My one-fan Athlon64

frostedflakes · Post by **frostedflakes** » Thu Jun 16, 2005 8:56 pm

Nothing too exiting yet, but check back soon - I hope to have some cool stuff.

Antec SLK3700-BQE
Seasonic S12-330 rev. A1
Mobile Athlon64 3700+ "San Diego"
Scythe NCU-2005
DFI Lanparty UT nF3 250Gb
2x512MB GeIL UltraX PC3200
XFX GeForce4 MX4000 64MB PCI
Seagate Barracuda IV 40GB
Rosewill 52x32x52x16 combo drive

Power consumption from the wall while folding: 82.11w.

With a low-power Turion64, I expect this can be cut down close to 50w.

The fan in the S12 barely revs up at all. While browsing the net it's sitting happy at ~770RPM. However, this 2.4GHz 3700+ gets the system damn hot while stressed. The CPU itself stays at a reasonable temp (~53*C), but the PWM IC gets warm enough (55-56*C) to cause instability. The top of my case actually gets very hot as well, about 7*C above ambient in the worst spot. This S12 seems to handle the load well, though - the internal fan only speeds up to ~800RPM.

Currently, there are four "hot spots" I'd like to get under control: Motherboard chipset, PWM IC, RAM, and HDD. 2x512MB TwinMOS BH-5 and a Toshiba MK4025GAS are on the way, so I should be able to check RAM and HDD off the list. I also think I'll switch to a 25w Turion64 soon, which should bring PWM IC temps down to an acceptable level, as there will be less current and voltage going through it. I still need to figure out what to do about the motherboard. Possibly an upgrade, to Via or another cooler-running chipset.

Also planned is a duct from the CPU heatsink to the S12 intake. I have some acrylic bought, and plan to build a shroud around the NCU-2005. From there I don't know, but I'll be sure to post pictures once I figure it out.

Comments/suggestions are welcome.

atomidude · Post by **atomidude** » Thu Jun 16, 2005 10:12 pm

glad to see there are more and more ppl here who think quiet mean not only good fans, but also less fans

hope you don't mind the criticism:
* you reduced the cooling to one fan only, but using the wrong one

the worst thing to do, imho, is to cool using the psu fan. that's because the psu generates a lot of heat itself, then circulating heated air from cpu through it...
* you should cut a blow hole at the top, above the psu, so the hot air can bleed there
* don't block the back venting holes, those could help actually, heated air from the cpu area would escape there, too
* blowing air is much more efficient than sucking, talking about case fans now, not hsf fans. try installing a quiet 120mm @700-800rpm on the side cover, just slightly under the cpu HSF, and watch your system getting much, much cooler. that fan would also cool down the vga and the mobos chipset
if you are concerned about dust, you could use a filter. however, a slow fan won't cause too much trouble. i clean my rif every 3-4 months, dust was never really an issue

hope all this would help, didn't mean in any way to offend

frostedflakes · Post by **frostedflakes** » Thu Jun 16, 2005 10:31 pm

I'd rather not cut any holes in my case.

I know cooling with the PSU fan isn't ideal, but it seemed like a good choice for my particular situation. I look at it like this: The power supply is rated for 330w, and at this power draw it should be producing ~66w of heat. So the power supply has surely been engineered to handle this heat plus whatever the 120mm fan happens to suck up from the other system components. So we'll assume that it's good for about 100w at least. Should be able to handle its own heat as well as that of the other components. My only fear was that the fan would ramp up like crazy, but difference between idle and load RPM is almost nothing.

atomidude · Post by **atomidude** » Thu Jun 16, 2005 10:54 pm

hm, i'm afraid your vga and mobo components (chipset, mosfets) will run hot, because the 120mm fan in the psu won't suck too much air. the impedance must be pretty high, the airflow is not ideal, passing through the psu... make sure you monitor the temperatures closely, using software like speedfan or similar, to avoid killing your system. good luck!

frostedflakes · Post by **frostedflakes** » Thu Jun 16, 2005 11:06 pm

Not too concerned about VGA. It's quite cool, but then again it's a measly MX4000, so I wouldn't expect it to get very warm.

