:19 pm

Re: overclocking

Tue Nov 22, 2011 11:24 pm

Thanks for the reply.
Did you have to increase the core voltage to get them to run at 900Mhz and higher? If so, how much did you try?
About the cooling, doesn't it become an issue if you increase the core voltage a bit? I mean, of course they are not like X86 CPUs where you need a big heatsink, but I would assume that if you stress it at ~1ghz with increased core voltage it will become pretty hot.

radu
Posts: 110
Joined: Mon Nov 21, 2011 8:19 pm

Re: overclocking

Wed Nov 23, 2011 12:52 am

And speaking of OCing, what is the standard clock for the GPU? Can it be changed? If so, did you do it?

onejay09
Posts: 3
Joined: Sun Jan 22, 2012 1:51 pm

Re: overclocking

Wed Jan 25, 2012 11:55 pm

Wouldn"t a thermal pad on the back of the PCB behind the cpu And inbetween a metal casing for mounting the r.pi be a good cooling solution?

gsh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 1442
Joined: Sat Sep 10, 2011 11:43 am

Re: overclocking

Thu Jan 26, 2012 7:15 am

No we didn't increase voltage beyond 1.2 V...  With a modest increase to 1.3V (which shouldn't do any harm) then it should hit 1G.

The GPU runs at 250MHz, I've had those hitting around 350MHz at 1.2V and this is where you really want to increase the clocking (this provides the memory logic clocking and therefore is very important for the time)

Gordon
--
Gordon Hollingworth PhD
Raspberry Pi - Director of Software Engineering

User avatar
Gert van Loo
Posts: 2485
Joined: Tue Aug 02, 2011 7:27 am
Contact: Website

Re: overclocking

Thu Jan 26, 2012 7:38 am

I played with SDRAM over voltageing and over clocking. Some manufacturers are faster then others. Unfortunately the better and faster ones are also more expensive so we don't use those on the product. So, yes I have managed to run the SDRAM much faster then 400MHz, but also reduced the lifetime a lot.

Hence: you can run it a lot faster but you have to buy a new one every few months. For the conspiracy theorists among you: No that is not a trick to sell more PI's.

IDE
Posts: 52
Joined: Fri Jan 13, 2012 12:28 am
Contact: Website

Re: overclocking

Mon May 07, 2012 6:45 pm

Is there any procedure already documented for those crazy enough to try some overclocking with the RPi.

Also, I have founded a simple cost effective way to get some cooling done.

Mineral oil based cooling.


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: overclocking

Mon May 07, 2012 9:16 pm

I understand the desire to overclock high-end processors to eek out the last possible iota of performance from them, as part of my work involves massively-parallel, high-performance computing.  However, overclocking makes no sense at all in what is essentially a low-power mobile device.

Even if you manage to run them at higher speeds, it will be for some period of time that will statistically be shorter the higher the clock speed and voltage.  The worst of it is that you won't know what the limits are for a particular device until you reach them, and you will have already shortened its life by then (see also the Heisenberg Uncertainty Principle - in measuring something, you affect the measurement, and the properties of the device itself, in this case).

I sure would like to acquire your boards before you intentionally attempt to destroy them as supply is extremely short, in case you hadn't heard.  If you can wait until people can order as many as they want and receive them immediately from stock, by all means, have as much of this type of fun as you want then.
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”