Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[mwillems]

So I've been trying to make my HTPC quieter because it's a full-sized PC and it's in the middle of my main listening area.  I initially thought about going entirely passive cooling, but for a variety of reasons that wasn't really a very good option for this system.  So I tried to find some suitably quiet fans, and demoed three or four well reviewed "quiet fans," and thought I should share. 

The best of show by a long way was the Noiseblocker M12-S1: http://www.amazon.com/Noiseblocker-NB-Multiframe-M12-S1-120mmx25mm-Silent/dp/B001BFTMQ6

The packaging claims a 7dB rating.  External testers have measured it at around 12 dB http://www.silentpcreview.com/article1266-page4.html, but their anechoic chamber's noisefloor is 11dB, so the fan may actually be slightly quieter than that. 

Needless to say, my own noisefloor is much higher so I can't meaningfully measure the fan noise myself.  It's the quietest case fan I've ever heard: no audible motor noise or bearing whine, just soft air turbulence, barely audible from about a meter away.

It's an 800 RPM 120mm fan so it doesn't move lots of air, but it's so much quieter than other fans you can run several of them and it will still be much quieter.  For example, the Scythe Gentle Typhoon 120mm 1450 RPM version is another reasonably quiet fan that moves much more air, but it measures around 21dB at full speed.  That single fan is eight or nine times as loud as the Noiseblocker.

I'm running four of the Noiseblockers in my case at full speed in place of two Gentle Typhoons that used to be in the case. My system is drastically quieter than it was when I was running two higher velocity fans, and I get almost the same cooling performance (within a degree or two). 

They're not cheap for a case fan, but if you want a very, very quiet fan this is it. 

[dean70]

I have been looking at options myself. Thinking of moving the rear fans (2 x 80mm) to inside the case and installing mesh grills over the holes to pull air through the case and out the Fanless PSU vents. Also cutting a hole in the lid with a mesh grill mounted over the CPU fan which is horizonal to exhaust air out the top.

Currenly have lots of foam sponge packed around and above the case to dampen the air noise. The fans themselves are quiet, but the air noise is noticable.

[mwillems]

I have been looking at options myself. Thinking of moving the rear fans (2 x 80mm) to inside the case and installing mesh grills over the holes to pull air through the case and out the Fanless PSU vents. Also cutting a hole in the lid with a mesh grill mounted over the CPU fan which is horizonal to exhaust air out the top.

Currenly have lots of foam sponge packed around and above the case to dampen the air noise. The fans themselves are quiet, but the air noise is noticable.

All my fans are inside the case currently with mesh like you're describing, and they're definitely quieter than they were when I tested them mounted on the outside of the case, but honestly I can barely even hear the air they're moving at this point. 

I also have a fanless PSU, but it doesn't seem to need much cooling for whatever reason.  It's the same wattage as my old power supply (500W), and my average system temps actually went down 2 degrees when I replaced my old PSU with the fanless one    If I'd known that would happen I'd have gotten one years ago 

[glynor]

I really like Noctua's fans, for general usage.  They have a variety of sizes, they are well built, and quiet.

Also, decibel measurements aren't everything... It is also the quality of the noise (pitch, tone, etc).

[mwillems]

I really like Noctua's fans, for general usage.  They have a variety of sizes, they are well built, and quiet.

Also, decibel measurements aren't everything... It is also the quality of the noise (pitch, tone, etc).

I looked at the Noctuas, but I didn't wind up grabbing one.  They were reviewed really well and looked pretty quiet. Maybe I'll get one and see if I like the noise level/overall sound better.

I definitely agree that tonality is pretty important, and that's one of the things I was looking for in a fan: only air noise, soft whooshing, etc.  In addition to being the quietest, the Noiseblockers were the nicest sounding of the four brands I tried (I tried a Scythe model, an Antec TrueQuiet, and a CoolerMaster, in addition to the Noiseblocker). 

One of the nice things about silent PC reviews (where I found out about the Noiseblocker) is that they include narrative about fan tonality (sometimes with recordings) as well as measured SPL, which is helpful. http://www.silentpcreview.com/article63-page2.html

That said, for my money, the best sounding noise is noise that's too quiet to hear from my seat 

[InflatableMouse]

NEXUS D12SL

I can personally testify they are awesome. Got all my systems equipped with them.

Believe the test reports like silentpcreview, but not the manufacturer's specifications.

I got the PWM version, 600-1200rpm (or 1400?). Obviously at full speed they make quite a noise but its not so much the fan itself, its the grille and components the air is pulled and pushed through.

Up to 1000rpm, they are audible, but very silent. At 600-800, virtually inaudible. But, as I said above, what matters most is the grille in front, dust filter and how close the components are to the fan. Holding a fan in open air with nothing close to it, its silent even at higher speeds. At max speed you'll hear it too of course. Try these things out for yourself and experiment in your case.