Chipset and motherboard components are a concern, though. I'm sure it would be a good idea to do the finger test on some of the MOSFETs/ICs and heatsink any that give off a good amount of heat. Then again, I've looked at quite a few rigs here in the gallery section, and have seen people running similar components fine under what I would consider to be nearly as stressful conditions, so who knows.

m0002a · Post by **m0002a** » Thu Jun 16, 2005 11:20 pm

If you use a Nexus 120mm case exhaust fan at about 7V, I doubt you will hear it. I run mine at 12V and the noise is barely noticable.

frostedflakes · Post by **frostedflakes** » Thu Jun 16, 2005 11:29 pm

I actually tried this (Yate Loon low-RPM sleeve at 5v, basically the same thing), and it seemed to F my airflow all up. As crazy as it sounds, I am getting slightly lower temps with the rear exhaust hole taped off and the power supply as the only exhaust.

dago · Post by **dago** » Thu Jun 16, 2005 11:29 pm

atomidude wrote: * you reduced the cooling to one fan only, but using the wrong one the worst thing to do, imho, is to cool using the psu fan. that's because the psu generates a lot of heat itself, then circulating heated air from cpu through it...

According to the SPCR test, efficiency for this PSU at the given wattage (82W) should be around 80%, that means that the PSU itself would radiate about 16W, while the CPU should talke most of the rest (say 50W).
So the CPU is the warmer component and that's effectively the one that should get the cooler air.

Or did I made a mistake somewhere ?

frostedflakes · Post by **frostedflakes** » Thu Jun 16, 2005 11:35 pm

This is my logic also. At such a low power draw the PSU should be giving off very little heat. Power consumption on the DC side is probably ~65w. The CPU surely accounts for the majority of this, although I doubt it is 50w. My guess would probably be more like 35-40w. With an undervolted Turion64, I bet I can get CPU power consumption down to ~15w.

alleycat · Post by **alleycat** » Fri Jun 17, 2005 1:07 am

Neat project. I'd like to make a few (hopefully helpful) comments. I know that you don't want to cut any holes, but I feel that it is essential to remove the intake grille at the front of the case. You'd be amazed at how easy it is to cut holes with Wiss tin snips (see sticky in cases forum). By ensuring an easy airflow path, you can more effectively direct the cool air, rather than letting it leak in through other gaps. Make sure the rest of your case is well sealed (sorry atomidude, but the rear vent should stay sealed also).

I don't think you need to make a duct because CPU temps aren't bad, and the last thing you want to do is deprive the RAM and motherboard of moving air. It could be that the size and shape of that HS blocks air movement in certain directions, as evidenced by your experiment with an additional fan.

I know that 120mm fans are the fashion at SPCR, but I think they move too much air for a truly inaudible system. At low speeds it's not the fan motor you can hear, but air movement. Using a 120mm fan, a small increase in fan speed increases air movement quite a bit. That's why I'm not so keen on 120mm one-fan systems. I believe you get better results with two 80mm fans at low speed, then you can better balance where the air goes. If done properly, the system will be quieter - isn't that what it's all about? At a given speed, two 80mm fans move slightly less air than a single 120mm fan. The increase in noise from having two fans compared to one is negligible. As I said before, the noise of air movement drowns out any motor noise (assuming quality fans such as Nexus). I know this probably doesn't help you much, but it might give others something to ponder when planning a new system.

len509 · Post by **len509** » Fri Jun 17, 2005 1:15 am

frostedflakes wrote:This is my logic also. At such a low power draw the PSU should be giving off very little heat. Power consumption on the DC side is probably ~65w. The CPU surely accounts for the majority of this, although I doubt it is 50w. My guess would probably be more like 35-40w. With an undervolted Turion64, I bet I can get CPU power consumption down to ~15w.

You should go ahead and try the undervolting. As this motherboard does it very well. By the way, I think you meant to say your CPU core is a ClawHammer/Hammer. The San Diego is a Socket 939 processor.

frostedflakes · Post by **frostedflakes** » Fri Jun 17, 2005 4:51 am

It is in fact a San Diego. This is one of the new 90nm 62w mobiles. Not a CH core. And it is actually already undervolted to 1.3v, which is as low as I can undervolt it at 2.4GHz.

alleycat: Thanks for the suggestion. However, I don't think this can get much quieter. From a couple feet away, the power supply fan at ~800RPM is inaudible. I really have to put my ear up next to the power supply grill to hear this, and even then, it is mostly a faint "clicking" from the Yate Loon's motor.

alleycat · Post by **alleycat** » Fri Jun 17, 2005 5:59 am

I'm sure your system is quiet! My main concern is that air is not circulating properly, resulting in the hot spots you mention. Hopefully you will be able to resolve this.

