Why a separate air feed to the PSU is good.

Enclosures and acoustic damping to help quiet them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Devonavar

Post Reply
Olle P
Posts: 711
Joined: Tue Nov 04, 2008 6:03 am
Location: Sweden

Why a separate air feed to the PSU is good.

Post by Olle P » Wed Feb 04, 2009 6:57 am

Background
In the old school PC case design more or less all ventilation was done by the PSU fan alone.
As CPUs became more power hungry, this wasn’t enough, so one or two extra fans near the PSU became a standard feature.
In a fairly generic mid tower case the inside air is heated mostly by the GPU, north bridge and CPU. This air is then evacuated by two fans; the PSU fan and a case fan in the rear wall, just below the PSU. The PSU is thus fed pretty warm air, that need good flow to add cooling to the PSU as well.

Recently there have appeared cases that let the PSU draw fresh air more or less directly from the outside instead, claiming this to result in a quieter computer since the PSU fan then can run slower.
I’ve been very sceptic about this, realising that the explanation cant be that simple. Guess what? The explanation isn’t that simple, but the total result still stands!

Here’s my reasoning towards the more complex explanation and result.
1. Feeding the PSU cooler air should make it quieter. No complaints there.

2. By removing the PSU from the standard air flow design, the overall flow will be reduced, indirectly heating up the other components. Therefore additional air flow is needed, either by adding another fan or by increasing the speed of the original fan, both options adding more noise. Question: Will that be more or less noise added than the one reduced by cooling off the PSU?

3. Insight: With the original design there was a need for high air flow just to keep the air fed to the PSU cool enough. When the PSU is gone, a higher temperature on the evacuated air is tolerable, so less air flow is required!

4a. Conclusion 1: Low air flow through the PSU plus relatively low air flow through the rest of the case equals a fairly quiet computer!
4b. Conclusion 2: If maximum cooling is desired, then the PSU needs to be replaced by another fairly high speed fan. That fan will be quieter than the PSU at the same air flow, but still probably quite noisy, so the overall noise effect of keeping the PSU separated is reduced considerably.

Cheers
Olle

ntavlas
Posts: 811
Joined: Mon Jul 16, 2007 2:35 pm
Location: Greece
Contact:

Post by ntavlas » Wed Feb 04, 2009 7:14 am

I agree. Furthermore, the case and psu fans are usually placed right next to each other at a 90 degree angle, not a very efficient setup as the two fans are competing for airflow. Even with a duct, the low pressure remains plus the duct creates extra resistance to the psu fan.

vertigo
Posts: 647
Joined: Sun Apr 17, 2005 6:09 am
Location: UK

Post by vertigo » Wed Feb 04, 2009 8:44 am

Question: Will that be more or less noise added than the one reduced by cooling off the PSU?
Yeah, it'll be less because the PSU fan has to spin fast enough just in case someone relies on it alone to cool their system. The PSU maker doesn't know how open the case is.

Another point, high exhaust temperatures mean you're removing heat. A quiet system will very often have high exhaust temperatures.

The key is having high exhaust temperatures but a low ambient case temperature. When the ambient temperature in your case is low (and no hot spots) and the exhaust temp is high, your airflow is highly efficient. It is always good to have some extra airflow of course but that's the principle.

But there is a caveat, fans will tend to wear quicker in hotter air so you probably want to keep your fans below 50 degrees. So a common action to reduce noise is to move the fans to cooler air so they can be slowed down, raising exhaust temperatures safely.

Of course, once your system is quiet you don't need to reduce the airflow any more.

By splitting the airflow into two "lanes", the heat is separated and you can be more confident about the exhaust temperature. You know that all the heat of the system can't accidentally go out one exit. This helps in hot climates.

dukla2000
*Lifetime Patron*
Posts: 1465
Joined: Sun Mar 09, 2003 12:27 pm
Location: Reading.England.EU

Post by dukla2000 » Wed Feb 04, 2009 11:54 am

Well, I disagree. Or at least, it depends. I think there is often a case to be argued that more than 1 fan is over-engineering, especially in these days of 12cm fans.

To cause an argument let me describe my system. It is running 70W at the wall, lets assume the PSU is 70% efficient. That means 21W being dissipated in the PSU, 49W by the rest of the PC. Which means the PSU is not necessarily the primary cooling problem.

But it does have a nice 12cm fan running a bit over 8V, more or less 700rpm, in theory 26cfm in free air. Obviously there is static pressure but on the one hand I cant measure it, and on the other even if I could the tiny curves published by Yate Loon for the Nexus D12SL-12 are not suitable to plot/read the resulting cfm. So to continue the argument lets assume my Nexus is actually shifting 12cfm.

Then delta T (temp rise across the system in Centigrade, myopically ignoring spot cooling requirements etc!) is just over 10C. With an ambient usually under 20C that means my exhaust is under 30C (and intake to the PSU is less than 27C). None of which gives cause for any concern whatsoever, and certainly not to the extent that I want my PSU to have a separate air feed.

ntavlas
Posts: 811
Joined: Mon Jul 16, 2007 2:35 pm
Location: Greece
Contact:

Post by ntavlas » Wed Feb 04, 2009 12:39 pm

In a system like you describe, more so in a 35 watt htpc, using the psu as an exhaust is perfectly sane. In a 120 watt system and with an ambient of over 30 degrees though you`d be much better served with a dedicated exhaust and more than one fans.

vertigo
Posts: 647
Joined: Sun Apr 17, 2005 6:09 am
Location: UK

Post by vertigo » Wed Feb 04, 2009 3:17 pm

Dukla2000 wrote:49W by the rest of the PC.
What CPU is that? It's pretty low powered, not at all typical.