I also have 2 or 3 lower RPM versions, they start at 200, up to 600 or 800. I thought it was a good idea. However, not every fan controller or motherboard PWM mount supplies enough current to start them up at 200. They regularly didn't startup, I could see them move a few mm's as it if was trying and sometimes they'd spin up, sometimes they didn't so I replaced them for the 600 variants.

These Nexus fans come with silicon mounting thingybobs, they are great. Never mount a fan with metal or hard plastic screws.

Also, airflow from front to back is important, make sure you mount them correctly. Front fans should suck air in and back/top should blow out. Some PSU's have a switch to reverse them, make sure yours functions as an exhaust if its at the back/bottom.

Last but not least, its typically best to have sucking fans rotate slightly faster than exhaust fans. For a large case this is less important (but can still help) but its particularly helpful with a push/pull configuration on a large CPU heatsink.

[6233638]

Noctua fans are generally considered to be the best if noise is a concern, though I will say that I don't have any in my own system. Not that I don't want any, but I'm not sure I can justify the cost right now. I plan on building a new system next year, so I'll definitely be using Noctuas everywhere then.
 
In my case there are three 180mm "Air Penetrator" fans and I replaced the others with three Corsair SP120 fans after testing showed them to be very high performing fans for the money. (high static pressure, low noise)
 
I run them all at fixed speeds around 1000 RPM which seemed to help make the noise less noticeable than when they were all running at different speeds, but I wish my system was quieter.
The problem is not so much the air moving around, but the oscillations of the fans, and this is something that Noctua seem to have paid attention to.
 
 
If noise is a concern, I would be hesitant to buy PWM fans.
There are two ways to reduce the speed of a fan:
 

• Lower the voltage and send a constant low voltage signal to the fan.
• Keep voltage constant, and pulse the signal.

 
Lowering the voltage is inefficient, and many fans will not reduce their speeds very much before stalling when using this method.
 
It is much more efficient to keep voltage constant and pulse the signal rather than reducing the voltage, and this lets you reduce the speed much further on cheap fans than reducing the voltage allows for.
 
Rather than going from say 12V to 6V for 50% speed, you pulse a 12V signal so that it's only on half of the time to reduce the speed to 50%. (the numbers may not be strictly correct, but you get the idea)
You might find that a cheap fan stalls below 9V though, so that fan can only be reduced to 75% speed when reducing the voltage compared to 50% using PWM.
 
So why aren't all fans PWM? Well many people - myself included - notice a pulsing noise from the fans when PWM is used to reduce the speed this way. I have yet to hear a PWM fan that I can tolerate below 100% speed. I would rather have a faster voltage-controlled fan than a slower "quieter" PWM one.
 
Noctua claim to have solved this with their PWM fans but I have not heard them to confirm it.
PWM without the distracting pulsing should definitely be better than a voltage controlled fan.

Also, airflow from front to back is important, make sure you mount them correctly. Front fans should suck air in and back/top should blow out. Some PSU's have a switch to reverse them, make sure yours functions as an exhaust if its at the back/bottom.

If you have a top fan, you probably want it blowing in to keep positive pressure in your case to avoid dust build-up.
 
I actually think part of the problem with PC noise is that few cases are actually designed with airflow in mind at all. They just build a decent looking box that can fit in the parts, and put a bunch of holes for fans in it, without actually considering how components are going to be cooled, and how it will affect the system noise.

[mwillems]

NEXUS D12SL

I can personally testify they are awesome. Got all my systems equipped with them.

Believe the test reports like silentpcreview, but not the manufacturer's specifications.

I looked at those too, but didn't grab one.

These Nexus fans come with silicon mounting thingybobs, they are great. Never mount a fan with metal or hard plastic screws.

The Noiseblockers entire mounting frame is made out of rubber (including the screw holes) and they ship with an additional gasket just in case.  There's no way to actually make the plastic part of them make contact with the case or the screws without breaking the fan  

Also, airflow from front to back is important, make sure you mount them correctly. Front fans should suck air in and back/top should blow out. Some PSU's have a switch to reverse them, make sure yours functions as an exhaust if its at the back/bottom.

Last but not least, its typically best to have sucking fans rotate slightly faster than exhaust fans. For a large case this is less important (but can still help) but its particularly helpful with a push/pull configuration on a large CPU heatsink.

Oh for sure. I tried several different configurations to see which would give me the best cooling.  I also found this article very interesting: http://www.bit-tech.net/hardware/2012/02/10/the-big-cooling-investigation/1   They measured various fan placements in a large case when using different numbers of fans, etc.  Their ranked results proved to be about right in my setup.

Noctua fans are generally considered to be the best if noise is a concern, though I will say that I don't have any in my own system. Not that I don't want any, but I'm not sure I can justify the cost right now. I plan on building a new system next year, so I'll definitely be using Noctuas everywhere then.  
[...]
The problem is not so much the air moving around, but the oscillations of the fans, and this is something that Noctua seem to have paid attention to.

