Page 1 of 3

Does "Evacuate the heat" role for the PSU work?

Posted: Mon Jan 05, 2004 8:10 pm
by MikeC
ColdFlame (from another thread...) wrote:I actually prefer the fan at the bottom of the PSU because it sucks hot air from the CPU.
Here's where I deviate from accepted "convention" (as I do in so many ways... bwahahahahahah). This post might get a few PSU makers and sellers ticked, but what the hey... mine is only one voice in the wilderness.

The whole dual fan PSU idea seems to have been promoted by Enermax & AMD around the time the original Athlon appeared. [Some of may recall how the Athlon pushed the then-standard 250W PSUs (especially no-name generic ones) to overload, causing all kinds of mysterious instability issues that were eventually traced to inadequate PSUs.]

Anyway. Back then there was not much thought given to heat and airflow in PCs. If you think cases are airflow impeded now, you should take a look at the old ones. They were worse. So with the extra heat from the Athlon, the extra fan in the PSU sucking air from the CPU area made a lot of sense, it got rid of the heat that the systems weren't handling well otherwise.

Now you know those Athlons were hot in their day, but early ones didn't exceed ~35W. The extra heat didn't really change the PSU operation much, especially as the industry was not paying much attention to noise. (Of course, later the T-birds reached ~70W...) For the last couple years, AMD has even had a diagram in their popular system assembly guides showing this blow-thru-PSU heat exhaust as the preferred case cooling.

Skip ahead to the present.

You now have CPUs pumping >80W of heat. For quiet enthusiasts, PSUs have thermal control to keeps its fan(s) running slow except when the internals reache a high temperature. PSUs of 400W rating are commonplace, as are VGA cards that dissipate >60W. Dual HDDs are almost ordinary. Yes, PCs in North Am seem to have followed the auto industry lead -- big gas guzzlers are the norm.

So when a fan sucks the rising heat from the CPU (and everything else below) and blows it through the PSU, what do you figure happens?

Right -- the PSU & that controlling thermistor gets HOT. This then leads the PSU fan control circuit to... speed the fan up! Which leads to more... NOISE.

So depending on the particulars of the PSU fan controller, your PSU fan could be going up & down in speed and noise like a yoyo. To me, this is much worse than a steady noise.

With CPUs and systems pumping out as much heat as they do, the idea of of using the bottom mounted fan in a PSU to suck more hot air out only makes sense if you...
1) don't care about how noisy the PSU fan gets as it speeds up in response to the extra heat, and
2) don't care what the extra 50-150W of heat going through the PSU will do to its longevity & stability.
3) have a well ventilated, modest-heat system that does not cause the PSU fan to speed up.

This is why I think the bottom mounted fan PSUs need to be used very carefully if you want to have a quiet system. No question, if the overall system heat is modest, it can still work fine. I'm thinking something like max 125W DC power draw. Maybe less. It depends how quiet is quiet enough for you.

My ideal case config for higher power systems is a straight-through airflow path for the PSU (with a duct from emty top CD bay to PSU intake side) with the PSU fan at the absolute minimum. Yes, this means you need to have good heat exhaust around the CPU area, but you need this anyway, and 2 quiet slow fans are almost always quieter than one fast noisier one.

This is not new, it's the idea Lilla ran with in Building a PSU channel but first explored here by early contributor Leo / powergyoza in his Quiet Dualie article

OK, let's hear the counterpints. uh... I mean points. :wink: I wonder if any PSU engineers will say anything?

Posted: Mon Jan 05, 2004 8:30 pm
by nap_loaf
This wasn't something I considered properly when choosing a PSU. Looking at the Silencer and Tornado, I thought to myself: "Hmm...the Silencer uses an 80mm fan while the Tornado uses a 120mm. It said somewhere on SPCR that a 120mm fan can move the same amount of air but with less RPM and noise than a 80mm. The Tornado it is!"

That was my reason for choosing it - the fan orientation didn't come into it. But while on that topic, it doesn't seem like a good idea to have two 120mm fans (the other one being the case exhaust) at right angles in close proximity. Wouldn't they be competing with each other?

