Underclocking

i dodn't rember where i saw it but somebody took a 1.33 ghz axia tbird down to 1 ghz with a 1.3 vcore and he used a passive heatsink with a temp around 40 degress celcius

Cool - thanks. That at least lets me know that underclocking is possible, enough that it's worth playing around with some.
- M.S.

Yeah, but I'd rather spend $30 on a good HSF and run it at normal (or slightly above) clock speed than spend $100 on a higher processor and underclock it.

Why are you trying to run a CPU bare? Noice? Space?

Apple? Macintosh? What are these strange words you speak?

The "running the processor nekkid" requirements come from the fact that due to the specifications for this project, we must run the final product in a sealed box. The outside of the box will be about 180 degrees Farenheit. (ouch) Now, we can plaster a monster heat sink on one side of the box, and the processor can be positioned to be on the inside wall of the box that has the heat sink. But when all is said and done, the most heat we could theoretically disapate is around 1watt! Realistically, around 1/2 or 2/3 of a watt is more do'able. There will probably not be enough room in the case for anything to stick up above the processor more than about 1/8 of an inch. (and if we decide to butt the processor up against the inside wall to dissapate some heat through the wall to the outsie of the case, then obviously there will be no room to put any sort of heat sink physically on the processor.

Cost is not an isue. <big grin> The folks buying this product from us have made that clear. They have also stated that we may NOT build a custom processor for this project. We must use off-the-shelf, "commonly available" parts.


Hmmm... interesting problem.

well there would be a few problems... what is this processor doing going to jupiters moons? underclocking is a good idea... if you can i advise you do for a transmeta as i doubt speed is that big an issue as youre wanting under a watt... how long is this thing gona be running for? the air in the case will get hotter as well if you think about it from just the processor itself and the motherboard... and if the temperature outside is that high then having a vacuum flask like case wouldnt be at all a bad idea as that temperature would travel through the case and heat the prcessor anyway unless you have a vacuum flask type sealed case... what project is this for?

you do not strengthen the weak by weakening the strong

Due to my NDA, I can't tell you what the processor will be doing other than it will be running a RTOS, and a single (processor intensive) application in a networked environment. Which RTOS and application we use are up in the air until we find a processor cool enough to run them.

We've been talking to Transmeta! <grin> Good call. (The Register.co.uk had an announcement this morning about their upcoming processor which may solve the problem for us... in two years.)

A vacuum in the case. Hmmm... we hadn't considered that. I'd have to check the spec to see if that would be allowed.
My rule of thumb when talking to non-technical sales people from different processor companies has been to tell them that if they're willing to LICK their 400MHz processor, after it's been running under heavy load for 6 hours, we'll buy it on the spot! <grin> no takers so far!

you might want to look into a proccser that will stepp its speed down when it gets hot or it is not needed you find most of these in the moblie market

No personal experience here but I think your best bet will be running one of the new mobile processors, much lower power consumption resulting in lower heat production. If cost is no object you might even want to strip a laptop and use the core hardware for your project.

I think you have difficulties then?

You are having an env with 180F outside - right?

Any cpu will generate heat - as will all the other components. If your outside is 180F then the inside will at least get there. 82C may be survivable for a CPU - just - but you'll lose memory, storage and chipsets at that temp.

The coolest/best you maight get is something like a strongarm processor - 200Mhz or so and able to run CE. Check the iPaqs for comparison....

-* This Space For Rent *-
email for application details

How about a Peltier, with the hot side attached to the case?

Cast not thine pearls before the swine

I don't think the TEC effect is two way like that is it? I think you have to cool the hot side and apply current to super cool the cold side OR cool the cool side, heat the hot side and generate current. I don't think it can bypass the laws of thermodynamics in the way you suggest?

