Priorities for my next PC are:

1. High power efficiency. I do a lot of DivX encoding, which brought up some interesting power efficiency questions I asked in Lowest power consumption for DivX encoding. That thread and the Deep UnderVolt/Clock article introduced to me the possibility of changing FSB/multiplier/voltage from windows, allowing me to scale the CPU power consumption to fit the load on it. I don't necessarily have to do this automatically with 8rdavcore, though: anything that gives me a wide MHz range from windows will do (CPUMSR, proprietary programs).

2. Small size. My girlfriend and I share a < 400 sq. ft. studio apartment, so space is at a premium. After buying a new 17" widescreen LCD, I noticed four mounting holes on the back that could easily support a slim PC (they're intended for a VESA wall mount). With a little ingenuity, I might even be able to mount my Klipsch satellites to the monitor, supporting the whole system (minus subwoofer) on the ~12"x4" LCD base. I don't think the case could be any bigger than 18"W x 13"D x 5"H.

3. Quiet operation. I have to be able to sleep while the computer's encoding DivX in the same room. I guess I can't quantify this any more than to say that a mild "shhhhh" sort of sound at a distance of about 10' is my personal noise ceiling.

4. Simple styling. No bling necessary (or wanted) here.

5. Gaming prowess. I only play games occasionally, and usually old ones at that.

I'll use a laptop hard drive for the system drive and hope to set up a desktop hard drive and a laptop dvd drive through firewire so I can switch them on and off as necessary to minimize power consumption & noise (see How to switch drives on & off). So now to find a case that can fit all of these components onto the back of the LCD. Cheapest I could find is the Silverstone Lascala SST-LC02 (~$150 shipped). This case can actually accept full ATX motherboards and has two PCI slots, but the best motherboard option seems to be an all-in-one microATX board using the nForce2 chipset, which would leave me with some space in the case (for the firewire bridge?). So here's my proposed system:

Case: Silverstone Lascala SST-LC02
Motherboard: Shuttle MN31N
Processor: AMD Athlon XP 1700+ Thoroughbred
Memory: 2x Kingston 256MB DDR PC2700 CAS2.5
System HDD: Fujitsu MHT2060AT 2.5" 60GB
Storage HDD: Maxtor Diamondmax 9 160GB
DVD: Panasonic CW-8123-B laptop drive
Firewire bridge: Two-device Firewire 400 (1394a)

Questions:

1) How much will I need to do to make this reasonably quiet? The PSU has a 60mm fan, and there's a spot for an 80mm directly above the CPU. I should be able to find space to suspend the laptop drive, and I anticipate using the Maxtor rarely enough that I won't care about the noise.

2) What are the chances that I'll be able to get 8rdavcore or CPUMSR to work on the Shuttle board? 8rdavcore supports several Shuttle boards, including the ATX AN35N - close enough? What about CPUMSR - does it work on most nForce2 boards?

3) Any other comments on the setup?

Thanks for the review!

Here are some pictures of my current setup:



Hopefully this shows why I'd like to bolt the PC to the back of the LCD - total system depth would be less than 8"!

Any help or comments? Anyone?

(edit: picture links)

rather than mess with bolting stuff onto your display, why not replace the books holding your monitor with a PC? And, if the black box on the bottom shelf is a sub, getting a smaller pc won't help that much.

but who am i to rain on your ambition?

Fancontrol has a good point there. I wouldn't get obsessed about hooking it on the back of the LCD either. If that's your ambition, you might be better off to buy all-in-one PCs of that config from the likes of Gateway, Dell, etc... tho you might have to hack with them to make them really quiet...

Anyway, going to your choices:
-- The PSU fan size is not promising
-- There's no guarantee that the PSU is high efficiency
-- The idle noise of the maxtor will drown out any notebook HDD silencing efforts. Porbably best to run it off a USB2 external box and hang it in a closet.

fancontrol: The books are no more than 10" deep, and the only way I could build a computer that small is by going with a mini-ITX and losing DivX encoding speed. You can see in the first picture that my current PC (in flexATX case) fits very well on that lower shelf next to the subwoofer. But I can't squeeze a microATX mobo in that case, and I can't find any other cases that small that can fit one either. Also, I'm considering building a little cable management box to go under the LCD that I could hide extra cabling in.

MikeC: I'm definitely concerned about the PSU. It should supply plenty of power, and it does claim to have PFC (active or passive?), but you're right, efficiency is a big question mark. This review claims that it's the same PSU found in the Shuttle SN85G4, which is apparently not the same size as the PC40 SilentX 250W. If it's too loud, maybe I could mount a panaflo to the back exhause to pull more air with less noise. At some point, I also hope to have this whole system running off a solar-panel-charged 12V battery, in which case I'd probably end up using an Opus 150W DC-DC power supply (90% efficiency).