Posted: Mon Jan 05, 2004 8:48 pm
by GamingGod
Well I have a fairly high powered system (see sig) and I dont really notice the fsp300 ever speeding up. It might ever so slightly but when im gaming I dont notice anyways. It does push alot of hot air out, Much Much more than the 120mm case fan ever catches, but it doesnt seem to be a problem for the power supply. It would probably be better if the power supply took care of just cooling the hardrive and was isolated towards the bottom of the case. And the top portion of the case was cooled by 80 or 120mm fans to cool everything else. I do feel that my case fan does very little, It helps case temps a bit, but not as much as I would like.

Posted: Mon Jan 05, 2004 8:57 pm
by Trip
If you like ducts, then are cases with doors bad? I've never used a Plextor or comparable Optical Drives, but mine are pretty loud. A door would provide some sound insulation, esp. if lined with thin foam, but would block a duct somewhat.

You prefer 2 80mm to 1 120mm case exhaust fan, correct?

Would this design be superior to a typical PSU: Image

(ignoring this particular PSU's price, failure rate, and its failure to meet ATX length standards.)

Actually, this one doesn't even have an exhaust for the air I don't think... What I mean is would a fan blowing from inside the case across the heatsink and out the back be preferable to one on the outside of the case?

Posted: Mon Jan 05, 2004 9:13 pm
by MikeC
If you like ducts, then are cases with doors bad?

You prefer 2 80mm to 1 120mm case exhaust fan, correct?

Would this design be superior to a typical PSU:
1) No what's wrong with doors? There are ways around doors.
2) My point about the PSU has little to do with 80 vs 120mm fans per se, but about the evacuate the heat role for the PSU. If the 120mm fan is quiet enough great, let's use it, I mean I have no hard and fast rule about this, it's not some kind of strict principle... Maybe I should re-title this thread?
3) If the pic is of a passive design meant to go into an ATX form factor, then no I don't think it is superior. The thermal airflow just doesn't seem right for that passive PSU style. I think you really have to integrate a cooling pipe -- a chimney I mean -- and straight up heat exhaust path in a passive PSU (with any serious power) to achieve some acceleration of the the rising convection cooling, otherwise, overheating will always be the limiting factor. Like the Apple Mac CUBE. Of course, the type of passive PSU I am describing is not possible to fit within the standard ATX case, we're talking custom design.

Posted: Mon Jan 05, 2004 9:56 pm
by Trip
1. What do you mean "ways around doors", removing them?

2. This may be too OT, but I was thinking the "ideal" case exhaust could be different depending on the orientation of the PSU fan and the distribution/amount of heat produced by the PSU and the rest.

I think this is 120mm bottom vs. 80mm front if you stick with ATX standard, so the 120mm vs. 80mm inevitably gets pulled in. Heh, who would argue that an isolated PSU in general is not preferable?, all other variables (inc. ventilation) constant. Dual 80mm could be brought in too I suppose, but I want to say that 120mm is better.

3. The purpose of this is that a bottom orientation produces sound through an indirect path, whereas a front orientation produces sound through a direct path. A back orientation could perhaps follow an indirect path, use a duct, a quieter 80mm - basically, eliminate the indirect path advantage a bottom fan PSU has. Thanks for the info. about the need for heatpipes, that's another question I've had, but was too OT for this thread.

Oh, the pic was of the silenmaxx prosilence 420 with the fan in the back. I was referring to a PSU with the fan in the same place, but with an exhaust for the air to exit from. The air must come out from a vent that I don't see. Heh, if there's no vent, I don't think a fan sucking in air would work right.

EDIT: Add-on

The big question is whether the extra case ventilation, extra CFM/noise, and indirect noise path is worth the reduced heatsink, extra heat in PSU, reduced PSU lifespan, and alternating noise. and how is case airflow affected? > vibrations of 120mm could be another issue. Also for me, since db is an exponential, do 120mm truly produce better CFM/noise?

Posted: Tue Jan 06, 2004 1:04 am
by powergyoza
Trip wrote:The big question is whether the extra case ventilation, extra CFM/noise, and indirect noise path is worth the reduced heatsink, extra heat in PSU, reduced PSU lifespan, and alternating noise. and how is case airflow affected? > vibrations of 120mm could be another issue. Also for me, since db is an exponential, do 120mm truly produce better CFM/noise?
Unless PSU efficiencies increase far past 70-80% and other devices drastically cut their heat production, I'd vote for cooling the PSU will fresh air. Besides, the noise from a 5v panaflo (all I need for my big PSU) does not really justify indirect noise paths. Come on folks, take a hint from the WTX specs!