You should check out the measurements over at silent PC review: http://www.silentpcreview.com/article63-page2.html  They measured some Noctuas and a lot of other models with specific reference to the noise level and the character of the noise (oscillation, etc.)

[glynor]

One of the nice things about silent PC reviews (where I found out about the Noiseblocker) is that they include narrative about fan tonality (sometimes with recordings) as well as measured SPL, which is helpful. http://www.silentpcreview.com/article63-page2.html

That said, for my money, the best sounding noise is noise that's too quiet to hear from my seat  

Totally agree (and I, too, like Silent PC).  I meant to add, thanks for the recommendation.  I wasn't criticizing at all, just pointing it out (perhaps for others).

I also like my Noctua fans, so I figured I'd mention them.  I've never done any serious comparative testing, though.

[mwillems]

Totally agree (and I, too, like Silent PC).  I meant to add, thanks for the recommendation.  I wasn't criticizing at all, just pointing it out (perhaps for others).

I also like my Noctua fans, so I figured I'd mention them.  I've never done any serious comparative testing, though.

Yeah and they're super well-reviewed, and much beloved.  I'm interested in checking them out; I'll probably grab one to play with at some point. I'm kind of enjoying hearing all the different fans, I had no idea going into this experiment that there was so much variation!

[6233638]

I just wanted to post an update on this topic. Iʼve been putting it off for a while, but I decided that enough was enough, and I had to do something about my PCʼs noise.
I use a Silverstone FT02 which is currently in the room adjacent to the TV/music room with wires routed through the walls.
Even with it being in the next room, I would still hear the PC during quiet sections, and it always seemed rather loud at night despite the fact that itʼs using large fans, and I had them all running at their minimum effective speeds. (all below 1000 rpm)
 
At some point this year I plan on upgrading my PC and changing the case—frankly, I really dislike the FT02.
I was hoping to wait until the new build, and Noctuaʼs CPU cooler/fans with active noise cancellation to be released, but I hear those have been pushed back to Q2/3 now.

The vertical airflow used by the FT02 sounds like a good idea, but I canʼt say that Iʼve really noticed it to make a big difference compared to a regular tower, and despite the case being huge and weighing a ton, itʼs actually rather cramped inside due to the internal layout.
I had always planned on using dual GPUs at some point, only to find that they donʼt fit inside the FT02 if you have any optical drives installed. You might be able to get away with it if you have short CD/DVD readers, but not a full size Blu-ray drive.
The hard drive mounting system is a disaster too:
 

• Only one of the “hot swap” bays actually comes with a backplane.
• It takes far longer to mount the drives inside the rails with their anti-vibration grommets than it would to just normally fit a drive in a regular cage.
• The rails are all made of thick plastic which seems to provide rather good insulation.
• The drives are right up against each other with very little airflow.
• There are almost no gaps for air to escape from the top of the hotswap bay.

 
In order to keep the drives adequately cooled, I need the fan below them running much faster than Iʼd like from a noise point of view.
I will say though, it does do a good job of vibration isolation. I donʼt know about drive-to-drive vibrations, but thereʼs only a very slight “rumble” from the case caused by the drives sitting idle. (I donʼt let my drives spin down)
 
 
So I spent far too much money and replaced everything with Noctua hardware today.
 

• I replaced three Silverstone AP181 fans with NF-A15 PWMs
• My Thermalright True 120 with Corsair SP120 Quiet Edition fans (previously Scythe Gentle Typhoons) was replaced with an NH-D14 SE2011 and an 115x mounting kit. (the regular NH-D14 does not include PWM fans)
• And the AF120 Quiet Edition exhaust fan was replaced with an NF-S12A PWM

 
I was not prepared for the sheer size of the NH-D14. I thought I had an idea of what a big heatsink was like, but I did not expect it to be this big.
I suspect its actual cooling performance wonʼt be significantly better than my True 120 though, as its fin density seems to be a lot lower (possibly half that of the True 120?) but thatʼs because itʼs engineered for silence rather than cooling over everything else.
I really like the fan mounting system they use—the True 120 fan clips were a nuisance and I just zip-tied the fans to it instead.
 
Even with the “heatsinks” removed from my Corsair Vengeance memory (it actually came off when I went to pull out one of the RAM sticks—I think theyʼre just held on with double-sided tape!) I was having trouble fitting the NH-D14 without raising the 120mm fan off a few rows of fins at the bottom, so I changed it from push-pull with the 120/140mm fans to having the 120mm on top.
Thereʼs so much airflow around the CPU area in this case (150mm fan directly below, then the NH-D14, and a 120mm fan directly above it) that Iʼm not concerned about it affecting cooling performance in the slightest.
 
