OSDev.org

os64dev wrote:well it can be that the cache is poorly used on the blackbox with the new version or it is the temperature stepping ? I am still waiting for my harddisk so i can test on my AMD X2 4400+ with 2 GiB memory. Well we will see. One advantage candy it not warm outside so you have some free heating

I strongly think it's the temperature stepping - if your code overheats it in 4 seconds, it'll trigger one hell of a speedstep downwards. I'm not sure how to cool it better though, it has a pretty good cooler on it right now.

if your code overheats it in 4 seconds

it not his code, it's probably any program that uses 100% CPU, i would invest in better cooling

Candy wrote: I strongly think it's the temperature stepping - if your code overheats it in 4 seconds, it'll trigger one hell of a speedstep downwards. I'm not sure how to cool it better though, it has a pretty good cooler on it right now.

You might have to use one of those water coolers. As far as I know the only specialty item is the heat sink with the intake and out flow port. One of those impeller type submerged water pumps will work. I can find them as low as 29.99 U.S. dollars here. Although, I am not sure how good one will work like that or the GPM but I am sure it would help. You might even get away with just using a bucket or something such as a cooler and let the water fall back into it as a waterfall to help dissipate the heat some (not a big drop just a couple of inches). Then, buy a bubbler (dual or single) or two and some air stones (like a aquarium uses) and let these help cool the water by transferring the heat from the water to the air.

... I know the water cooling systems work, but I do not know how well the simple and cheap idea above will work. What I do know is that by using the bucket you should at least have the potential to maintain high heat transfer for a limited amount of time (until you heat the water too high). The GPM matter - the more GPM the more heat gets removed since lower temperature water stays in contact more and warmed water gets returned to the container where it should try to stay closer to the surface so place the pump intake at the bottom.

I think most water setups use a old car radiator with a fan blowing on it or something similar so the same mechanism is at play since the water will only cool as low as the air surrounding it -- unlike a air condition which alternates the pressure in the medium (freon) that is circulated.

If I had that problem I would be at the store right now, but that is mainly because I am a scientifically adventurous person.

I think you are making something very similar to a cooling tower on a nuclear reactor except the river or body of water is the container, right?

I tell you I just thought about it. You might not want the water cooler to run while the CPU is idle or doing nominal tasks since this could keep the water temperature high instead of staying low until those moments you need it to kick in to remedy this you can use a really easier method.

There is a way (which I have done before) to allow direct access to the LPT port on a computer. Write a small shell script to get run by cron or a program in C/C++. Have this script/application read information from /proc/acpi or whatever it may be that will specify the processor temperature (it is in there, I have found it before - might be /proc/cpuinfo).

When the temperature rises to a certain point you send the data byte 0xFF to the LPT port. Using a old cable (which I have done) probe with a meter until you find one of the data pins. You can do this by writing a really simple program to output a 0x00 then a 0xff with a delay. Now of course you would have already cut the cable and exposed the wires or how ever you might like to do it. Find a small relay.. most if not all should have 120V. Get a NPN transistor or a couple in parallel to increase current capacity if needed. Connect the data pin/write to the middle pin of the transistor (most likely middle - need to check or just experiment no big deal if it pops or does not work). Connect the bottom base pin of the transistor to the ground pin/wire. Connect a voltage source such from a old 12V transformer or something ripped out of a old radio. You connect the positive lead to the top collector pin of the NPN transistor and connect the relay to the bottom emitter pin (along with the connect for the LPT ground pin) on the transistor. Run the other lead for the relay back to the voltage source. You might have to insert a diode between the voltage source on one leg or find a full wave rectifier (harder to find). Oh.. and make sure to insert a resistor (you might try a 120V dial for use with ceiling fans) between the ground pin for the LPT port and the transistor to keep from pulling too high of amps on the LPT port.. safe is better with the LPT port since it is the only critical component that you should worry about.

The transistor will amplify the LPT data pin voltage using the voltage source to turn on the electromagnetic relay and in turn runs the pump when needed.

I strongly think it's the temperature stepping - if your code overheats it in 4 seconds, it'll trigger one hell of a speedstep downwards. I'm not sure how to cool it better though, it has a pretty good cooler on it right now.

