Only just saw this here. So, to address much that already went on, more completely than by line items...
By 2003, I had my cooled the following way:
Athlon XP with an AX-7 and 120mm Nexus, undervolted (don't recall if 5V or 7v)
Geforce 4 Ti with a huge north bridge cooler, and Gamma28 at 5v pointing to it.
PSU fan forced undervolted.
I moved to a custom case, with no direct paths for noise to get out, once completed. The droning noises of fans tends to be rather annoying, so I was thrilled to be rid of them. I would love nothing more than to have good performance, yet get back to that point, with an SSD (even quiet HDDs need some work to be hidden in such a system). These days, I do sometimes need my PC in a professional capacity, so highly custom 'projects' involving cooling it are pretty much out of the question, unlike back then.
At work, there's enough noise around that I don't hear my Optiplex, usually. At home, the same class of PC (same thing with less RAM and an i3, I have worked on at home) is, at idle, about as loud as my desktop under GPU load (a bit quieter than synthetic tests, but subjectively similar compared to gaming).
Every step from that Athlon XP system has been some form of regression, during times with a gaming GPU. After ATI's amazing drivers (I <3 AMD), I sidegraded from a 9600XT to a FX 5900XT. I managed to keep it acceptable with a Zalman ZM80C-HP, and a couple weeks of experimenting with a 120mm nexus (I had to add one to stir air around in the case). That ended up being audible, if barely, all the time :\.
After an upgrade not for gaming (obviously it was easy for the ~3 years I was out of gaming, with a passive 7300GT and Core 2 Duo), and getting back in, I got a decent (Gigabyte) GTX 460. It was OK for a little bit, especially after BIOS modding to set the fans lower, and with voltage and clock control in early drivers that went both ways. But, after awhile (6 months, maybe?), bearing noises began, and I replaced the cooler with an MK-13. That was OK, but not what I wanted (the GTX 460 was at the edge of its real capability, so the fans couldn't always be super slow). Later, the MK-26 finally was enough to
really quiet it down, to where it was only just audible under load. However, it also used up
all my PCIe slots. I would have gotten USB 3.0 on that PC, were a slot available, and passed on that because of only having available PCI slots (for my current PC, I had to go find new bolts, because the backplate and bolts stuck out too far to fit, rather than just blocking a slot, which also meant just going by feel from the screwdriver for getting even torque--
I'll pass on a repeat of that!).
I ended up needing to change to a normal case, just due to space constraints, in which I was able to get it inaudible, partly by putting the PC itself in an inconvenient spot, that, through some luck, did not reflect back to me what little noise was being made.
I've built PCs with nice quiet cases, like the P180, P183, Define, R4 and Define Mini. Even with added absorbing material, they cannot mask fans, once the fans reach a point where they could be heard if freely spinning. Simple dBA SPL is a useful number, but a weighted average level. Any sound absorbing material thin enough (like <2" thick) to go into a PC simply cannot attenuate enough, in the low kHz or lower, to make a difference, except at high frequencies. If using parts don't make much high frequency noise at all, and don't transfer vibration to the case, the added mass and materials don't do any more good that in a case without them, but it otherwise well designed (things like a door v. open from fan grill, FI, still make big differences).
Here is an example (I don't want to hotlink, so scroll down to the frequency graph):
http://www.silentpcreview.com/article1325-page4.html
The noise made is directly correlated with RPM, increases at a rate equal to or greater than RPM, and is low enough in frequency that much of it will simply pass right through the barriers of the case. The only way to reduce its presence is to reduce the amount made (as can be seen, Asus did well removing the really annoying high frequency tones, already, for that cooler). That can be done with bigger fans running slower, to a point. But, once slow enough, like say a ~300 RPM 180mm fan, the pressure changes slowly enough to start hearing, "wub-wub-wub-wub," noises from the air chopping action, even feet away, if the intake isn't indirect.
Watts generated have to go somewhere. Once looking at Watts in = Watts out, there aren't even expensive drinks
, much less a free lunch. While low idles got better than Fermi over the last few years, light loading, like video watching, accelerated desktop programs, and old games, still brought GPUs from both vendors to high enough power consumption that, in terms of managing the noise, a new card would be only a minor upgrade (almost no upgrade in noise terms, while still requiring effort I did not want to repeat). In light of nVidia's response to the ROPs and RAM issue (I can accept that some things slip through the cracks, but they are not handling it well), if the upcoming Radeons are as good, I might just side-grade, like I did the other direction way back when. If it's going to use 50W+ outside of a game, it needs fans spun up, and in a 2-slot space, they can only be made so quiet, without reducing the amount of heat that needs to be evacuated.
I've heard them. I've built and worked on PCs with them. I am not denying that there are whisper quiet options for either camp. There are, and that's good (especially from Asus, MSI, and Sapphire), but they still sound like annoying little fans, by and large, and my room gets quiet enough to readily hear them, in cases, at whatever dBA the annoying frequencies happen to be. Maxwell's scaling in actual power used, however, much like Haswell's, when in between maxed out and idle, is exceptional, compared to what has been released so far. Asus, MSI, and Gigabyte are the ones that did the real legwork, here (GB's may not turn the fans off, but they did very well on the VRMs and load noise). But, the GPU's capabilities and performance are what made it possible, and there are only three models of the same series on the market right now equally capable (at very different price points, and with the 960 really needing a Ti performance gap filler).
My PC aught to only indicate that it is on by peripherals responding. I should be able to hear squirrels rustling in the trees at night, rather than PC fans. I have reached that point at idle, with an unmodified card, which I consider to be quite feat, given prior needs to get there costing a hefty added chunk of money and time, as compared to a <$20 premium over the cheapest models. If I could achieve that under load, without the time and expense of a custom WC loop (assuming I can trust reports of noise of certain pumps, mounted well, the last time I checked it would have been ~$250, and likely more in the end, plus the time and effort involved), or major pain of a huge air cooler, I would go for it. Dealing with Prolimatech's monstrosities really turned me off to doing that, again, however effective they are (but, with up to 220W possible when gaming, smaller won't work).
I've been a member at SPCR something like 70-80% as long as I've been one here. It's something I'm rather obsessive about. That a couple of major manufacturers partly cater to those of like mind, made the choice pretty simple, it's not merely hype, and alternatives are currently quite costly, in relative terms. Now, granted, if they could run the fans at 200-300 RPM, but not 0, at idle, the results would be basically the same, with sufficiently good fans. That particular decision was primarily marketing (though, it also allows cheaper fans, since small fans that can run slowly for a long time tend to be maglev, or high-friction 'fluid' bearing types). But most of the time, that's either not going to be good enough to do the job, with the heatsink restrictions, or the fans simply won't stay spinning at such low speeds.
Facebook
Twitter