 
Switching the AP181 fans with NF-A15 PWMs confirmed something I had feared—the “fan oscillation” I had been hearing was not being caused by the AP181 fans, but by the mounting system in the case.
There are three filtered intakes and each fan is mounted to a bracket above them. If you open the side door, you can just slide the filters out.
Remove two screws, and the fans will slide out on their mounting bracket—which does make things nice and easy for cleaning.
Itʼs a nice idea, but having to take the side off the case to clean the filters is a nuisance, the way they are mounted creates a gap inside the case below the fan filters (which I had not realized, so that was full of dust) and itʼs just not rigid enough.
 
Even with the NF-A15 PWMs running normally, I was still hearing the annoying oscillations from the case. Thankfully though, it turns out that the intake below the hard drives is unaffected, so that fan can run at a decent speed to keep the drives cool, and with low noise adapters on the other two, they run at 540 rpm which is low enough to prevent this annoying noise from occurring.
As theyʼre PWM fans, I would have preferred to let the BIOS fully control them (there shouldnʼt be a need for the LNA) but its minimum speed for case fans was too high—it turns out that only the CPU fan can be reduced to 25%, the case fans have a minimum speed of 60%.
 
 
The NF-S12A impressively quiet. This thing is totally inaudible when the other fans are running, even though Iʼm running it off a non-PWM header.
Now if you had an otherwise fanless system thatʼs only got an SSD in it, Iʼm sure you could probably hear it, but I canʼt tell whether itʼs switched on or off with the PC running.
 
 
I have the CPU fans set to run at 300 rpm most of the time (the minimum for the Noctuas) with the “silent” fan profile in my BIOS, and it stays around 40℃.
Under sustained load itʼs hitting 65℃ on the hottest core and thatʼs about 4℃ hotter than all the others.
I could probably do better with a re-application of Noctuaʼs TIM (Iʼm used to Gelid GC Extreme) but Iʼm not too concerned about it. The fans are only spinning up to about 900 rpm which is very quiet (12.6 dBA according to the Noctua site) so Iʼm just going to leave it.
 
Unfortunately when setting this all up, I found out that my motherboard only has two PWM fan headers, and I think the other just supplies 12v.
Well, Iʼm not sure about the latter—the fan speeds reported are lower than Noctua suggest they should be when running off 12v, so perhaps itʼs doing something to reduce their speed, but I donʼt have any real control over them in the BIOS.
 
So the CPU is running at 300 rpm via PWM, and two of the 150mm case fans (with a splitter and low noise adapters in place) stay around 540 rpm due to the limitations mentioned above.
The third 150mm fan under the hard drives is running off a splitter with the NF-S12A. That fan should be running at 900 rpm, but itʼs reported as being 750 rpm even without the splitter in place, so Iʼll have to investigate that tomorrow.
What I will probably do is remove the splitter and just put the NF-S12A on that header, and run the NF-A15 off a molex adapter. The hard drives are running a little hotter than Iʼd like (though it turned out that someone blocked half the intake of the PC tower earlier ) so hopefully running it off 12v with a low noise adapter (900 rpm) will be suitable.
 
 
The end result after all this: silence.
 
Well, not quite silence—there is a gentle rush of air from the tower when youʼre near it but this is a significant improvement, and itʼs only running slightly warmer than before. (around 5℃)
I cannot tell whether the PC is on at all from another room now, even in the dead of night.
 
The loudest fan in the case is now the “blower” on my GTX 570, which Iʼm considering replacing with something else now (if only Noctua made GPU coolers...) or I might just wait until I upgrade later in the year—itʼs not an objectionable noise as it rarely ever goes above the minimum speed.
 
One other thing which would bother me if I was in the same room as the PC while using it, is the power supply. I use a Corsair AX850 which should basically never be running its fan, but you can actually hear some noise from the components inside the PSU now that the PC is so quiet. I can see why Corsair now sell the RM series where every component is picked for low noise—fanless does not mean silent.
I will almost certainly be replacing the PSU in my next build, and Iʼm unsure about whether or not I should go with a low wattage one and replace the fan with a Noctua, or go with a high wattage one so the fan never spins up. Iʼm actually leaning towards the former.
 
 
I think I'm probably a Noctua fan (hah!) for life now.
They really have solved the problems I hear with every other PWM fan—oscillations or ticking noises—even at their minimum speed of 300 rpm.
There are basically no mechanical noises from the fans, and all you hear is the air theyʼre moving.
Other fans may offer good performance in the 900-1200 rpm range, but Iʼve never encountered another fan that doesnʼt produce some kind of distracting noise at these low speeds.
 
I really regret having bought the Corsair fans over Noctuas a year or so ago.
Reviews said that the Corsair fans were quieter than the Noctuas when pushing the same amount of air, and while that may be true if youʼre using an SPL meter, I donʼt think the two even compare when it comes to the quality of noise they produce.
 
