Can somebody recommend me a really quiet 1000w modular power supply. I am about to build a sandy bridge PC with 2 high end nvidia cards in SLI.

Welcome to SPCR !!

I'm quite sure you don't need a 1000W PSU even for 2 GTX 580 (244W TDP each) and an i7-2600K (95W TDP).
How about the Corsair AX850 ?

[EDIT : 719W with 2 GTX580 and and OCed i7 965 ; link : http://www.guru3d.com/article/geforce-g ... -review/14]

More probably that not the most quieter gaming PSU are currently the Nexus RX-series (33dB at 850W DC tested by SPCR).

If you want to stick with 1kW power level (not recommended, IMO), among a few others maybe you could consider the new Antec HCP-1200 (the Antec CP-series should be a too much louder at full power but quite good around 750W DC), the Zalman ZM-1000HP (more than a tad loud at full power, but acceptable under 850W DC), and eventually the Cooler Master Silent Pro M (not the newer Silent Pro Gold, which seem to be far louder, according to OklahomaWolf). I would avoid the newer Enermax Revolution-series, highly overpriced and however a footstep backwards with reference to the old Revolution 85+-series (IMVHO).

Personally I would go for a Nexus RX-8500 (850W, if I gamed very often), or a Corsair/Seasonic X (on the contrary: likely Seasonic should give an advantage of a couple of dB at load over the Corsair).

A lot of it's down to the case. Using a HAF932 that vents at the bottom, My ZM850-HP is dead quiet [idle fan speed as far as I can tell] even when drawing 700W (800W mains). If you look at the hotbox vs open-air tests of some similar large PSUs, that isn't that surprising, as there's up to a 300W offset in noise/load in some cases. Since the Zalmans remain at idle fan speed beyond 400W, being idle beyond 700W in a system with separate ventilation for the PSU is very believable.

Of course, the HAF932 is terrible for silencing the rest of the case, just good at cooling it, but nonetheless this highlights the importance of case design.

frenchie: Given TDPs for geforces are actually false, two GTX580s will use a max of around 580W in some titles. That's still ok on an 850 for a 95W CPU with moderate overclocking, but with a higher speed, or when using a 130W CPU, or if using several drives, an 850 is going to be 1) near its safe limit and 2) on the whole, very noisy.

samuelmorris : I realized that after I posted, hence the link in the edit (2xGTX580 and overclocked i7 965)

If anything, the 850W choice was more a case of 'I refuse to cave to a kilowatt PSU'. I could afford one, but decided to go a bit lower. four figure PSUs are insane
Still, it's done pretty well so far, 2 years old next month, and I have yet to ever hear it make any noise.

Or I haven't well understood what you said, or I think it is just not possible: the ZM1000-HP at 700W DC scream out around 40dB, even if in the SPCR hot-box (IIRC). In a normal operation it could be lower by 10dB, or 15dB, but even 25dB is plainly loud, not "dead quiet".


Do you have some source for this state?
Actually TechPowerUp! (a somewhat respectable source) maxed out the SLI at 444W when gaming, while giving 627W when benching and disabling the built-in current limiter.

I'm not lying to you, this PSU is quieter at 700W than my Corsair HX 520W at idle. Look up SPCR's Corsair AX 850W review, and look at the difference between hotbox and open.
The ZM1000-HP is 20dB at 400W. Apply the same 300W to that that the AX850 has, and you get 700W at 20dB. [That's in the live test room too, not the anechoic chamber]

I wasn't thinking you're just plainly lying.
If in case you're just referring a subjective sensation: so it cannot (IMO) be granted to other people/situation (I mean that what you call "quiet" may be different from what I call "quiet").

However the above quoted AX850 is 25dB at 700W, and 25dB is loud to me: "deadly quiet" to me is under 12-13dB.
Frankly I'm not confident that a 2007 product might match a 12-13dB noise level at a 700W DC power level.

