PWS-920P-SQ noise?

[anlin]

Hi everyone! Long time lurker, first time poster.
I just got a pair of PWS-920P-SQs for my backup server project, and was very surprised to find out that while running, the 920P-SQs (individually) are quite a bit louder than the 1k28P-SQ in my main file server. The noise isn't bad by any means, but it is clearly audible across the room (with no other fans running in the server), whereas I cannot hear the 1k28P-SQ more than a few feet away even in a dead quiet room. With the other (3x NF-F12, 5x NF-B9, 2x NF-A8) fans plugged in, I can just about hear the slight buzzing (motor) noise the PSU fans make over the CPU and chassis fans if I'm close to the server. Both of the 920P-SQs produce the same amount of noise, so it would be rather unlikely that both PSUs would be defective. IPMI/PMBus sees both PSUs correctly and the fan speeds are about 2300-2400rpm for FAN1 and 3400-3500rpm for FAN2. For comparison the fans only run at about 1200-1500rpm on my 1k28P-SQ. At what speeds should the 920P-SQ fans typically run? I must say I am a bit disappointed that the 920P-SQs are so audible compared to the 1k28P-SQ.

 

[i386]

The psu fans in my 836 run a little bit faster, but I can't hear them as most of the noise comes from the rear fans or the hdds (hgst 6tb nas drives).

 

[j_h_o]

How much load are you generating? I imagine the fan speeds are a function of max load for the PS; what happens if you cut the load in half on the 920P-SQs?

 

[anlin]

I'm pulling about 80 watts at the wall, unfortunately I can't go much lower than without powering down the system (which stops the PSU fan as well). That being said, I'm still at less than 10% of the rated capacity of these PSUs.

The thing I'm most perplexed by is why the supposedly louder 1k28P-SQ is so much quieter than the 920P-SQ. Based on everything I've read, I was hoping the 920P-SQs would at least be as quiet as the 1k28P-SQ (ie. almost inaudible). I wonder if I have a particularly quiet 1k28P-SQ, or louder than normal 920P-SQs.

 

[Matt Lund]

I'm mostly a lurker too. Interested in your conclusion as one of these PSU models is probably in my near future and I now live in a house without a lot of remote placement options.

 

[Fritz]

Don't have any 920P-SQ's but I can attest to the 1k28P-SQ being very quiet. I have one running right next to me and I can't hear it over the noise of the rest of the fans. Overall, the system is as quiet as any desktop even tho it's fully loaded including 8 HD's.

 

[BLinux]

something doesn't seem right. i have 2 pairs of PWS-920P-SQ in both my 846s. they are dead silent... my idle is about 120W and peak load is about 360W and I never hear them.

 

[anlin]

BLinux - What kind of fans are you running in your 846s? Could you do me a favour and check the speed the PSU fans run at (this can be accomplished with IPMIView or ipmitool)? Compared to the San Ace fans that come with the chassis, the 920P-SQs are indeed whisper quiet, even one fan running drowns out the PSU completely. With the original fans replaced I can just about hear the motor noise from the 920P-SQs (a difficult to describe but very distinct rapid ticking/buzzing sound, doesn't sound like a broken or worn out fan at all). With the 1ks8P-SQ I can hear a similar buzzing sound but only when I'm very close to the server and with all other fans unplugged or spun way down. Noise from air movement is also louder with the 920P-SQ (understandable considering the 920P-SQ spins its fans about twice as fast as the 1k28P-SQ), but is easily drowned out by the case fans. It's the motor noise that is the most audible from a distance.

 

[anlin]

Attached are a few recordings of the PSUs I made with my phone. Unfortunately the quality isn't great, but as far as I can tell, the phone doesn't apply much if any compression to the recordings. I haven't artificially boosted the audio at all, so you might need to turn the volume up quite high. There shouldn't be any loud clicks or other sounds present as these were recorded in a dead quiet room. I recorded each PSU about a hand's width from the rear of the case and from inside the case. The rear of the case recordings give a rough idea of the kind of air flow noise each PSU generates, whereas the inside case recordings emphasise motor noise. I had to leave a single NF-F12 running to cool the CPU, the fan can be heard ramping up and down in some of the clips.

The rear of the case recordings sound quite similar to the ones recorded by @OutOfSync here: Supermicro PWS-1K21P-1R, PWS-501P-1R, PWS-920P-SQ noise comparison, which leads me to believe the 920P-SQs might simply be a louder PSU than the 1k28p-SQ. The PSU fan RPM readings I'm getting are also close (if not exactly the same) to those reported by @OutOfSync.

 

[svtkobra7]

Interesting post ...