I was still hesitant to buy the Noctuas after reading reviews/comparisons recently, but for me, they were absolutely the right fans to get.
The funny thing is that Scythe's Gentle Typhoons (the fans I replaced with Corsairs last year) are rated highest from a noise/airflow perspective on a site that takes a “scientific” approach to fan testing/reviews, and I hated those.
I donʼt know if itʼs a flawed testing approach that gives those results, or simply that an SPL meter doesnʼt measure the quality of the sound, but from my point of view, the Noctuas are worth every penny.
 
Maybe if youʼre water-cooling and the fans are being used on a thick, dense, radiator at relatively high rpms the Noctua fans donʼt do as well as the Gentle Typhoons, but at low speeds as case fans or on a lower density cooler such as the NH-D14 (which makes up for its fin density with sheer size) I havenʼt heard anything that comes close.

[mwillems]

The end result after all this: silence.
 
Well, not quite silence—there is a gentle rush of air from the tower when youʼre near it but this is a significant improvement, and itʼs only running slightly warmer than before. (around 5℃)
I cannot tell whether the PC is on at all from another room now, even in the dead of night.

That's great news; it's very rewarding to not be able to hear your computer (especially after spending a good amount of time on it!)
 

One other thing which would bother me if I was in the same room as the PC while using it, is the power supply. I use a Corsair AX850 which should basically never be running its fan, but you can actually hear some noise from the components inside the PSU now that the PC is so quiet. I can see why Corsair now sell the RM series where every component is picked for low noise—fanless does not mean silent.
I will almost certainly be replacing the PSU in my next build, and Iʼm unsure about whether or not I should go with a low wattage one and replace the fan with a Noctua, or go with a high wattage one so the fan never spins up. Iʼm actually leaning towards the former.

You can get very efficient 500 or 600 watt power supplies with no fans at all now.  I'm using a fanless Rosewill right now that actually runs cooler than my old fan-cooled PSU.
 

I was still hesitant to buy the Noctuas after reading reviews/comparisons recently, but for me, they were absolutely the right fans to get.
The funny thing is that Scythe's Gentle Typhoons (the fans I replaced with Corsairs last year) are rated highest from a noise/airflow perspective on a site that takes a “scientific” approach to fan testing/reviews, and I hated those.
I donʼt know if itʼs a flawed testing approach that gives those results, or simply that an SPL meter doesnʼt measure the quality of the sound, but from my point of view, the Noctuas are worth every penny.
 
Maybe if youʼre water-cooling and the fans are being used on a thick, dense, radiator at relatively high rpms the Noctua fans donʼt do as well as the Gentle Typhoons, but at low speeds as case fans or on a lower density cooler such as the NH-D14 (which makes up for its fin density with sheer size) I havenʼt heard anything that comes close.

I had a similar experience with the Gentle Typhoons, and couldn't quite figure out why they were underperforming for me noise-wise.  The best I can figure is that, according to silentpcreview some of the gentle typhoon models are drastically quieter than others, and some of the higher RPM ones don't scale down nicely (they start whining loudly at lower RPMs, whereas the lower RPM Scythe fans are quiet at those same speeds).  Unfortunately the specific model of Gentle Typhoon that silentpcreview recommended is out of production, so after my underwhelming experience with one of the higher RPM models, I gave up on the Scythes entirely.

[6233638]

You can get very efficient 500 or 600 watt power supplies with no fans at all now.  I'm using a fanless Rosewill right now that actually runs cooler than my old fan-cooled PSU.

Yeah, I'm just not convinced it's good for power supplies to be fanless. My AX850 basically is - it only turns on the fan at about 40% load if I recall correctly, and the components I have are well below 340W most if not all the time.
Corsair do a 1000W RM-series PSU, and I can't imagine the next Intel processors and Nvidia GPUs to draw anything like 400W under load, so it would never be on.
Capacitors and other PSU components don't like heat though. When the Noctuas are this quiet, I think that buying an appropriate PSU (say the RM450) and replacing the stock fan would probably add a negligible amount of noise and help prolong the life of the PSU.
Operating the power supply below 40% load is not great for efficiency either.
 

I had a similar experience with the Gentle Typhoons, and couldn't quite figure out why they were underperforming for me noise-wise.  The best I can figure is that, according to silentpcreview some of the gentle typhoon models are drastically quieter than others, and some of the higher RPM ones don't scale down nicely (they start whining loudly at lower RPMs, whereas the lower RPM Scythe fans are quiet at those same speeds).  Unfortunately the specific model of Gentle Typhoon that silentpcreview recommended is out of production, so after my underwhelming experience with one of the higher RPM models, I gave up on the Scythes entirely.

The ones I had (I forget the specific model) were recommended as being quiet fans at the speeds I was running them at. They didn't whine, but they didn't sound good either.

[6233638]

Unfortunately when setting this all up, I found out that my motherboard only has two PWM fan headers, and I think the other just supplies 12v.
Well, Iʼm not sure about the latter—the fan speeds reported are lower than Noctua suggest they should be when running off 12v, so perhaps itʼs doing something to reduce their speed, but I donʼt have any real control over them in the BIOS.