I made the additional comments about fans for the benefit of anyone else considering a one-fan system, ie it's not simply a case of less fans = better. I didn't expect that at this stage you would want to redesign your entire system!

frostedflakes · Post by **frostedflakes** » Sun Jun 19, 2005 10:57 pm

I don't think many systems are candidates for single-fan cooling. I agree with you - in certain cases, two low-speed fans would be more acoustically efficient than a single fan that was forced to ramp up to higher RPMs. However, many of the components in my computer are low-power, so I believe a single-fan solution is very doable.

Keep in mind that all the earlier testing was done with a 3700+ mobile at 2.4GHz. Even undervolted, this chip pumps out a good amount of heat. However, when underclocked/undervolted to 1.8GHz/1.2v, temps drop across the board. While folding CPU stays at 42*C and PWM IC at 48*C (compared to 53*C/56*C at 2.4GHz/1.3v). Even at the same speed/voltage, the Turion64 should consume less power. Not to mention it can be undervolted stable to 1.1v (from 1.2v default).

EDIT: I've been thinking, and wonder if an intake duct would be a more effective solution. CPU temps are not really a problem, so what would I benefit from a CPU duct? However, components of the motherboard, such as the chipset and VR circuitry seems to get rather warm.

So I was thinking, why not have a sealed tube that goes from the front intake of the case and stops near the edge of the motherboard. With the location of the nForce3 chipset on the this board, air will be directed almost right onto the chipset heatsink. From there it would basically be sucked straight up, over a majority of the motherboard components and through the CPU cooler.

This low-heat laptop hard drive gives me much more flexibility when mounting than I would've had with this Barracuda IV. I picked up a SilentDrive real cheap on eBay, so I plan to throw the drive in there and mount it up under my optical drive. This way, I can rip out the entire 3.5"/floppy cages, leaving me plenty of room to build a duct.

Sound like a much better plan?

frostedflakes · Post by **frostedflakes** » Wed Jun 22, 2005 11:19 am

Here's a crude drawing of what I am planning.

Comments, suggestions, ideas?

IsaacKuo · Post by **IsaacKuo** » Wed Jun 22, 2005 11:41 am

I'd recommend fashioning a duct from the CPU heatsink to the PSU. Otherwise, only a fraction of the air being sucked into the PSU will come from the CPU heatsink. The gap between your PSU and CPU heatsink is rather large and it's offset. I'd guesstimate that only a third of the air passing through the PSU passed through the CPU heatsink first.

Now, I have another more interesting idea--flip the PSU fan so the airflow is the reverse of what you're proposing. This will give your efficient PSU nice cool fresh air. It's efficient, so it won't heat the air up that much. Therefore, the air it blows at the CPU heatsink won't be too much warmer than the fresh air from outside.

frostedflakes · Post by **frostedflakes** » Wed Jun 22, 2005 12:01 pm

If CPU temps were a problem I'd duct the heatsink to the power supply, but they aren't. I've been folding non-stop for the last couple days, and the CPU is at 39*C (Turion64 MT-34 undervolted to 1.1v).

As crazy as it sounds, I believe the hottest part of my system is the nForce3 chipset, along with a few MOSFETs and ICs along the motherboard. This is why I was considering this intake duct, as it should direct airflow directly over these components.

IsaacKuo · Post by **IsaacKuo** » Wed Jun 22, 2005 12:30 pm

frostedflakes wrote:As crazy as it sounds, I believe the hottest part of my system is the nForce3 chipset, along with a few MOSFETs and ICs along the motherboard. This is why I was considering this intake duct, as it should direct airflow directly over these components.

If that's really the case, then your ducting plan is BAD. You'll just end up heating up everything else if you put your hottest component "first in line" via the duct. As it is, a lot of incoming air simply never passes nearby the hot chipset.