Posted: Tue Jan 06, 2004 4:20 am
by Bluefront
The new setup I'm finishing up uses the Fortron 120mm PSU as the only exhaust fan. I replaced the stock fan in the unit with an Evercool fan, and to avoid the up/down fan noises, I control the speed with an external controller. This setup with a duct around an Alpha heatsink can operate as the only fan in the whole computer.

The temps were slightly too high for my liking, so I added another Evercool 120mm fan on the bottom of the case. This enables the computer to run cool using only the two fans.....both running at about 1100rpms. I'll have final pics of this setup ready soon.

Posted: Tue Jan 06, 2004 6:29 am
by Ralf Hutter
Trip wrote:If you like ducts, then are cases with doors bad?
Most doors have enough air leakage around them, or like the 3700 and Antec SX10x0 series actually have openings around their periphery, so you get halfway decent airflow without a direct (i.e. "noisy") path to a PSU duct. Seems to me like a slightly ventilated door would be quite preferable on a case with a PSU duct.



And as far as MikeC's topic, I'd have to say that it kind of seems like a no-brainer to me. Why would you possibly want to pump more heat through a PSU that's thermally controlled to run quiet? The heat vs. longevity issue is probably as, or more important than the noise issue too but impacts us less here at SPCR where out typical priority is silence.

Posted: Tue Jan 06, 2004 7:13 am
by Bean
Does more fans at low rpm exceed the noise created by less fans but with one fan at higher rpm? For example lets say I use three fans at low rpms. Is that quieter than two fans at the same low rmps and one at slightly higher rpms (say 7-9 volts) which would be the case if you introduce hotter air into a psu.

If the answer is less fans then the PSU fan controller needs to be addressed as in eliminated or reprogrammed to be less intrusive when it speeds up the fan.

Posted: Tue Jan 06, 2004 8:11 am
by powergyoza
Ralf Hutter wrote:Most doors have enough air leakage around them, or like the 3700 and Antec SX10x0 series actually have openings around their periphery, so you get halfway decent airflow without a direct (i.e. "noisy") path to a PSU duct. Seems to me like a slightly ventilated door would be quite preferable on a case with a PSU duct.
How very true! Still, I saw enough of a difference between door closed and door open that I needed to open a hole in my SX1040 door.
Bean wrote:Does more fans at low rpm exceed the noise created by less fans but with one fan at higher rpm?
No, more slower fans always make less noise fewer faster fans (given the same aggregate cfm)

Posted: Tue Jan 06, 2004 9:48 am
by Sam Williams
Well, the PSU fan - an 80mm Panaflo L1A - is the sole extractor in my system (schematic here). Fresh air is pulled in through a duct by the CPU heatsink fan; it then rises from the heatsink, into the PSU and away. There's a slow fan stirring the air near the HDD, to keep it close to case ambient, but nothing else. The PSU never speeds up audibly. The situation is helped by two angled exterior ducts - the exhaust stream is turned upwards to the vertical, while the intake is oriented downwards to avoid pulling in exhausted air. Case ambient is 32C; typical CPU temperature is 42C.

The main advantage of this arrangement is that it places all the serious airflow at the back of the case, away from my ears - and it's more than cool enough. Admittedly, this is a low-powered system (1 GHz Athlon, only one HDD) but there certainly are scenarios in which the PSU can quietly get all the necessary extraction done.

Posted: Tue Jan 06, 2004 11:22 am
by Joe DeFuria
Just to add my $0.02...

I also have a 120mm PSU (SS Tornado), and it works effectively as the only air "exhaust" in my kid's computer.

It's a P3 Celeron / Tualatin at 1.4 Ghz, running under constant F@H load. I have a front case fan (120mm) pushing in air at about 1000 RPM, no fan currently running on the CPU.

The PSU fan (sitting "above" the CPU) runs at about 900 RPM constant. CPU temps are in the low to mid 50s. I can turn the CPU fan on a low setting (80 MM low profile fan) to push the CPU temps down to well under 40.

The PSU fan may be drawing in some fresh air from the large, gaping 120 MM hole (grill with no fan) in the case rear too. ;) I'm thinking of blocking that grill off to make sure the air comes the front, to minimize dust.