Well I figured this out. Firstly, it turns out that my motherboard has five fan headers rather than three.

• CPU
• CHA1
• CHA2
• ACC
• PWR

Only 1 & 2 offer PWM control, and PWR is fixed at 12V.

While CHA1 offers PWM control over the fan speed, I wrongly assumed that without PWM control on the CHA2 header, it would be running at 12V.
I guess ASUS did not actually expect people to be using PWM fans, because one setting controls both Chassis headers, which explains why the minimum speed is locked at 60% even with only PWM fans connected.
So with a setting of 60% speed you have 12V at 60% duty cycle PWM on CHA1, and 7.2V on CHA2.
 
ACC was actually in the center of the motherboard hidden behind a panel, so I had to use two fan extension cables to route it over to the fan.
This now gives me independent control over the fan cooling the hard drives, so I can properly control their temperature now.
 
Hopefully everything will be PWM controlled on newer motherboards. (never thought I'd be saying that)

[glynor]

I donʼt know if itʼs a flawed testing approach that gives those results, or simply that an SPL meter doesnʼt measure the quality of the sound, but from my point of view, the Noctuas are worth every penny.

Agreed completely.

Like I said above, I haven't done serious comparative testing myself (but you did, so thanks), but I agree completely.  Like you, I bought a couple for my HTPC, and then quickly (within a month or two) replaced all the other fans I possibly could replace.  They are kinda expensive for a fan, but honestly, they're still pretty cheap in total cost.

GPU blowers are annoying.  Unfortunately, aftermarket coolers for those GPUs are often LOUDER than the original stock ones (optimized for OMG TEH OVERCLOKZ) and not silence.  The ones that are "silent" are often of dubious quality.  Plus, that's a warranty-voiding adventure.  Apple got some things right in their trashcan.

Only last thing I figured I'd mention...

Seasonic.

[6233638]

They are kinda expensive for a fan, but honestly, they're still pretty cheap in total cost.

The US prices are a lot more reasonable. I was about $250 compared to the $160 these parts cost. (at least going by Amazon's prices) But I still think it was worth it.
 

GPU blowers are annoying.

Honestly, I must be lucky because my GTX 570 is a relatively quiet card.
I actually chose a blower design due to the vertical airflow in the case, so maybe that is actually working, as it rarely ever moves above the minimum speed. If I set the fan higher than about 50% it does start to get really obnoxious though. (previously I could get to about 65% before you heard it over the other fans!)

Unfortunately, aftermarket coolers for those GPUs are often LOUDER than the original stock ones (optimized for OMG TEH OVERCLOKZ) and not silence.  The ones that are "silent" are often of dubious quality.  Plus, that's a warranty-voiding adventure.

Actually, EVGA allows you to replace the cooler/TIM and still covers your card. I think the newer ones have a more restricted warranty policy that is limited to 3 years now though. (mine was 10)
 
I am almost certain that my next card will be an MSI one with their "Twin Frozr" cooler - it seems to be exceptionally good: http://www.techpowerup.com/reviews/MSI/GTX_780_Ti_Gaming/22.html

Apple got some things right in their trashcan.

When it's hitting 95℃ on the CPUs, 97℃ on the GPUs, and throttling, I'm not sure that I agree. It might be quiet, but I don't consider that to be providing adequate cooling.
 
I'm not even too keen about my CPU hitting 65℃ on its hottest core, but that's an overclocked chip.

Only last thing I figured I'd mention... Seasonic.

Yes, my previous power supply was a Seasonic and they're very good. I do like that Corsair specifically picked every component in the RM series based on its noise level though. While Seasonic are very well engineered, I'm not sure they make the same claims. (no coil whine etc.)

[glynor]

When it's hitting 95℃ on the CPUs, 97℃ on the GPUs, and throttling, I'm not sure that I agree. It might be quiet, but I don't consider that to be providing adequate cooling.

They were simultaneously running Furmark (which most GPU vendors throttle immediately in software) and Prime95 on 24 threads.  In other words, they were deliberately trying to light up ALL of the silicon inside there at the same time.

That is outrageously outside of anyone's expected use-case.  Anand was deliberately trying to make it throttle.  It did not throttle under any of their normal tests:

The same was true for the CPU. Even with all 12 cores taxed heavily, I never saw any drops below the CPU’s 2.7GHz base clock.

Next I tried a heavier workload on the CPU: a H.264 video encode. Here I just ran the x264 5.01 benchmark in parallel with the LuxMark workload. Once again, I saw no drop in CPU or GPU clocks although I believe I was approaching the limits of where that would hold true. The system was pulling an average of 410W at that point, with peak power draw at 429W.

...

So far I wasn’t surprised by the platform’s behavior. The Mac Pro’s thermal core and fan was enough to handle a real world workload without throttling. But what about a power virus? For this I ran a combination of FurMark at 2560 x 1440 and Prime95 (in-place large FFTs) across all GPUs and CPU cores.