The nice thing about having good cooling is that you know nothing can go wrong. It may become 36C that day and you know that none of the heated air is being trapped in the case, raising the ambient to worrying levels. When I hear about people running their cards at 90C, etc, I shudder. Especially NForce chipsets running > 70C is madness (it may be within spec but you can do so much better). In those cases they have definitely underengineered the cooling.

forester joe
Posts: 46
Joined: Thu Jun 08, 2006 3:32 am
Location: IL

Post by forester joe » Wed Feb 04, 2009 10:39 pm

dukla2000 wrote:Well, I disagree. Or at least, it depends. I think there is often a case to be argued that more than 1 fan is over-engineering, especially in these days of 12cm fans.

To cause an argument let me describe my system. It is running 70W at the wall, lets assume the PSU is 70% efficient. That means 21W being dissipated in the PSU, 49W by the rest of the PC. Which means the PSU is not necessarily the primary cooling problem.

But it does have a nice 12cm fan running a bit over 8V, more or less 700rpm, in theory 26cfm in free air. Obviously there is static pressure but on the one hand I cant measure it, and on the other even if I could the tiny curves published by Yate Loon for the Nexus D12SL-12 are not suitable to plot/read the resulting cfm. So to continue the argument lets assume my Nexus is actually shifting 12cfm.

Then delta T (temp rise across the system in Centigrade, myopically ignoring spot cooling requirements etc!) is just over 10C. With an ambient usually under 20C that means my exhaust is under 30C (and intake to the PSU is less than 27C). None of which gives cause for any concern whatsoever, and certainly not to the extent that I want my PSU to have a separate air feed.
in the original ATX design, as is the case in many builds to date, the PSU fan will pull hot air directly off the CPU heatsink. this will lead to higher intake temps on the PSU, which in turn does two things. a) it will speed up the PSU fan -> more noise. b) it will heat up the PSU -> lower efficiency and shorter life span.

arguably both factors may be negligible, but I would rather go the other way.

dukla2000
*Lifetime Patron*
Posts: 1465
Joined: Sun Mar 09, 2003 12:27 pm
Location: Reading.England.EU

Post by dukla2000 » Thu Feb 05, 2009 1:08 am

vertigo wrote:
Dukla2000 wrote:49W by the rest of the PC.
What CPU is that? It's pretty low powered, not at all typical.
4850e with CnQ running it @ 1000MHz. At full speed the whole system (780G based) is 100W: assume the PSU efficiency is up to 75% that is about 75W.

[At 1000MHz it folds about 30 pts/hour which is what it does most of the time.]

vertigo
Posts: 647
Joined: Sun Apr 17, 2005 6:09 am
Location: UK

Post by vertigo » Thu Feb 05, 2009 6:37 am

dukla2000 wrote:
vertigo wrote:
Dukla2000 wrote:49W by the rest of the PC.
What CPU is that? It's pretty low powered, not at all typical.
4850e with CnQ running it @ 1000MHz. At full speed the whole system (780G based) is 100W: assume the PSU efficiency is up to 75% that is about 75W.

[At 1000MHz it folds about 30 pts/hour which is what it does most of the time.]
If you have a low powered PC then you wouldn't need to go the extra mile for better cooling. I suppose it comes down to whether you prefer to upgrade the cooling when you upgrade the PC or to use cooling that can live over upgrades. I think the benefit is that better cooling gives peace of mind. One does not have to worry about upgrades, LAN parties, hot days, etc, it just works. Swap mobo, CPU, memory? So what?

That's also why I prefer negative pressure. You can more easily determine exhaust temperatures, removing the side panel will most likely not cause the system to overheat, and it lasts better over upgrades.

Seawane
Posts: 8
Joined: Thu Feb 05, 2009 6:44 am
Location: Bulgaria

Post by Seawane » Thu Feb 05, 2009 7:06 am

I used to have a FSP PSU with a swapped fan, spinning at 1000rpm. It wasn't as quiet as I wanted it, that's why I bought a Modu82+ PSU. My case is CoolerMaster Mystique, standard ATX case with the PSU on top. The Modu82+ is supposed to run at around 450rpm. Well, mine is probably deffective, cause it starts at 535rpm and there is a strange ticking noise from it, but it's still almost silent as long as it stays at this rpm. But I'm using a 95W Phenom CPU with Xigmatek S1283 blowing towards the PSU. As a result it starts to increase it's rpm to around 900 and I could do nothing to slow them down. That's why, I drilled holes at the top of the case, above the PSU and turned it upside down. Now it's getting fresh air and never goes above 540rpm. CPU is a little bit hotter now, but still reasonable. This case, the Mystique, though looking really good, is not suited for silence. It's still pretty quiet, but you really need the PSU getting fresh air. Especially if you're using a more hot CPU and video card.

vertigo
Posts: 647
Joined: Sun Apr 17, 2005 6:09 am
Location: UK

Post by vertigo » Sat Feb 07, 2009 3:41 pm

Hi Seawane, it sounds like you did a good job to make your PC quieter. Is it now as quiet as you need?

Seawane
Posts: 8
Joined: Thu Feb 05, 2009 6:44 am
Location: Bulgaria

Post by Seawane » Tue Feb 10, 2009 10:16 am

Probably it's even quieter than I was thinking it was possible, but now I want it silent. It's not possible though with this case. The thing is, that I started making it quieter steb by step. I didn't started it as a quiet project, cause for the last 5 years I've been using only laptops and noise has never been such an issue and I forgot how noisy desktops can be. So I started reading here and slowly made it the way it is now. During the day it's not noticeable and in the night it's possible to sleep and listen quietly to music. But I want it silent, and for that I need a case with the PSU on the bottom and fan on the top(cause since I turned the PSU upside down, the CPU idles at 4-5C higher...).

Post Reply