(edit: added paragraph) As I said in the post, the Maxtor would be mounted internally, but connected via firewire so I could switch it off when not in use (~99% of the time). I plan to use it only for mass storage - i.e., turn it on, copy a movie to the laptop drive, turn it off again, play the movie. For ~5 minutes at a time, I honestly don't care how loud the thing is unless I can't carry on a phone conversation while it's on (doubtful).(/edit)

I found a Thermalright SLK-900 for cheap, which should help me get by with the lower airflow available in this kind of case. Hopefully there will be enough height in the case to mount it with an 80mm japanaflo.

Since I'm still hoping to mount the case vertically, should I be concerned about pockets of hot air rising to the top (front) of the case?

So I decided to give it a go with the parts above. Everything is here but the motherboard, and now I'm thinking ahead about fan placement & airflow. I'm hoping that mounting this case vertically (as it looks below) will actually help the cooling, since convection will work to my advantage (the SLK-900A fins are even vertical). The following seems like a good airflow pattern to me, with cool air in blue and warm air in red:


The blue arrow to the CPU is actually coming through a vent in the top (now side) of the case.

The problem is getting hot air from the PSU to the front (now top) 80mm panaflo. This PSU was obviously designed to blow warm air out of the back (now bottom) of the case, which makes great sense when it's horizontal. In this vertical orientation, though, the panaflo covers up the vents on the front (now top) - see the yellow arrow in this picture:



So, my questions:

1. Is this airflow pattern promising?

2. With some additional vent holes in the PSU near the front (now top) panaflo, do you think the PSU would stay cool enough to remove that 60mm fan on its side? Maybe I could remove the PSU cover to get better airflow?

3. Should I try to find a way to mount the front (now top) panaflo away from the side of the case, where it won't obstruct the existing PSU vent? I'm thinking where some of those power switch cables are in the bottom picture.

Thanks for any help; I'm psyched to assemble the system and see how quiet I can get it in such a small case!

(edit: picture links)

I think you are in for a though struggle.
You want high performance, small case and low noise.
These things don't combine easilly. And it seems that you want to combine it with low cost. That makes it even harder.

Here are a few problems that I see:
1. Your case has far from optimal airflow. A first step might be to remove/replace fan grills with something less restrictive.

2. That PSU is far from my choise for a quiet PSU. It that the version with a 40mm fan at the back and a 60mm fan in the side? Both with rather restrictive grills? And the PSU still gets very hot during normal operation?
Pushing the hot air from that PSU as the intake air for you CPU does not sound like a good move.

3. The best way to get a silent CPU is to make it low wattage. The easiest way to make it low watt is to have a fast CPU that you underclock and undervolt. Preferably start with a multiplier unlocked CPU and a motherboard that is good for undervolting.
The XP 1700+ is unlocked, so that is good. Unfortunately almost no mATX boards are undervoltable, and undervolting is what makes most difference. The underclocking is mostly to enable much more undervolting.

Do you have any space to put dampening materials in that case? I think you could benefit quite a bit from some.

You might also want to think about structural rigidity. Putting the case at the back of the LCD might be a good idea but bolting it on is IMO a recipe for disaster. Those LCD stands aren't built to take so much weight, I'm not going into center-of-mass at all.

silvervarg - Thanks for the feedback. I definitely plan on opening up existing fan grills a bit, especially because the two stock fan locations are meant for 60mm's (there's enough room for 80mm's though).

The PSU only has the one 60mm fan on the side. The back (now bottom) grill is fairly open but somewhat small; the ones on the side and front (now top) aren't too great, so I'll work on them some. What I was hoping to do is remove the 60mm fan on the PSU, maybe cover the hole, and have hot air from the CPU & PSU meet in the front (top) and be expelled by that panaflo. I don't know if that level of cooling will be sufficient for the PSU, though - maybe going ultra-low power (switched drives, laptop components) will keep it from working too hard.

On to undervolting. Several Shuttle SFFs and mobos have been reported as undervoltable in the sticky thread (SN61G2, SN45G, AN35N), and I've heard of the SN45G, AN35N, and SN41G2V2 working with 8rdavcore, so I'm hoping I can use 8rdavcore with my MN31N to underclock/undervolt my unlocked 1700+ to ~600MHz (or lower) at ~1.1-1.2V (like the Deep UnderVolt/Clock article), where it will stay most of the time.

I get the impression that the two panaflos will be sufficient cooling for sub-GHz running, but I have no idea how high I'll be able to go before things start crapping out. If I need more airflow, maybe I could take off both card slot covers (for an intake) and add another panaflo exhaust? (there's another fan opening in the side opposite the other one, near the front -now top)

As for damping, I'll be throwing in little bits of mousepad material wherever I can to minimize vibration transmission, but I seriously doubt I'll have anywhere near the space I'd need for AcoustiPack. With laptop components and fan isolators, do you think there will be that much vibration to damp?

burcakb - I'm pretty sure it will be fine. The LCD weighs 15 lbs, so the VESA mount should be able to support well over that amount. The case & PSU weigh ~6 lbs., so I can't imagine going over 15 with a motherboard, SLK-900A, and two or three panaflos. The LCD itself is currently supported by its proprietary stand, made of very thick steel, which should be able to support everything together. If center of mass is an issue, I can add lead weight to the base.