-* This Space For Rent *-
email for application details

Ok, let's figure this out using basic thermodynamics. You remove some heat from the processor (using the cool side of the peltier). Now where does that heat go? To the hot side of the peltier. Big deal, so where does it go from there? Like I said, attach the HOT side to the CASE. Now the case becomes a heatsink. But, you say, the case is already in an 180 degree environment? Well guess what, your case WAS 180F. Now it's 190F, big deal. He even said he could put a heasink on the OUTSIDE of the case. Good, the bigger the heatsink, the closer the case will be to 180F.
Of course it would have to be a BIG peltier, because it would loose some efficiency at the higher temp, but peltiers are thin, which meets his requirements.
In fact, if he LINED THE CASE with peltiers, using a double-walled case, and insulated the dead space, the whole INTERIOR of the case would be cooler, which may make the memory hold up as well. In effect he could turn the case into a mini cooler! And he would have to attach the cooled inner surface of the case to the processor, and use it to cool the die. Now do we understand each other?

Cast not thine pearls before the swine

not inside the actual case... between the inner case and the outer case... so that the temperature from the outside doesnt get in... how long must this processor run for that is the main question... to run at 400mhz at 1 watt is damn hard at the moment... i seriously doubt it is possible with the standard chips sold by the companies... you would have to heavily underclock the chips to get anywhere near that and knowing the volume of the case cold be useful... if the box only has to run for a limited time a good idea would be to either fill it with a non-condructing fluid with a high specific heat capacity... lets say you have it filled with something similar to water... shc of water is 4200J/degree C/Litre... if you can get it down to 1 watt... and your case has a volume of 1 litre... the liquid starts lets say at 20 degrees... the processor will run until the temperature reaches 60-70 degrees if youre lucky lets say 60... discounting any heat coming in from outside the case thats 3600 J given out per hour by the processor at 1 watt... that about .85 of a degree per hour... you can have a 40 degree rise in temperature... giving you almost 2 days running time in a totally sealed off environment...if you double the capacity of the case it doubles the running time... double the processor energy release and halve the running time... it would be possible to lower the temperature without using fans but instead a vapochill like mechanism and as "cost is not an issue" you could convert a fridge with the compressor on the outside to lower heating of the inside... that would cool the fluid within the case lengthening the life of the processor... i would seriously advise against having the processor against a wall which has a heatsink on it... that would heat the processor to 180 degrees F before it even runs and thats not the best of things at all... good luck whatever youre doing...

you do not strengthen the weak by weakening the strong

Underclocking is possible - I've underclocked my 1.5GHz TBird to 450MHz!!

---------
Grass is a beautiful weed

Unless you actively cool the hot side of a peltier, they both get hot - and then melt. I am not sure what the dissipation is, but presuemably you must be able to dissipate at least as much heat as is generated = equilibrium.

Thinking about it - if he can disspiate the heat off the outside of the case - then why not? As long as there is disspiation of heat rather than build up. But if the outside of the case is 180F = 88C then something might fail. Also, who do you protect the rest - line it all with peltiers and fill it with freon - might work I guess - you could even turn it into a super-cooler and get a 2G P4 or 1.7G A4 in there....

I'm sure there is a gerat reason why all this won't work, but I'd love to bench it and see why....

-* This Space For Rent *-
email for application details

how are you getting power to this box


couldn't you run a a cooled nonconductive liquid in a pipe along side the eltrical power and pump it in and cool it down and then run the hot liquid out the only prblem i for see in this is that the inline would have to be insulated to make sure that it dosen't pick up much heat form the ari

I'd say your only options are water cooling and freon, but since you have 1/8th of an inch, both of those would be hard to get set up.

Apple? Macintosh? What are these strange words you speak?

Can you tell us why the computer can't be run in a different room? I'm assuming that it has to interface with the equipment that's causing it to be 180F (or that the other equipment can't run at less than that). Otherwise, you could string cable instead of having the box in there.

Apple? Macintosh? What are these strange words you speak?

Wow - I certainly didn't expect this much input in a mere 24 hours. My thanks to those of you who have replied.

The peltier idea may be worth investigating. This won't be a battery powered device, so we could concieveably mount peltier wafers on the inside of the case, cool side in. (fine, let 'em heat up the outside of the case to 200 degrees... as long as the inside stays cool)

In answer to the "how is power getting in thie magical box: "I have no clue!" <grin> That's a headache for another engineer, not me. The box isn't REALLY air tight, there's just no place for air to go outside this box, and no holes to speak of in the box walls.

Liquid cooling on the other hand probably wouldn't be an option. One of the stress tests that we will have to pass is an impact test. (you know, the fun stuff like hitting it with a sledge hammer. <grin>) I think that we'd run into too many problems with liquid leaking. (even non conductive liquid, once it leaks, won't do it's cooling duties.