All workstation class machines would have throttled at that load (and AMD runs their GPUs hot lately, those GPUs are qualified for operation up to 100℃).  Nvidia and AMDs consumer GPU drivers for Windows automatically throttle when you even load Furmark, without fiddling with them.

[6233638]

They were simultaneously running Furmark (which most GPU vendors throttle immediately in software) and Prime95 on 24 threads.  In other words, they were deliberately trying to light up ALL of the silicon inside there at the same time.

That is outrageously outside of anyone's expected use-case.  Anand was deliberately trying to make it throttle.

As someone that used to build custom PC's for a living (that was a long time ago now) that's standard testing procedure.
Just because they hit that with an artificial test does not mean it can't happen if they're rendering video on the GPUs and doing something else which is similarly intensive on the CPU. (though I agree it's difficult to load up a 12-core CPU that high - I'd expect to see it more frequently on the 6/8 core parts which have the same TDP and higher clocks)
Even if it's not throttling under "normal" loads (consider that the GPUs are underclocked out of the box too...) I don't think those temperatures are acceptable.

And if you read to the end of that page:

With both subsystems combined you'll get the 437W figure I talked about earlier.

You don't even need a power virus to get there. Running two realistic workloads that are heavy enough won't get you up to the 463W point where I saw throttling, but it'll get you close to the max average power I recorded on the Mac Pro.

Considering that it has a 450W power supply, I don't like it. They're pushing things far too close to the power and thermal limits with the Mac Pro design.

P.S. calling this a "power virus" is just making excuses for Apple. That's thermal testing/stability testing. A power virus is malware which does this without the user's interaction.

All workstation class machines would have throttled at that load (and AMD runs their GPUs hot lately, those GPUs are qualified for operation up to 100℃).  Nvidia and AMDs consumer GPU drivers for Windows automatically throttle when you even load Furmark, without fiddling with them.

I've seen a surprising number of Apple users saying that they plan on upgrading from a Mac Pro 1,1 to the new Mac Pro.
 
If it's consistently hitting those temperatures/power consumption, I can't see the new Mac Pro lasting another 7 years. It might just last your 3 year AppleCare warranty period though.

[glynor]

As someone that used to build custom PC's for a living (that was a long time ago now) that's standard testing procedure.

I still build them.

It isn't a long time ago anymore, and Furmark isn't a regular benchmark.  It is designed to throttle GPUs.  Running it at the same time as Prime is crazy-pants.

AMD and Nvidia both call it malware.

[glynor]

Just because they hit that with an artificial test does not mean it can't happen if they're rendering video on the GPUs and doing something else which is similarly intensive on the CPU. (though I agree it's difficult to load up a 12-core CPU that high - I'd expect to see it more frequently on the 6/8 core parts which have the same TDP and higher clocks)
Even if it's not throttling under "normal" loads (consider that the GPUs are underclocked out of the box too...) I don't think those temperatures are acceptable.

Anand tested all of that.  They did a simultaneous x264 encode while running a 3D benchmark.

You don't know what Furmark is, which is fine, but modern GPUs are NOT designed to have all the active parts lit up simultaneously (which is what Furmark does).  No game, 3D modeling program, or anything else, no matter how demanding, uses them this way.

It isn't a real-world kind of "worst case test" (ala Prime95).  It is a "there is no way this could ever happen in the real world, just beat on it till it cries momma" test.

[glynor]

A power virus is malware which does this without the user's interaction.

Again, you don't know what Furmark is.

That's Nvidia's name for it, not Anands, and AMD agrees.  Furmark, and applications like it, have been widely referred to as Power Viruses (since Nvidia coined the term) for probably 3 years now.  There are a few others of these applications around now, but Furmark was the first.  Like I said, it isn't a regular benchmarking test like Prime95 is or running other GPU tests.  Furmark is designed to make the GPU generate as much heat as is possible.

Both Nvidia and AMD had to specifically design their throttling to take it into account, because people were burning out their GPUs running it.

EDIT:  I looked it up.  It was actually AMD that started calling it a Power Virus, way back when the Radeon HD 4xx0 line was brand new (2008/2009), not Nvidia.  But both companies have agreed on this point (and if both of those two agree, well then, that's something).  I'm not saying Furmark isn't useful in some cases, though prolonged use is a terrible idea and is a good way to damage your PC.  It isn't a benchmark.  It is designed specifically to test the thermal design limits of a device and intentionally make it fail, in order to see where it failed, to make improvements when overclocking (though it is debatable if this is valid, I think sometimes it can be).  It is not a valid "overclocking stability test" tool, though, which is where I took issue.

I'm glad Anand did the test.  It was interesting and amusing.  But, it has to be viewed in the right context.  Running that test isn't a "stability test" (like running Prime for 24 hours on a new rig).  It is the equivalent of the part of a MythBusters episode when they stop caring about the validity of what they're testing, and just try to make the thing blow up.