There's certainly lots of things that might not go as I'd planned with this, and it will be a challenge working in such a small space, but I'm pretty confident I can come up with something that will work for me.

Linus, it seems that you have though a bit about this. From your earlier notes I thought that you maybe got in to this a bit hasty, but now I feel you have a good chance to get this to work fine.

It is hard to tell how things will work out in reality, but I have a few suggestions that is likely to work fine.
Assuming your motherboard allows multiplier change and lowering of vcore and you want to run really silent:
Start with multiplier 5 and 133 FSB. By keeping the FSB high you maintain speed on things that are not entierly CPU bound. By lowering the multiplier to 5 you should be able to drop vcore to 1.1-1.2V. There is a huge difference between 1.1 and 1.2V for stability. When you try to find the lowest vcore I suggest that you follow Ralf Hutter's advice. Start very low. E.g. 1.1vcore. You might not even get your board to post, but that is the idea. If it really fails if was a very quick test. Bump up vcore one step at a time until it boots. Then test stability with prime95. Once you find something that pass prime95 with zero errors you either stop there or increase vcore one more step so you have some margin.

For the airflow I would suggest that you try to have rather slow fans blowing are out both to the right and the left. Block all air intakes except through the PSU. Either have the PSU fan run rather slow or remove it compleately. Make PSU grills less restrictive if possible. Seal any major gaps in the case. Almost all air intake should go through the PSU, and there should be a decent underpressure in the case.

I also suggest that you put something to dampen sound just above the CPU (on the inside of your case panel). This will at least force noise from CPU fan to turn a few times before it can escape out of the case. By doing this you should be able to have the CPU fan at low speed without becoming disturbed too much. This should make it possible to run CPU at quite high speed. Experience from posted undervolts usually suggest that if you can just sacrifice 20% speed by underclocking you can often undervolt quite far. This seems to be true for most CPU cores, both AMD and Intel.

Thanks for the underclock/volt advice. I can't wait to get my mobo so I can see what the chip can do. Just about anything will be an upgrade from my Coppermine Celeron (850 MHz).

I'm not sure if I mentioned this before, but there is an air intake in the top (now side) of the case, more-or-less directly above the CPU. So sound damping right there is pretty much out of the question. However, once all is mounted to the back of my LCD, this will be pointed directly away from the user, with the LCD providing some natural attenuation. Hopefully that will help.

In the meantime, I like your idea of having more exhaust fans than intake to pull as much air as possible through the PSU. Once I assemble everything and play around a bit, I'll consider different methods of speed control to fine-tune things.

Thanks a bunch for your knowledgeable advice; I'll post back once I've got everything and experimented a bit!

Well, after several hurdles and hardware changes, my project is looking like it might work out after all.

The Shuttle MN31N motherboard didn't have any provision for multiplier or Vcore adjustments, and 8rdavcore didn't work with it. So I replaced it with a Biostar M7NCG 400, knowing that I could at least have BIOS adjustments if 8rdavcore didn't work, and 8rdavcore works (see New 8rdavcore release - more boards supported?)! I fried my Tbred 1700+ in a moment of stupidity, but that gave me an excuse to get the new Mobile Barton 2200+ 1.35v, which is quite the performer. It boots into windows with BIOS settings anywhere from 5x100 @ 1.3v to 11x200 @ 1.75v, and preliminary testing with 8rdavcore indicates Prime95 stability down to 5x60 @ 1.225v and 11x125 @ 1.225v after booting with the settings above. Wattage varies from ~40W @ 300MHz to ~120W running Prime95 at 2200MHz. What a range!

The testing above was done with the following hardware:

Case: Silverstone Lascala SST-LC02, still horizontal for now
Motherboard: Biostar M7NCG 400
Processor: AMD Athlon XP 2200+ 1.35v Barton
Memory: 2x Kingston 256MB DDR PC3200 CAS3, running at DDR333 & CAS2.5
System HDD: Fujitsu MHT2060AH 2.5" 60GB
Airflow: one 60mm blowing hot air from the PSU to the CPU where a 92mm Chinaflo @ 12v blows out the top (eventually back) of the case
(the storage HDD & optical drive will be connected via firewire, & off most of the time)

Note: it isn't quiet yet. The next step is turning the system vertical, unplugging the PSU fan, installing an 80mm Japanaflo as in the pictures above, and seeing how far I can undervolt the fans without affecting stability. Then I'll think about opening up grills and whatnot.

Just realized I never linked this thread to the General Gallery thread of the final result: Small & Flexible: Back-of-LCD variable power consumption.