So, I would basically concur with what others have been saying. The 120 MM power supply fans can be particularly useful when in low to moderate power situations...when it can be sufficient to be perhaps your only case fan, or certainly your only exhaust fan.

Otherwise, if you have a separate exhaust fan, the best bet is to duct outside air to your "standard" power supply. Let the power supply fan "only" be concerned with cooling itself, and have your case fan(s) cool your components.

Posted: Tue Jan 06, 2004 2:44 pm
by snutten
Ralf Hutter wrote: ... And as far as MikeC's topic, I'd have to say that it kind of seems like a no-brainer to me. Why would you possibly want to pump more heat through a PSU that's thermally controlled to run quiet? The heat vs. longevity issue is probably as, or more important than the noise issue too but impacts us less here at SPCR where out typical priority is silence.
That should just about wrap this up shouldn´t it?
If the rest of the system is quiet, why make the PSU the big problem by having it coping with extra heat? It already makes for one of the hardest components to shut up.

Posted: Tue Jan 06, 2004 3:06 pm
by MikeC
snutten wrote:That should just about wrap this up shouldn´t it?
If the rest of the system is quiet, why make the PSU the big problem by having it coping with extra heat? It already makes for one of the hardest components to shut up.
Not exactly... One reason I started this thread is because of the number of posts by people planning to use (or having problems with using) 120mm fan PSUs in hot systems. My point is obvious to lots of people who have been thinking about this whole cooling + silencing thing for a while, but apparently not so obvious to lots of other people.

Here's a challenge to SPCR readers:
What's the best practical way to use a 120mm PSU in a hot system to maximize cooling without allowing the fan to speed up beyond minimal?
If there are enough responses, I'll move them to a new thread. And maybe we can get a sponsor to put up a 120mm fan PSU as a reward for the best answer! In fact, if we can't get a sponsor to ante up, SPCR will give one away! :!: :)

Posted: Tue Jan 06, 2004 3:28 pm
by Bean
Is the PSU 120mm fan the only exhaust fan allowed?

If not then proceed as follows: I'd determine what temperature will make the PSU fan RPM increase. Once I knew that, keep the inlet temperature to the PSU below that point. You could do that with a rear exhaust fan on a fan controller. As the PSU fan speed increases turn up the rpms on the exhaust fan.

Posted: Tue Jan 06, 2004 3:37 pm
by MikeC
Bean wrote:Is the PSU 120mm fan the only exhaust fan allowed?

If not then proceed as follows: I'd determine what temperature will make the PSU fan RPM increase. Once I knew that, keep the inlet temperature to the PSU below that point. You could do that with a rear exhaust fan on a fan controller. As the PSU fan speed increases turn up the rpms on the exhaust fan.
OK, let me clarify further...
What's the best practical way to use a PSU with a CPU-facing (bottom) intake fan in a hot system to maximize cooling while keeping the noise level as low as with the best setup using a quiet single 80mm rear fan PSU? (No limit to number of fans in PSU)
Is that clearer?

Also, regarding your solution, how would you implement that practically in a system?

Posted: Tue Jan 06, 2004 3:54 pm
by Zyzzyx
MikeC wrote:What's the best practical way to use a 120mm PSU in a hot system to maximize cooling without allowing the fan to speed up beyond minimal?
Durn that 'practical' word. ;)

My first random thought was cut a hole above the PSU, flip it upside down, and have the fan be right in fresh ambient air. Not practical though, and not too silent either.

Posted: Tue Jan 06, 2004 4:19 pm
by lenny
MikeC wrote:What's the best practical way to use a PSU with a CPU-facing (bottom) intake fan in a hot system to maximize cooling while keeping the noise level as low as with the best setup using a quiet single 80mm rear fan PSU? (No limit to number of fans in PSU)
Do you mean "no limit to number of fans in case" rather than PSU?

Assuming it is true that a single 120mm fan PSU (fan located inside the case) moves the same amount of air at a lower noise level than a single 80mm fan PSU (fan located at the back of the case).

The best practical way, in my opinion, would be ducting the CPU and video card to exhaust hot air immediately rather than allowing it to recirculate.

Edit to add clarification:

If assumption is incorrect (120mm quieter than 80mm at same airflow) then there really is not much point in using such a PSU.

