Heat management question

[sjr]

Hey all. I have a heat-and-CPU question, debated putting it in the CPU area, but thought I'd start here.

I'm getting ready to build a mini-ITX box to replace an ageing Shuttle XPC that's in my living room. Noise is the #1 consideration for this build. I don't need a powerful rig, the current box is an Athlon 2800+ running Linux that does just fine streaming audio to a couple of speakers. It's given me five good years, but the PS is starting to go, it's got an IDE HD, and upgrading it's not cost-effective.

Right now, I'm thinking of building a machine around a Zotac Ionitx A-U board -- I like the idea of the external brick PS, and the Atom 330/Ion combo strikes me as a good tradeoff between low voltage and capability, especially for a machine that's going to do very, very basic things.

Question for those smarter and with more experience -- what the practical upper limits for the system's power consumption to go fanless? I've seen heat pipe designs (HFX micro has a nice case that endpcnoise.com sells, but they're asking a lot for an Intel 945 chipset and no optical drive -- IMHO, anyway) but have no experience with them. Am I safer just sucking it up, getting a quiet fan and a case I can put some dampening in?

Welcome any thoughts...

-- sjr

 

[rodney_ws]

I profess neither to be smarter nor more experienced on this topic... but I will say that I think you're on the right track with the dual-core Atom + Ion combo... that has HTPC written ALL over it. It sounds like you're going for a silent setup... and as with most anything in life, it takes extra effort (i.e. money) to make it that last little bit.

http://techreport.com/articles.x/16340/2

This article mentions the feasibility of a passively cooled Ion system... but they point out that going passive is really dependent on interior volume and venting... something a small enclosure is going to lack. My gut says aim for a single, ultra-quiet fan... and if the one that the system comes with isn't up to your standards, look for a replacement fan to swap out the noisy one. Yes, Ion uses more power than the new Pinetrail Atom platform, but it's still low enough that you should be able to build an ultra-quiet PC around it.

 

[sjr]

My gut says aim for a single, ultra-quiet fan... and if the one that the system comes with isn't up to your standards, look for a replacement fan to swap out the noisy one. Yes, Ion uses more power than the new Pinetrail Atom platform, but it's still low enough that you should be able to build an ultra-quiet PC around it.

Rodney_ws, thanks for the link. I looked at that and am coming to the same conclusion you did. Also saw some incredible Ion case mods here:

http://www.modders-inc.com/modules.php?name=Forums&file=viewforum&f=49

It looks like the finalists all went for fans, and if that's good enough for guys who use machine tools for their hobby...

I'll probably still look at/research passive solutions, at least to educate myself about what's there.

--sjr

 

[a4mula]

I don't know anything about low power consumption htpc's. What I have been digging into lately however is oil cooling. If you don't mind an aquarium filled to the brim with mineral oil sitting in your living room it might provide a perfect medium for a silent pc that will cool much more efficiently than a fan-less design.

Oil Cooled

There are certain things to consider. First you cannot submerge the hdds. Second you have to be careful not to have any cord going directly into the oil because of wicking.

It's an out of the way solution, but one you might dig into. It also has the added benefit of a zero-dust environment for your pc, greater stability due to the equalized heat distribution, and a certain aesthetic appeal.

 

[Houndsteeth]

Unless you are ready for a big mess and a lot of work, stick with air cooling. In your case, you can try to substitute less active cooling with more passive cooling (i.e., a larger heat sink). You still need to have a system that dumps that heat into the environment, so even if you have large heat sinks, you will need to have airflow through those heat sinks to dissipate the heat into the air. Whether this is done using a fan or convective action doesn't really matter, though convection takes a lot more finesse and engineering to work properly.

Water or oil cooling should really only be considered if (a) you want to achieve a better clocking , (b) the case is so small that no means of active air cooling will make the build viable or (c) noise is the overriding factor, and you must have the most efficient cooling solution possible for the least amount of noise.

Stick with passive or active air cooling. For the needs you described, you could very well be able to do this with a large heat sink on the CPU and use a series of fans in push/pull config on the case to keep a steady flow of air across that heat sink, or engineer your case design to take advantage of convection currents, so as the machine gets hotter, more air cycles through the machine to keep it cool.

 

[a4mula]

Unless you are ready for a big mess and a lot of work, stick with air cooling. In your case, you can try to substitute less active cooling with more passive cooling (i.e., a larger heat sink). You still need to have a system that dumps that heat into the environment, so even if you have large heat sinks, you will need to have airflow through those heat sinks to dissipate the heat into the air. Whether this is done using a fan or convective action doesn't really matter, though convection takes a lot more finesse and engineering to work properly.