I therefore have to go with my latter, more radical suggestion. Create the duct you're thinking of, but flip the PSU fan. With your efficient PSU and cool CPU, your case temperatures will not suffer much from reversing the airflow. At the same time, the duct will direct exhaust air past the hot chipset. This will cool your hot chipset and exhaust the hot air directly out of the case.

edit:

Actually, now that I think about it, the above may not make sense. The way things are right now, the heat generated by the chipset has to go SOMEWHERE, and I think the only place it can ultimately go is out through the PSU. Therefore, the heat from the chipset heats the air going through the PSU one way or another either way. Your ducting plan might heat up the CPU slightly, but of course that isn't a problem.

So I think your duct plan could work with airflow in either direction.

dago · Post by **dago** » Wed Jun 22, 2005 12:49 pm

IsaacKuo wrote:
frostedflakes wrote:As crazy as it sounds, I believe the hottest part of my system is the nForce3 chipset, along with a few MOSFETs and ICs along the motherboard. This is why I was considering this intake duct, as it should direct airflow directly over these components.
If that's really the case, then your ducting plan is BAD. You'll just end up heating up everything else if you put your hottest component "first in line" via the duct.

Aaaaah, again

You get the most efficient cooling when the hottest part meet the coolest air.

IsaacKuo · Post by **IsaacKuo** » Wed Jun 22, 2005 12:55 pm

dago wrote:
IsaacKuo wrote:
frostedflakes wrote:As crazy as it sounds, I believe the hottest part of my system is the nForce3 chipset, along with a few MOSFETs and ICs along the motherboard. This is why I was considering this intake duct, as it should direct airflow directly over these components.
If that's really the case, then your ducting plan is BAD. You'll just end up heating up everything else if you put your hottest component "first in line" via the duct.
Aaaaah, again
You get the most efficient cooling when the hottest part meet the coolest air.

Umm...you may get the most efficient cooling for the hottest part if the hottest part meets the coolest air. But overall cooling suffers if this warmed air remains in the case and is then used to cool other components.

The ultimate contrast is between the CPU and the hard drive. The CPU can tolerate high temperatures and generates a lot of heat. The hard drive can't tolerate high temperatures and generates little heat. Which of the following is more efficient?

1. fresh air -> CPU -> hard drive -> exhaust

or

2. fresh air -> hard drive -> CPU -> exhaust

The former will indeed cool the CPU marginally more efficiently, but it will cook your hard drive. The latter feeds the hot CPU slightly warmer air, but it will keep both components sufficiently cool.

frostedflakes · Post by **frostedflakes** » Wed Jun 22, 2005 1:00 pm

Well, it's the hottest component (just an estimate, there is no sensor for NB temp, but the heatsink/PCB surrounding it is rather warm), but that doesn't neccessarily mean it gives off the most heat. If I could attach an NCU-2005 to my NB, I'm sure it would run significantly cooler than my CPU. However, passive heatsink selection for motherboard chipsets is limited, and what I'm using now is basically the best that's available. So the next logical step in my mind is to try and direct as much air over the current heatsink as possible in order to increase it's effectiveness. With the CPU at 39*C, and the air blowing out of the S12 only slightly warmer than ambient (with the fan at ~765RPM, mind you), I think adding a couple watts of heat to the CPU/power supply is the least of my worries. Just my $0.02, though.

As far as flipping the power supply fan, I could see it being effective in certain setups, but not this one. What could it possibly gain me, other than lower power supply and CPU temps, which I don't need. The areas that need attention now are the chipset and lower motherboard components. With a 120mm intake at the top of my case fighting natural convection, I couldn't see how these areas would be any better off - likely they would run hotter than they already are.

IsaacKuo · Post by **IsaacKuo** » Wed Jun 22, 2005 1:09 pm

frostedflakes wrote:As far as flipping the power supply fan, I could see it being effective in certain setups, but not this one. What could it possibly gain me, other than lower power supply and CPU temps, which I don't need. The areas that need attention now are the chipset and lower motherboard components. With a 120mm intake at the top of my case fighting natural convection, I couldn't see how these areas would be any better off - likely they would run hotter than they already are.

The power of natural convection is really really really small, especially when you don't have any major heat sources powering it among the lower motherboard components.