Posted: Tue Jan 06, 2004 4:19 pm
by Bean
MikeC wrote: OK, let me clarify further...
What's the best practical way to use a PSU with a CPU-facing (bottom) intake fan in a hot system to maximize cooling while keeping the noise level as low as with the best setup using a quiet single 80mm rear fan PSU? (No limit to number of fans in PSU)
Is that clearer??
If you can use more fans in the power supply unit modify the PSU with a rear fan and open the front of the PSU. Provide the front of the PSU with a ducted intake to the front of case – cdrom opening as discussed here before.
MikeC wrote:Also, regarding your solution, how would you implement that practically in a system?
Well you send me that PSU and I’ll get right on it :lol: j/k
My thinking is a rear exhaust fan will share the temp load with the PSU and if you increase the exhaust fan rpm it will suck more hot air before it reaches the PSU.

How to do it? I’m just guessing here as I have not done this (I just disqualified myself :roll: but since you asked: You could do it trial and error. Using worse case temperature (run prime and turn the heat up in house) start dialing up the voltage to the exhaust fan controller until the PSU resumes its steady state low rpm.

More elegant would be thermally controlled fan that ramps up the speed with temperature increase. If the manufacturer of the PSU provided the temp that the fan increased, make the exhaust fan increase the rpms at that point.

Posted: Tue Jan 06, 2004 4:24 pm
by lenny
Another random thought : wire the PSU so that the fan control circuitry controls the case fan, and have the PSU fan hard-wired at a fixed voltage.

Obviously some fiddling required to ensure that the case fan is doing its job. A failsafe to kick the PSU fan into higher voltage when the PSU fan control ramps up too high, for example.

Posted: Tue Jan 06, 2004 4:26 pm
by MikeC
Further clarification of the task:
What's the best practical way to use a PSU with a CPU-facing (bottom) intake fan (such as a 2-fan Enermax, Antec, etc or 120mm fan Seasonic, Fortron, Nexus, etc) in a hot system to maximize cooling while keeping the noise level as low as with the best setup using a quiet single 80mm rear fan PSU?

Posted: Tue Jan 06, 2004 4:38 pm
by Zyzzyx
MikeC wrote:
Bean wrote:You could do that with a rear exhaust fan on a fan controller. As the PSU fan speed increases turn up the rpms on the exhaust fan.
Also, regarding your solution, how would you implement that practically in a system?
How to have the case exhaust fan ramp up in speed along with the PSU fan?

Sounds fairly easy to me, just a bit of custom wiring involved. Wire the two fans in parallel so they see the same voltage, and if the same fans will see the same speed. Use the + lead from the PSU, split it, and run one lead each to the PSU and case fan.

And actually, if I hadn't fried the fan circuit on my Seasonic SS300, I'd probably give this a try too. If I had a case (and $$, of course) with a 120 rear fan, I'd pick up an ST300 to give it a try.


edit add: (wrote above as MikeC and Lenny posted) I'd still think about giving this setup a try with a 120mm bottom PSU fan and an 80mm rear case fan. Seems like it would work better with a matched 120mm rear case fan though.

Similar to lenny's idea, yet I'd have variable speeds for each fan, both controlled by the PSU circuitry.

Posted: Tue Jan 06, 2004 5:05 pm
by Bluefront
As I said in my previous post, right now I am running a P4 2.66, two maxtor hard drives, two optical drives, an AIW8500, etc, using a fortron PSU modified with an Evercool 120mm fan as the only exhaust.

The PSU fan is controlled manually from a front fan control unit. How quiet it runs depends entirely on how hot you are willing to run. At 1300rpms, right now the CPU is idling at 35c. If I run the fan down to it's lowest speed (about 800 rpms) the idle temp goes up to about 40c....but the computer is almost silent to me.

To maximize the cooling potential of the PSU fan, I constructed a duct around the PSU fan which forces most of the air to travel through the cpu heatsink. I suspect a different CPU heatsink designed for this airflow would perform much better than my Alpha (no fan on it). Still working on that......but this can be done.

I've always been amazed that so few people attempt to maximize the cooling potential of their 120mm PSUs. The fan is right there, an inch or so above the CPU heatsink. Why not use it more effectively?