Water or oil cooling should really only be considered if (a) you want to achieve a better clocking , (b) the case is so small that no means of active air cooling will make the build viable or (c) noise is the overriding factor, and you must have the most efficient cooling solution possible for the least amount of noise.

Stick with passive or active air cooling. For the needs you described, you could very well be able to do this with a large heat sink on the CPU and use a series of fans in push/pull config on the case to keep a steady flow of air across that heat sink, or engineer your case design to take advantage of convection currents, so as the machine gets hotter, more air cycles through the machine to keep it cool.

I'm not really sure where this misconception of "a big mess" or "alot of work" comes from, but it seems to come up all the time when considering this option.

I fail to see how it's alot of work. Water Cooling is alot of work, Oil Cooling? It's easier than setting up air cooling. You drop it in an aquarium and fill it with oil, not exactly rocket science. Stick your HDD on top of your tray and under the lid. If you want to prevent wicking then either a) keep your psu out of the oil, or b) use a mounted power-socket that takes two minutes to install in the lid.

Unless you're a tinkering type, you install, you pour the oil and you forget about it. Where's the mess? As far as using oil as a OC option, then no, you kinda missed that too. Water is about 4x more efficient at dissipating heat. You can use oil as a substitute for a nb/mosfet/memory block as well as evenly distributing any remaining heat as I'm doing, but water will cool the gpu/cpu. Today's high-end parts produce far too many watts to even consider passive oil cooling.

What oil does give you is the ability to run low watt systems that might be too large for a mere heatsink. It also gives you a system that is going to be much more stable and extend lifespan by evenly distributing heat. Can an passive aircooled system claim that? All of this and it's maintenance free, no dust screens, no air cans, no bleeding lines or changing of water. Not to mention the aesthetic value, which admittedly is quite relative to one's own taste. Most of all, it'll give the OP his primary objective: silence.

 

[Paperdoc]

We built a TV server machine recently using a small case but did stick with a fan-cooled AMD CPU system. To be as quiet as possible, we substituted for the stock CPU fan one by Noctua; we also used the same fan design as the main case cooling unit, and let the mobo control fan speeds. The vid card is by XFX based on an ATI 4350 system with passive cooling only. With everything off in the Living Room you can hear the fans gently as it continues to download files from the 'net. But turn on the TV and watch something, and you'll never hear the computer.

 

[Houndsteeth]

I'm not really sure where this misconception of "a big mess" or "alot of work" comes from, but it seems to come up all the time when considering this option.

I fail to see how it's alot of work. Water Cooling is alot of work, Oil Cooling? It's easier than setting up air cooling. You drop it in an aquarium and fill it with oil, not exactly rocket science. Stick your HDD on top of your tray and under the lid. If you want to prevent wicking then either a) keep your psu out of the oil, or b) use a mounted power-socket that takes two minutes to install in the lid.

Unless you're a tinkering type, you install, you pour the oil and you forget about it. Where's the mess? As far as using oil as a OC option, then no, you kinda missed that too. Water is about 4x more efficient at dissipating heat. You can use oil as a substitute for a nb/mosfet/memory block as well as evenly distributing any remaining heat as I'm doing, but water will cool the gpu/cpu. Today's high-end parts produce far too many watts to even consider passive oil cooling.

What oil does give you is the ability to run low watt systems that might be too large for a mere heatsink. It also gives you a system that is going to be much more stable and extend lifespan by evenly distributing heat. Can an passive aircooled system claim that? All of this and it's maintenance free, no dust screens, no air cans, no bleeding lines or changing of water. Not to mention the aesthetic value, which admittedly is quite relative to one's own taste. Most of all, it'll give the OP his primary objective: silence.

1) Trust me. Oil is messy. No way around it. If you have to pull anything out of the oil, you have to give it space/time to let the oil drain away before you can work with it. Wicking aside, unless you seal your system, you will eventually end up with oil on the floor. Hopefully you don't have carpet like I did. And yes, the carpet discolored where the oil landed, and no amount of shampooing can get all the oil out. If you do it, seal your system or do it on a non-permeable floor surface.

2) Oil will destroy carbon-based plastics and rubbers. This is a fact, not controversial. Some plastics will become extremely hard and brittle when oils remove the esthers that make them pliable and "plastic." Granted, it takes some time, but this is accelerated since the oil is being heated. The PCI risers on the board I was working with became so brittle that the act of pulling the mainboard out of the oil cracked the plastic riser that supported the video card. Superglue fixed the crack, but this is something to be aware of.