Could you do me a favour and check the speed the PSU fans run at (this can be accomplished with IPMIView or ipmitool)?

I know this question was not for me, but I've always been curious - what is the difference between the two fan speeds I've highlighted?



FWIW I have two PWS-1K21P-1Rs one is rev 1.2 and one rev 1.4 (no clue what the difference is) and the rev 1.4 was massively louder than the rev 1.2 one (and fan speeds MUCH higher). I removed it and only run one PSU now.

I wish I had the "problem" of a PSU being the noisiest component in my server.

 

[anlin]

I know this question was not for me, but I've always been curious - what is the difference between the two fan speeds I've highlighted?

Higher power Supermicro PSUs have stacked contra-rotating fans - my guess would be that IPMIView pulls the rpm readings for both individual fans but for some reason shows the reading for one fan under the PMBus tab and the other under the FRU tab.

 

[drewthor]

Attached are a few recordings of the PSUs I made with my phone. Unfortunately the quality isn't great, but as far as I can tell, the phone doesn't apply much if any compression to the recordings. I haven't artificially boosted the audio at all, so you might need to turn the volume up quite high. There shouldn't be any loud clicks or other sounds present as these were recorded in a dead quiet room. I recorded each PSU about a hand's width from the rear of the case and from inside the case. The rear of the case recordings give a rough idea of the kind of air flow noise each PSU generates, whereas the inside case recordings emphasise motor noise. I had to leave a single NF-F12 running to cool the CPU, the fan can be heard ramping up and down in some of the clips.

The rear of the case recordings sound quite similar to the ones recorded by @OutOfSync here: Supermicro PWS-1K21P-1R, PWS-501P-1R, PWS-920P-SQ noise comparison, which leads me to believe the 920P-SQs might simply be a louder PSU than the 1k28p-SQ. The PSU fan RPM readings I'm getting are also close (if not exactly the same) to those reported by @OutOfSync.

Thanks for posting the audio recordings! Can confirm the PWS-920P-SQ is rather quiet, and much quieter than the non-SQ versions, but it ends up being "loud" in my experience as there is a very distinct and annoying high pitched almost whine that it emits when the fans are spinning. From your recording for the PWS-920P-SQ I can pick out the high pitched whine as well. Sounds like I should either consider getting a PWS-1K28P-SQ (after verifying if my PDU even supports those) or removing the psu wall and just plopping an ATX PSU in there.

 

[Fritz]

Thanks for posting the audio recordings! Can confirm the PWS-920P-SQ is rather quiet, and much quieter than the non-SQ versions, but it ends up being "loud" in my experience as there is a very distinct and annoying high pitched almost whine that it emits when the fans are spinning. From your recording for the PWS-920P-SQ I can pick out the high pitched whine as well. Sounds like I should either consider getting a PWS-1K28P-SQ (after verifying if my PDU even supports those) or removing the psu wall and just plopping an ATX PSU in there.

A PWS-1K28P-SQ will work with a little help from a dremel tool.

 

[drewthor]

A PWS-1K28P-SQ will work with a little help from a dremel tool.

I lucked* out that my CSE-846 has a PDB-846-8824 PDU so should be good to go with the PWS-1K28P-SQ

 

[drewthor]

To anyone in the future, the PDB-PT846-8824 PDB is a 19 pair pdb so the PWS-1K28P-SQ does not immediately fit in place. You will have to do what @Fritz mentioned to trim the extra grounds from the pdb to make it fit. PDB-PT846-2824 has 23 pairs and would be the pdb to get. $180 on ebay at time of this post so very spendy.

@Fritz are there any posts or videos you know of that show which pins to trim?

 

[Fritz]

To anyone in the future, the PDB-PT846-8824 PDB is a 19 pair pdb so the PWS-1K28P-SQ does not immediately fit in place. You will have to do what @Fritz mentioned to trim the extra grounds from the pdb to make it fit. PDB-PT846-2824 has 23 pairs and would be the pdb to get. $180 on ebay at time of this post so very spendy.

@Fritz are there any posts or videos you know of that show which pins to trim?

I just held the two versions together and trimmed the ground (I think that's what the solid section is) using a dremel tool until the two matched. I have several done and have never had a minutes trouble with any of them.

But I will embarrassingly admit that I learned about this the hard way. I shoved a PWS-1K29P-SQ so hard into a SC846 that it busted the PDU plug. Took awhile for one to show up on eBay that was priced what I was willing to pay but eventually one did. I've had no further problems of this sort since. There are several versions of the SC846 PDU so if you're in the market do your homework. The oldest one doesn't have the second 8 pin connector. I had to jury rig it to get me by until the proper PDU showed up.