I'll bet good money that your chipset and lower motherboard components will be cooler if you flip the PSU fan. Why? Because the air coming out of the 120mm fan will be a directional column of air (somewhat spiraling). This flow will be pointed at your motherboard components even without any ducting involved (I notice no VGA card in the way). Also, it will set up a good amount of circulation within the case.

But of course, this is all theory. Reality trumps all theory. How about you just try flipping the fan and reporting the effect on your temps?

dago · Post by **dago** » Wed Jun 22, 2005 1:19 pm

IsaacKuo wrote:Which of the following is more efficient?

1. fresh air -> CPU -> hard drive -> exhaust
or
2. fresh air -> hard drive -> CPU -> exhaust

1, no discussion, you even agreed [ok, ok, I'm pedantic on your question

]

The former will indeed cool the CPU marginally more efficiently, but it will cook your hard drive. The latter feeds the hot CPU slightly warmer air, but it will keep both components sufficiently cool.

It depends, but the discussion is over :
1. fresh -> hdd -> chipset -> cpu -> psu -> out
2. fresh -> psu -> cpu -> chipset -> hdd -> out

Another alternative would be to have a hole in the bottom of the case to get fresh air trough the whole MB (hdd alone should be ok with few air).

frostedflakes · Post by **frostedflakes** » Wed Jun 22, 2005 1:29 pm

dago wrote:Another alternative would be to have a hole in the bottom of the case to get fresh air trough the whole MB (hdd alone should be ok with few air).

Interesting idea, I hadn't even thought of that. The front intake would likely provide more airflow, but the bottom wouldn't neccessarily need ducting to be effective. Thanks for the suggestion.

IsaacKuo: I would in a heartbeat, but I don't want to lose the warranty on my Seasonic. Unfortunately I don't have any spare power supplies with 120mm cooling either. This best I could do is position a Yate Loon next to the power supply blowing down. Although this would probably not duplicate the intended effects of the PSU fan blowing down, it could give us a good idea of this method's cooling potential. I may tinker around with it later on.

IsaacKuo · Post by **IsaacKuo** » Wed Jun 22, 2005 1:56 pm

dago wrote:It depends, but the discussion is over :
1. fresh -> hdd -> chipset -> cpu -> psu -> out
2. fresh -> psu -> cpu -> chipset -> hdd -> out

Actually, in this case the hard drive is not anywhere in the main airflow path (it's in the upper front corner in an optical drive bay). That's okay, because it's a 2.5" drive and thus cooling it isn't a priority.

Anyway, the fan flipping idea isn't going to happen, so the point is moot.

As for putting a hole in the bottom of the case--I don't see the point. The duct as proposed is already feeding the bottom motherboard components fresh air, and it does so without major modification (like increasing the height of the case feet for bottom clearance.

dago · Post by **dago** » Wed Jun 22, 2005 8:53 pm

Actually, in this case the hard drive is not anywhere in the main airflow path (it's in the upper front corner in an optical drive bay). That's okay, because it's a 2.5" drive and thus cooling it isn't a priority.
[/quote]

was just thinking of the pictures on top.

IsaacKuo wrote:
As for putting a hole in the bottom of the case--I don't see the point. The duct as proposed is already feeding the bottom motherboard components fresh air, and it does so without major modification (like increasing the height of the case feet for bottom clearance.

That could also help cooling the video card (if it was needed)

But I agree it's probably not worth the modifications.

alleycat · Post by **alleycat** » Thu Jun 23, 2005 12:29 am

As long as you have an easy path for cool air to enter the duct, and the rest of the system is well sealed, it should work well. I don't really like front intakes myself, as they seem too restrictive. My system has a hole cut in the bottom of the case, and is raised a few centimetres, similar to what dago is suggesting. Depending on what tools you have, this could possibly be the easier option.

Sorry Isaac, but I disagree with your suggestion about inverting the fan. I know you have had success with this technique, but personally I would feel very uncomfortable about doing this. It would probably be okay in a system where the PSU is at the bottom of the case.

frostedflakes · Post by **frostedflakes** » Thu Jun 23, 2005 9:19 pm

Bought 3' of 4-1/2" OD 4-1/4" ID acrylic extruded tubing online for the duct. Hopefully I'll be able to get this project done sometime next week.

P.S. If anybody happens to need some tubing for a similar duct, let me know, as I'll only need about a foot for what I plan to do.