3) You have to cool the oil, using some form of heat transfer. Some of the heat is lost to the environment through the glass sides of the "case," but you cannot expect the oil to be a continual heat sink that magically makes all the heat disappear. If you look at the large transformers that electrical companies use, the large coils are immersed in oil-filled canisters that have large fins to allow for passive air cooling. The oil is nothing more than a dense dielectric medium between the metal coils, metal housing and eventually the environment. If you expect to run this machine continually, you will have to make sure that the heat going into the system does not overload the system's capacity to dissipate heat energy. Hence, more active or passive cooling of some kind. Granted, an oil system that continually gains heat energy will take longer to reach failure temperatures for the electronic components than most any other system, but failure is failure, and a slow one can be the most insidious (i.e., the "How to boil a lobster or frog" saw).

4) Water cooled systems are stilled air cooled in the end. The water cooling loop is nothing more than a more efficient heat energy transport loop. You still need to have enough radiator capacity to dump that heat energy back into the environment. The larger the radiator, the larger your thermal dump, and your system can handle more potential heat energy without compromising your components. All the water cooling loop does is allow you to concentrate that heat transfer to a single location instead of multiple locations. In the end, you are pushing just as much air as you do using air cooling alone, but you are using larger fans at lower RPMs, thus generating less noise.

Before you get up in arms...yes, I water cool...yes, I have had my oil-immersion experiments...and yes, I have done subambient cooling, both phase-change and dry ice slushbox. The only thing I haven't tried is thermoelectric (too inefficient) and LN2 (too dangerous).

 

[a4mula]

1) Trust me. Oil is messy. No way around it. If you have to pull anything out of the oil, you have to give it space/time to let the oil drain away before you can work with it. Wicking aside, unless you seal your system, you will eventually end up with oil on the floor. Hopefully you don't have carpet like I did. And yes, the carpet discolored where the oil landed, and no amount of shampooing can get all the oil out. If you do it, seal your system or do it on a non-permeable floor surface.

People are messy. Oil sits in its aquarium and does nothing. If you tackle wicking issues from the start and avoid something silly like adding an aerator for "liquid flow" aka bubbles (foam) then there's nothing that's going to will the oil to jump up out of its place. He's building an HTPC, not a tinker toy. What exactly do you think he'd have to pull out of the oil?

2) Oil will destroy carbon-based plastics and rubbers. This is a fact, not controversial. Some plastics will become extremely hard and brittle when oils remove the esthers that make them pliable and "plastic." Granted, it takes some time, but this is accelerated since the oil is being heated. The PCI risers on the board I was working with became so brittle that the act of pulling the mainboard out of the oil cracked the plastic riser that supported the video card. Superglue fixed the crack, but this is something to be aware of.

Fallacy number 2, right behind "it's messy". I'm fairly certain this little yarn was spun directly here at Tom's after they used cooking oil to cool with and made an assumption that the fatty acids in the oil could have a detrimental effect on the rubber and plastics. Maybe it would, maybe it wouldn't but that's a moot point since you'd be a fool to use cooking oil either way. Here's a fact, mineral oil is used as a rubber life extender. Transformers have been using mineral oil cooling solutions for decades with absolutely no issues with their seals.

3) You have to cool the oil, using some form of heat transfer. Some of the heat is lost to the environment through the glass sides of the "case," but you cannot expect the oil to be a continual heat sink that magically makes all the heat disappear. If you look at the large transformers that electrical companies use, the large coils are immersed in oil-filled canisters that have large fins to allow for passive air cooling. The oil is nothing more than a dense dielectric medium between the metal coils, metal housing and eventually the environment. If you expect to run this machine continually, you will have to make sure that the heat going into the system does not overload the system's capacity to dissipate heat energy. Hence, more active or passive cooling of some kind. Granted, an oil system that continually gains heat energy will take longer to reach failure temperatures for the electronic components than most any other system, but failure is failure, and a slow one can be the most insidious (i.e., the "How to boil a lobster or frog" saw).

Actually you can assuming you're running with a low enough wattage. Again I point to the Puget system. For going on two years now they've run a oil pc with no active cooling whatsoever. After about 12 hours at full load (stress testing) it reached a maximum temp of about 80c at which point it reached equilibrium with ambient temps and remains stable. It's idle stable times were 37c. Their pc has never once reached "failure temperatures for the electronic components". I'm not recommending this for an i7 system, we're talking about a low wattage HTPC.