No, that's what I was expecting. I'm well aware people have different opinions on what is 'quiet', but I assure you, I'm very strict.
You missed my point on the AX 850 article. That 25dB figure is the same at 700W open, as it is for 400W hotbox.
The ZM1000-HP scored 20dB in the live test room, at 400W hotbox. So, given the result above, it's not unreasonable to suggest it would also manage 20dB in the live test room, when open to the air.
Secondly, the acoustic profile of the fan in the ZM850-HP is pretty good, so I could easily see it scoring something like 15-16dB in the anechoic chamber if re-tested.

I've compared it to a Nexus fan [one without any bearing problems] and it matches the noise of the fan at around 9V, even at that load level.

Maybe not unreasonable, but IMO even a bit far from probably, as law of physics can't be easily overcome, due even to the different efficiency: at 700W, 82% of rated power, in the AX850 there are about 112w of waste heat, while in the ZM1000-HP at 82% of rated power there are a bit under 200W of waste heat (about 150W at the very same 700W level).
There's to say that the Zalman heatpipe-cooled PSUs historically privilege hot running temperature to grant a low operating noise.

However, as I don't own a Zalman heatpipe-cooled PSU, I have no facts supporting my guessworks.


If it was true, the Zalman would output 13dB at 700W: a bit hard to believe without a direct experience (but far from false).

230V UK mains voltage here, that improves efficiency, which will reduce heat output and thus noise.
At 599W DC output, the ZM1000-HP recorded 87% efficiency, thus 89W of waste heat. Not a lot for a 140mm fan and a heatpipe cooler to dissipate.

I can't categorically state that the ZM850 I have is 14dB at full load, but what I can say is the subjective noise pressing my ear against the side, along with the amount of airflow coming out, is exactly the same at full load as it is at idle. It's arguable the unit could be producing a bit more noise at load and me not notice, but frankly, that's a little irrelevant. When testing the unit at around 550W load with a very quiet card [an Accelero-cooled 4GB HD5970, albeit a faulty one], the PSU was still clearly spinning at idle fan speed. At 700W i think you'd struggle to get any components that would allow the PSU to be audible.

I'm not going out there saying the PSU everyone should buy is the Zalman. As good as mine has been, they are older units, and there are more efficient models, including like the AX, ones with fans that turn off completely at low load, which will be superior for silent idle systems. However, what I was trying to illustrate was that 'noisy' PSUs can become quiet PSUs in a case with a good cooling layout, and PSUs that would otherwise be written off are very viable with the right case.

Indeed I think that (now) I understand your point.

Just for the sake of talking about, I've only said that without hands-on experience it's not easy to me evaluating such a situation.

For example, I have a modest Wolfdale based system cooled by a Scythe Kabuto. We can't rely upon TDP as the CPU is mildly overclocked (3.33GHz vs 2.5GHz stock) and slightly undervolted (1.15V vs 1.22V stock), but to summarize I am able to cool (around 63°C with an about 22°C ambient) it when running 2 instances of Prime95 plus Furmark 1.8.2 with the fan at about 650-680rpm (even if the heatsink is also close to a 500rpm exhaust Slipstream).

So I can argue that a 600gr. six pipe heatsink with a total surface of 6042 cm2 may cool more than 65W (likely around 80W) with a modest 11-13dB output noise.

Therefore, theoretically it *could* be possible to cool 90W of waste heat (at 242V) with a slightly larger fan at a similar modest noise level (it would seem the Zalman kicks in when its temp rise excess the 14°C, while the Corsair fan ramps up when temp rise just above 7°C).

But does the single heatpipe/two blocks Zalman 60x80mm heatsink be comparable with a somewhat massive 120x130mm Scythe Kabuto?
And what about if and when the waste heat was more, as at 700W DC or 850W DC levels (or with lower voltages)?
Frankly, I don't know (and I have no more guessworks to make about).

Surely it will be very interesting if MikeC and his good friends would extend the open air tests to all the recommended PSUs to give us more information about real life conditions.

Thanks for sharing.

Sometimes it would be nice to go back and re-assess older products with newer review techniques, but it is not always possible. Those PSUs are no longer in our possession -- we have limited space in the lab, so unless something is considered an important reference, it gets moved on.