Just wondering what cooler you have. I have an artic cooler pro on my AMD and it doesn't get hot constantly between 35 and 40 degrees. It can also be that the cooling paste was not applied correctly(sounds worse than intended) and that will kill the cooling capacity.

os64dev wrote:

I strongly think it's the temperature stepping - if your code overheats it in 4 seconds, it'll trigger one hell of a speedstep downwards. I'm not sure how to cool it better though, it has a pretty good cooler on it right now.

Just wondering what cooler you have. I have an artic cooler pro on my AMD and it doesn't get hot constantly between 35 and 40 degrees. It can also be that the cooling paste was not applied correctly(sounds worse than intended) and that will kill the cooling capacity.

http://www.zalman.co.kr/eng/product/view.asp?idx=145

Review on tomshardware said it should be sufficient (50-55 degrees). It could be the cooling paste on top or the area between the heat spreader and the actual cores - the cpu is a dual core as in physically two cores type, so they might not be equally high causing it to fail on that. I'll try remounting it with more cooling paste later on.

How hot is it running? I have a core 2 duo 1.6ghz and have yet to see it get above 60C. It normally runs at about 48C.

frank wrote:How hot is it running? I have a core 2 duo 1.6ghz and have yet to see it get above 60C. It normally runs at about 48C.

Well... it automatically steps down when it hits 75 by BIOS work - so I guess 75. Average temp is around 65-70 so that's way too high anyway.

Candy wrote:

frank wrote:How hot is it running? I have a core 2 duo 1.6ghz and have yet to see it get above 60C. It normally runs at about 48C.

Well... it automatically steps down when it hits 75 by BIOS work - so I guess 75. Average temp is around 65-70 so that's way too high anyway.

Well now thats definitely too hot. I was thinking maybe it was tripping too early or something. Yea that cooler looks like it will do just fine, just make sure to apply the thermal grease correctly.

what do u guys think about immersing the entire machine - the motherboard, HDD (yea i've tried it. it is safe) in oil?

things like CDROM drives and FDD wont work inside oil, so they need to be above the oil level, and i think the complete assembly will fit into something like an aquarium like box with ATX dimensions.

Fans will still revolve, but i have no idea what effect the drag will have on the current drawn by the fans.

To keep the oil cool, we'll pump fresh outside air into the aquarium using regular aquarium air pump.

Sadly though, we cant put a real fish in there

what say ?

...Who would willingly do that to their computer? it would make reusing the components practically impossible.. removing it from the oil afterwards would also be a pain..

And then we haven't even started talking about the fire hazard...

Well one of the sites says that motor oil is the best type of oil to use. Car engines get a whole lot hotter than any computer could ever hope to get.

Brynet-Inc wrote:...Who would willingly do that to their computer?

One, whose friends weep because of burnt CPUs the same as his, with no water cooling or even good heat sink assemblies available in nearby marketplaces, shipping company demanding the same charges as the goods to ship them if purchased on web, i guess that is the solution to save his poor little CPU.


About the fire hazard, if we assume that any component wont get hotter than ~80C, we can safely use motor oil.

Even if it did get hotter you have to assume that oxygen is present, and I imagine it would not be deep in the bath of oil. I think convection would occur rapidly, before that happened, giving a clue the bath is getting mighty hot. If a processor dissipated enough heat to boil oil then, I think, it would be similar to running a electric stove all day which I think is somewhere around 3000W. That processor would not be turned on very long in my current financial situation...

Not to mention the cold spots through out the bath where the heat will move to through conduction in contrast to a hot pan with oil in which the heat covers the entire bottom.

The water temperature in a automobile is generally 82C -- 115C.

I think the fire hazard is much less or a worry, than not using some oil witch has a chemical reaction with the components on the board. Not sure if oils naturally have corrosive characteristics.. but I imagine there is a oil out there that does.

AFAIK, the two classes are lubricants and coolants in which either: lubricate, or lubricate and transfer heat. If you did only the later it would classify as a coolant, and some (coolants) are very corrosive or abrasive I know for sure. Oil is a lubricant and a coolant, but has a primary usage for lubrication. In automobiles the water is the coolant.

If I wonder why use water instead of oil as the primary coolant it may have something to do with the heat capacity of water versus a oil which is why the usage of water is dominant, and not to forget mentioning that water is cheap.