4) Water cooled systems are stilled air cooled in the end. The water cooling loop is nothing more than a more efficient heat energy transport loop. You still need to have enough radiator capacity to dump that heat energy back into the environment. The larger the radiator, the larger your thermal dump, and your system can handle more potential heat energy without compromising your components. All the water cooling loop does is allow you to concentrate that heat transfer to a single location instead of multiple locations. In the end, you are pushing just as much air as you do using air cooling alone, but you are using larger fans at lower RPMs, thus generating less noise.

The reason water cooled systems are more effective has nothing to do with just moving around the "air". It has to do with thermal conductivity of water being 4x greater than that of air. Then the thermal conductivity of copper (rad) being 300x greater than air. Air is introduced as the final link of the heat exchange, not because it's adept at it, but because it's basically free. Generally speaking the greater density of an object the more effective it becomes at absorbing heat. Air is not a prerequisite for heat absorption or transfer. Here's a great starting point for understanding the basics of thermal dissipation and conductivity.

Before you get up in arms...yes, I water cool...yes, I have had my oil-immersion experiments...and yes, I have done subambient cooling, both phase-change and dry ice slushbox. The only thing I haven't tried is thermoelectric (too inefficient) and LN2 (too dangerous).

I don't doubt that you've done these things. What I do call into doubt however is the authoritative manner in which you state your suppositions about this option. You're deliberately making statements that have no basis in reality thus leading the OP to an misinformed decision one way or the other.

 

[sjr]

We built a TV server machine recently using a small case but did stick with a fan-cooled AMD CPU system. To be as quiet as possible, we substituted for the stock CPU fan one by Noctua; we also used the same fan design as the main case cooling unit, and let the mobo control fan speeds. The vid card is by XFX based on an ATI 4350 system with passive cooling only. With everything off in the Living Room you can hear the fans gently as it continues to download files from the 'net. But turn on the TV and watch something, and you'll never hear the computer.

@paperdoc:

Sounds like we were shooting for more or less the same thing.

Slow progress so far -- I've ordered an SSD from newegg for the system, (Kingston SSD Now V 64GB - $105 on sale and with a rebate -- assuming that ever comes through -- Item#:N82E16820139006) but am still debating other options.

First option is putting the box on the back of the small TV in the living room. For this, I was looking at a case built for easy convection cooling (http://www.mini-box.com/M350-universal-mini-itx-enclosure). That should take the Zotac Ion board. I probably will replace the stock fan -- if needed at all -- with a Scythe Mini Kaze or similar.

Second option is going fanless with a different enclosure (http://www.logicsupply.com/products/gs_l05) and using an Intel mobo with the Atom CPU but the 945 chipset. It's lamer from the feature point of view, but the cheaper mobo somewhat offsets the much more expensive case.

Neither option is going anywhere until I pay the Christmas credit card bills, though.

@a4mula:

Thanks for the Puget Systems link. I love the idea, and have a buddy who's building one and passed him the site. Two downsides: the aquarium kit is way cheaper than I'd expected, but my budget would still force me to stretch out the build over a longer period than I'd like. The other thing is the SO, who I don't see signing off on a couple of gallons of mineral oil in the living room. (Not that any of my projects haven't all worked perfectly right the first time, or anything...)

Will update when I've gotten off my duff and put something together.

--sjr

 

[sjr]

OK, so first cut at this is up and running.

I wound up going with the new Intel D510MO motherboard (Tom's review here: http://www.tomshardware.com/reviews/atom-d510-d510mo,2507.html) rather than the Zotac Ion board I'd been considering. My first priority was silence and getting as close to fanless as I can, so the D510MO's low wattage came out ahead of the extra muscle from the Ion. I went with a Kingston SSDNow V 64GB SSD, 4GB Kingston low profile RAM, and a Silverstone SG06B case.

I set the fan to variable speed in the BIOS and the box is next to silent -- you have to almost touch an ear to the fascia to hear the case fan running. Running UNR 9.04 which I installed from a USB drive b/c I can't convince the mobo to boot from an external CD/DVD drive. The Intel graphics apparently max out at 1024x768, which is irritating but not fatal. Waiting to see if the Intel support forums offer any suggestions on those.

Next step is to replace the default power supply (300W 80+ deal) with a pico/power brick combination. The D510MO draws relatively little power, so I'm not sure this will offload enough heat to matter, but I want to see if I can make it work. Will also get an inexpensive slimline drive next week and see if that solves the booting issue.