It is possible that at >500W loads in open air conditions, the ZM850/1000 have lower noise than other more recent recommended PSUs. We never did test them that way until about 18 mos ago. With so many high performance cases putting the PSU on the bottom now, and providing intake for the PSU from the bottom of the case, the hotbox tester is becoming less broadly useful than it used to be. The standard hotbox results are most relevant when the PSU is conventionally mounted in the top/back corner of a tower-style case.

This may call for a change in our PSU testing procedure -- I would personally love it, as benchtop testing is much less time consuming than hotbox testing where the temperature in the box has to stabilize before any measurements can be taken at each and every power load level. I guess I need to take a look at how many popular DIY cases out there have fresh air intakes for the PSU -- and perhaps a poll in the SPCR forums. My immediate thought: PSUs up to ~500W should be tested both ways; the higher power ones will usually go into bigger, more gaming oriented cases where fresh air feed for the PSU is the norm, so they can just be tested in free air only.

Anyway, getting back to the topic on hand -- I am surprised at samuelmorris's comment that his PSU sounds the same at idle and at full load (800w AC?), that seems too good to be true. The 230VAC input and higher efficiency could be a factor... I wonder about the noise from the other components in his case. FWIW, the older live room SPL measurements match the newer anechoic chamber measurements when the SPL gets to 22~25 dBA/1m or higher. Finally, I recall/agree that this Enhance-made Zalman was very quiet, the quietest high load PSU at the time of the review (and we have reviewed very few such high power units since), the only super powerful PSU quiet enough for the EndPCNoise SPCR-designed Extreme Gamer PC.

Note that the model ZM850/1000P is no longer in Zalman's catalog. The ZM850/1000HP appears to have replaced it-- still just 80+, max 86% eff. at 230VAC input. There are also new HP Plus versions rated 80+ Silver.

As far as I know, no such product as the ZM850P/ZM1000P ever existed, it's the ZM1000-HP that SPCR tested, and the ZM850-HP I own. It was still on sale recently.
The new more efficient model could be an interesting product to test.
While I definitely think new PSUs should include open tests, I don't think the hotbox test should be eliminated, as certain cases still use that layout.


Again with fan speed forcing at full load, I was able to verify that the unit was as quiet as at idle at 600W ac. It's totally impractical to test the unit at 800W ac as the only components that can draw that much have earachingly loud fans. However, I can test the unit immediately after full load by closing a game after a long period at 800W a.c. and immediately then forcing the GPU fans down to a silent level. Doing so did not yield any noise I could hear from the PSU whatsoever beyond the quiet background environment. The amount of air coming out the back of the unit felt the same as at idle, and with my ear up against the exhaust of the unit I could hear no tone different to that of the PSU idling.

You're right -- re HP. My bad.

Your comments about the PSU's fan behavior may be right on, but then since "the only components that can draw that much have earachingly loud fans", the PSU's low noise at such power levels may be moot -- ie, you might not hear the up-ramp of the PSU fan (in another PSU) over the noise of the other components.

Something I considered, which is why I ran the PSU with my ear right up against the unit, and tested it immediately after the long period of heavy load, assuming there'd be at least some sort of cooloff period. Having tested at 550W DC with a very quiet card, I also did the same, and the unit honestly isn't audible over even the SFF21F on the CPU cooler idling at 900rpm.
It wouldn't surprise me to find perhaps that the unit had increased from, say 14dB to 19dB at the 700W load level and me not notice, but it certainly has not reached the level of making any considerable amount of noise, as I'm very sensitive to noises like that (An idling HX520 bugged me as I could hear it quite clearly).
In any case, from a purely mathematical standpoint, I think it's quite likely that most high-efficiency units of the same rating, with large fans and temperature-based (as opposed to load-based) fan controllers will have similar offsets to their noise/power results when run open-box versus hotbox, and in this case, the offset for the AX 850W at 700W open, is 300W, from 400 hotbox.
At 400W hotbox the ZM1000-HP is still at baseline, we only know it rises to 26dB at 500W DC. It's conceivable it may also have been 20dB as far as 450W. Thus, at 805W AC in the UK, or 700W DC almost exactly, that 300W offset seems to suggest the unit could quite easily still be running at 20dB.

Thanks for all your replies. I will change to a lower power supply. I will probably go for 800 or 850w.