Posted: Tue Jan 06, 2004 5:19 pm
by MikeC
Given that your P4-2.66 pretty much qualifies as a hot CPU and your dual HDDs & AIW8500 are all somewhat hot, your solution is pretty much the most practical so far. The Heatlane Zen 1000 would be the natural choice for HS. But you are more or less ignoring the actual heat in the PSU by having the fan manually controlled, which probably means it get toasty in there sometimes. How does the CPU do under a real stress test like Prime95 or CPUBurn?

I guess I will have to try your technique to determine how quiet/cool it can be. Evercool 120mm -- can you suggest a good source? Is that the best 120mm you've tried?

Posted: Tue Jan 06, 2004 5:38 pm
by Bluefront
Well I'm using an A-Open board with SilentTek. And I have a thing about heat....I avoid it. So I have two small fans on the duct, controlled by SilentTek. I have them set to turn on about 50c, so using CPUBurn, the cpu max temp is about 53c. These little fans rarely turn on under normal usage, and I usually run the PSU fan at 1200 rpm.....still plenty quiet for me.

There's a CompUSA on my way to work, so I bought the Evercool fans there. CompUSA sells the Aluminum EverCool fan for $17. They also sell their own brand of 120mm fan (plastic body) which uses the same fan motor, same specs, for $13.

I've been wanting to try the Zen, but it's slightly too tall to fit in that case setup. Maybe with a few more mods.....

Oh, I've been closely monitoring the temps of this setup. I have a sensor right at the output grill of the Fortron. It rarely goes over 36c. I also am using a small muffler on the exhaust without any heat problems so far. (and I don't expect any)

Posted: Wed Jan 07, 2004 5:18 am
by snutten
Sorry to bug you about this but can anybody tell me why you´d want to use the PSU as case fan?
Bluefront, you seem to know just about everything about this. Help me out please?

Either way I look at this, I think the PSU fan shall have to spin faster if the air it´s sucking in is hotter. Ducting cool air from the front (or something like that) makes possible a quieter PSU.
Adding quiet undervolted fans on rear or top of case has only minor comparative negative effects: the extra fans costs money, the extra fans need to be installed properly with isolated mounts and all, and it opens up the case so that noise comes out. But this can be adjusted with a soundtrapping cover to avoid the straight path out and still let air out.

Are these the reasons for using the PSU fan?

On topic it seems to me the best to simply duct the CPU heatsink to the PSU fan just like SilentMike and Bluefront suggested.

Posted: Wed Jan 07, 2004 6:35 am
by Ralf Hutter
MikeC wrote:Further clarification of the task:
What's the best practical way to use a PSU with a CPU-facing (bottom) intake fan (such as a 2-fan Enermax, Antec, etc or 120mm fan Seasonic, Fortron, Nexus, etc) in a hot system to maximize cooling while keeping the noise level as low as with the best setup using a quiet single 80mm rear fan PSU?
The answer seems obvious to me but maybe I"m missing the point: "Use a standard PSU with one 80mm fan on it's back wall".

Why use a 120mm PSU at all? Why try and cool the case with the PSU? Why not use the case fan on the rear wall to cool the case and let the PSU cool itself? It's not like anyone is forcing us to use only bottom facing 120mm PSUs. We do have a choice.

Posted: Wed Jan 07, 2004 6:59 am
by Joe DeFuria
MikeC wrote:Further clarification of the task:
What's the best practical way to use a PSU with a CPU-facing (bottom) intake fan (such as a 2-fan Enermax, Antec, etc or 120mm fan Seasonic, Fortron, Nexus, etc) in a hot system to maximize cooling while keeping the noise level as low as with the best setup using a quiet single 80mm rear fan PSU?
I would say this:

In a hot system, have two case fans, with both the the rear fan and front set as an intake fan. (Pull cool air from the rear of the case...)

Posted: Wed Jan 07, 2004 8:05 am
by powergyoza
So all the heated air goes thru the PSU, thereby heating it further?

I think the central issue of this debate is whether the heated air from the CPU, VGA, HDD, VRM, NB, etc. is still useful to us. Past a certain temperature, ie. beyond a certain power draw, I'd say NO. Where do I define that threshold? There are sooooo many variables, but to start:
  • Fan sizes, voltages, locations & qtys.
  • Case size, airflow efficiency
  • Power draw, locations of hotspots
  • PSU efficiency
  • CPU, VGA heatsink choices
  • Noise threshold
Any more?