And one last answer, as far as how long this box will need to run for any given time... I believe the new estimate is only 2 or 3 days of straight running time. (I could be wrong... we're designing it to run perpetually, as long as it's needed.)

Sounds like you have a tough job on your hands. Aren't Peltiers fragile, though? I don't know much about them.

Apple? Macintosh? What are these strange words you speak?

Mmmmm... good point - at least the peltiers that I have worked with were a little on the brittle side. Ahhhh... I'll just GALVANIZE the ceramic peltier and call it good.

So we're still not really any closer than we have been...

Apple? Macintosh? What are these strange words you speak?

i meaqnt that you have an out side source of the liquid which could run along the side the same line that take power into it and only hook it up once it is in place and then un hook it when it is done


be just like pluging it in to get the power to it instead you would be pluging it in to a hose that would bring a cool liquid to it


the only way i could see this as beging a problem is if you actuallt do a stree test on the machine when it is pluged
which i doubt you do



an all you would have to do is trun the pump on to move the liquid a few minute before you start the processer to file it up and then you could have a small pump in the box or just outside the box to pull must of the liquid out of the box

if you ran it in a vaccum the chip would melt, without any contact material to dissipate heat the chip would retain all of its heat, and melt no matter how cool it ran.(like a thermos)use a peltier and conntect it to the outer side of the case with a large heatsink, and run the pelt at a low setting.

~Matisaro~
"Friends don't let friends buy Pentiums"
~Tbird1.3@1.55~

Hmmmm... good point. A vaccum won't help this scenario at all. I should talk to some of the mechanical engineers and see if there is such a thing a a 'heat diode." (ie: only allow heat transfer one way)

Hmmm... you know, that kinda comes back to the peltier idea. I wonder if it's possible to make heavy-duty peltier devices. The ones I have seen were for PC's, (I'm sure everyone here has seen them plenty) but could one be made that was large enough, and strong enough to make a case out of? I supopose the outside of the case would be 250 degrees F or so, but the inside would end up like your fridge at home. And no mechanical moving parts. (BTW: the mechanical engineers nixed the idea of liquid cooling. Thanks for the idea though wapaaga.)

Oh well... the weekend's here. I'll stop worrying about it until Monday. <grin>

A big thanks to everyone for their resonses. Based on the suggestions and feedback from everyone, Rockwell has commissioned me to whip together a quick "underclocking lab" and get some hard numbers for various processors. (mostly PPC processors)

Thanks again for the input.
- Micah S.

a partial vacuum flask idea wouldnt be that bad... even if you used peltiers (however you spell it) you could somehow use a good heat conducting jelly to attach it to the case side... and the other 5 sides of the cubic case would have a small vacuum in the walls to lower their conductiviy... only allowing heat to be radiated... conduction not becoming an issue... allowing for the inside of the case to stay significantly cooler... the best idea i think liquid cooling with the liquid flowing with the power cables which was suggested... but anyway... whatever you do... make sure you enjoy your near limitless funds... and good luck...

you do not strengthen the weak by weakening the strong

You don't really want to reduce the heat conduction. The heat needs to be dissipated from the processor. Unless it is producing no heat whatsoever, the processor will eventually overheat without the heat being dissipated from the case.

Apple? Macintosh? What are these strange words you speak?

youd want to reduce it on all sufaces other than the peltier hot side... otherwise a lot of energy is going to be wasted just cooling the inside by the other walls allowing all the heat back in overheating the ram... hdd... etc...

you do not strengthen the weak by weakening the strong

Extruded aluminum case with the cooling fins acting as impact cushions, line the case with Peltier devices attached with JB weld, fill the dead spaces with non-conductive material, attach inner panels using more JB-Weld, press the CPU against the top of the inner panel using a spring mounted mainboard, fill the case with oil to reduce component vibration. Now you have a solid unit with oil-cooled components, the oil is cooled by the inside of the case. Because the outside of the case is so large, it would only have to be maybe 10F hotter than the inside in order to dissapate enough heat. You ask for a type of "diode" that only allowed heat transfer in one direction, the Peltier is such a device.

Cast not thine pearls before the swine