If the new Mac Pro had throttled with that previous test (multiple x264 encodes simultaneous with a 3D benchmark), then I'd have agreed with you.  Even that is an absurd test that would never happen in the real world, but it would indicate things about the longevity of the parts and whatnot, and indicate that a part might be more troublesome as it ages.  But, going by the furmark+prime numbers and saying "see, cooling system failure" isn't fair.

I'm also not sure your concern about the power supply is warranted.  PSU ratings are a fuzzy business.  Apple's published specs have always pretty much matched the thermal design of the systems they sell, and they are usually very conservative (my 2010 Mac Pro had a similar PSU rating, and could handle spikes well above the rating without issue).  Past success isn't always an indicator of future quality, but based on the history of these machines, I'd guess they'll be just fine, and many will still be kicking in 5-10 years.

We'll see.  I wouldn't ever buy one of these myself for many other reasons, but I think the cooling system they designed is interesting and innovative.  It is silly that we have separate "isolated" cooling systems for our devices, and it creates the "cheap, crappy blower GPU fan" issue.  Anand talked about it a little on the most recent AnandTech podcast.  Very interesting.

[6233638]

I still build them.

Oh, I still build PC's too, but not as a way to make a living any more.

It isn't a long time ago anymore, and Furmark isn't a regular benchmark.  It is designed to throttle GPUs.  Running it at the same time as Prime is crazy-pants.

It's not designed to throttle them - it's designed to push them to their limits. If I recall correctly, AMD called it malware because they did not have thermal protections in place with their older GPUs.
It can be an effective tool as long as you are monitoring what's happening rather than leaving it running overnight or something like that.

Anand tested all of that.  They did a simultaneous x264 encode while running a 3D benchmark.

Well actually I was thinking about GPU-accelerated encoding (which is what the Mac Pro seems built for) but the problem is that the review has no information on what kind of temperature/noise you get from the system when doing this sort of thing.
OK, the fans are about 45dB at 1900rpm - how often does it spin up to that? Is that the maximum speed?

It isn't a real-world kind of "worst case test" (ala Prime95).  It is a "there is no way this could ever happen in the real world, just beat on it till it cries momma" test.

Some games can do similar things to Furmark - it's just that Furmark is a well known tool for doing this - like Prime95/Intel Burn Test.
I seem to recall Starcraft 2 gaining a reputation for burning out GPUs as well before Blizzard patched it.

Furmark is designed to make the GPU generate as much heat as is possible.

Yes, and adequate cooling should prevent it from hitting temperatures like 97℃. But AMD's recent cards have been reaching ridiculous temperatures for a while now. Nvidia seems to target a maximum of 80℃ with their designs - which is still very hot, but not ridiculously so.

It is designed specifically to test the thermal design limits of a device and intentionally make it fail, in order to see where it failed, to make improvements when overclocking (though it is debatable if this is valid, I think sometimes it can be).  It is not a valid "overclocking stability test" tool, though, which is where I took issue.

I would primarily use Furmark for thermal testing, but it will also show issues with stability. There are much better tools specifically for testing stability though (especially for extended testing)

If the new Mac Pro had throttled with that previous test (multiple x264 encodes simultaneous with a 3D benchmark), then I'd have agreed with you.

Well, it's throttled down to 650MHz as standard, and the boost clock is lower than the reference cards as well. It might not be throttling any further under normal loads though.
But the problem to me is not that Furmark creates a lot of heat and the GPUs have to throttle back, it's that high GPU usage was causing the CPU to throttle back.

I think the cooling system they designed is interesting and innovative.  It is silly that we have separate "isolated" cooling systems for our devices, and it creates the "cheap, crappy blower GPU fan" issue.

I agree that the Thermal Core concept is very interesting, and I think Anandtech's testing was surprisingly lacking in this review. (though I see some sections have been updated since I read the review - which seems to be a trend there now)
It was more focused on Apple history and price-per-performance than actual real-world testing.
 
Where are the temperatures from running a GPU encode and using the CPU full tilt in "real world" tests? Is it still hitting almost 95/97℃ but not throttling back any further, or is it staying at reasonable temperatures?
What are the noise levels like when doing this? It seems like it should stay relatively quiet at idle, but Apple hardware tends to get really loud when you actually start using it.
I go back and forth on the shared heatsink design. On one hand, it means fewer fans, and looking at the individual components, they are potentially connected to a much larger heatsink.
But when you start sharing the heatsink, running your CPU full tilt also heats up your GPUs. You're relying on one heatsink and fan to cool three components now.
 
With individual components, if my CPU is running at 100% it has little-to-no effect on my GPUs and vice-versa. So those fans can stay at idle speeds and only the CPU needs to spin up.
With separate heatsinks you may have more fans running, but quieter fans overall. (because three 13dB fans ≠ 39dB)