[Supermicro Storage Cases] How quiet are the new chassis with the "super quiet" PSUs?

[bestboy]

I'm currently in need of growing my system and need a new case. The problem is that I need it to hold 12+ drives for now and to be silent.
First off I'm looking into the newer revisions of the Supermicro cases 836 and 846. They are now available with a new line of "super quiet" platinum PSUs (marked by the SQ designation: e.g. PWS-920P-SQ). In contrast to the old revisions of these cases, it is said that those PSUs are really nice and quiet ([1], [2]). But as noise levels are very subjective, I'm not sure what to think of it. For sysadmins who are used server gear, "super quiet" is not necessarily "super quiet" in terms of home users. So I tried to find out how "super quiet" sounds in reality, but I did not find much information about the newer cases.
When looking for examples I found very loud ones (e.g. here or here) and a few quieter ones that are still not what I'd call "super quiet" (e.g. here). Unfortunately in all those examples it is never clear what chassis revisions and what PSUs were used.
So in order to get a clearer picture of those popular cases, I though I'd ask around here. Are there any users of the Supermicro storage chassis running a SQ-PSU and a PWM setup? Could anyone provide a quick SPL measurement from a smartphone (Android app, iPhone app) or post a short video of it in action?
I'm sure that would be very helpful to many (home) users, who are looking into the excellent Supermicro gear.

 

[depasseg]

I have the SuperMicro SuperStorage Server 5028R-E1CR12L (12 drive/ 2 U) chassis and an SC847. I posted a couple videos around here somewhere, but I will say that the 5028R is really quiet. Like I couldn't hear it quiet. Even without the SQ PSU's. The SC847 OTOH was really loud, but once I was able to move the system fans to the Motherboard (equivalent, since it's only a jbod chassis) instead of the backplane connectors the fans dropped speed and it was a lot quieter (but still noticeable). I ran a quick test with just the PSU's (I was considering swapping them out), and I couldn't hear the PSU's with the chassis open, so the SQ's wouldn't get my system quieter.

Bottom line - look at the 5028R and don't worry about the SQ PSU's (if they come with the system, that's great, but don't limit yourself to that).

[edit: I would love to gather some SPL data, but that might require a very difficult outage. In the mean time, I can take some readings directly in front and behind each individual component and see if it provides anything meaningful.]

 

[tvsjr]

The system in my sig block has the R1K28 SQ Platinum power supplies. The PSUs basically contribute no noise to the overall system - the big noise creator are the 7 80mm fans, primarily noisy because they are high static pressure fans to move air through all the small holes in the backplane and front panel. I moved all of the fans from the backplane fan connectors (where they run wide open all the time) to the motherboard (where they are PWM controlled) which dramatically improved things.

Using the app you suggested, I measured the system at 51dB at 5 feet during boot (fans on full) and 43dB once the PWM kicked in. Obviously, if you heavily load the system the fans will spin up further... I added active CPU heat sinks as well in an attempt to keep the fan speeds low. I value quiet - my office/study measures about 35dB with the system off. I would find the system mildly irritating if I were to have it in the office all the time, but not enough to make me totally nuts.

I also have two 827s which have the 740-watt gold-class power supplies (not marked SQ) as VMware hosts. These two systems are quieter than the 4U box.

In my research on the 846s/847s, I did come across several people with the older 900- and 1200-watt (if memory serves) PSUs, which were apparently real screamers. So, I would avoid those. I typically found those supplies paired with earlier chassis revisions that had SAS1 backplanes, which weren't acceptable for me anyway (note I said typically - as always, verify all part numbers!)

I hope my ramblings are useful :)

 

[bestboy]

As @depasseg and @tvsjr suggested, it seems that the PSU is no longer the most relevant source for noise. Apparently the high efficiency PSUs have become so cool and quiet, that the 80mm high static pressure fans are now the dominating factor - even when they are PWM-controlled. That's good to know.

If this is the case, then the 826 chassis of @depasseg's 5028 and the 827 chassis of @tvsjr VMs may simply be considered truely silent, because they lack the 2 80mm rear fans. Those are usually present in the 3U and 4U chassis to feed the air shroud.

Otherwise I could not really explain why the 826/827 cases are considered silent and the 846/847 cases considered "still noticeable" at best. If

• the PSUs are truly irrelevant and
• the mid-walls with the 3 80mm fans are identical and
• all cases use a blocking backplane in front of the hard drives,

then I'd assume it is the 2 extra rear fans that make this noticeable difference.
Anyone up for a comparison? 2U case vs. 4U case having the rear fans temporarily disconnected? :)

@depasseg: According to the parts list the 5028R is already using a 920W SQ PSU. Unfortunately tho I could not find the videos you posted. I have to take another look.

@tvsjr: Yes, your ramblings have been helpful and are very appreciated. Also many thanks for doing a SPL measurement. 43dB is probably not bad for such a dense behemoth like the 847, but still a bit much for a living room.

 

[bestboy]

Otherwise I could not really explain why the 826/827 cases are considered silent and the 846/847 cases considered "still noticeable" at best. If

• the PSUs are truly irrelevant and
• the mid-walls with the 3 80mm fans are identical and
• all cases use a blocking backplane in front of the hard drives,

then I'd assume it is the 2 extra rear fans that make this noticeable difference.

Yeah, after taking another look at the 847 design I can probably answer my own question. Those conditions do not hold for the 847. It's actually very different from the 8x6 line and does not have a 3 fan mid-wall + 2 fan rear exhaust design. It seems to have a double stacked 4 + 3 mid-wall and no exhaust fans at all. It's probably noticeable, because it has 7 7000RPM fans - plain and simple.
Well, I'd still be interested to see what difference the rear exhausts of the 836/846 make in comparison to the 826.

 

[tvsjr]

Correct. Other than the CPU fans, if you choose to add them, the 847s just have 7 fans mid-span. They need to be bigger, faster, and higher static pressure than the 827s (which are stacked 40mm high-depth fans) because you've got potentially a bit more than 4X the heat load (36 drives total, versus 8 in the 2U variant). There is a shroud with some plastic to "shape" the air as it comes out of roughly three fans and push it across the CPUs... I'm sure this works, but I wanted to give it a bit more help with the addition of the active CPU heatsinks.

I definitely wouldn't want an 847 in the living room, at least not by itself. You could always consider something like a Quiet Rack, AcoustiQuiet, etc. to house the systems if they're going to be in a living space... fair warning, these types of cabinets are Not Cheap.

 

[depasseg]

@depasseg: According to the parts list the 5028R is already using a 920W SQ PSU. Unfortunately tho I could not find the videos you posted. I have to take another look.

The PSU I was testing was the one in the 847, which is not the SQ (it's the PWS-1K41P-1R, i believe).

And you are correct, all the fans are stacked in the middle (drives in the rear of the chassis), and they are 7k screamers. I tried putting some Noctua's in there (someone else mentioned it), but I couldn't keep the drives cool enough (with only the front drive bays filled). PWM has made a HUGE difference though.

I think this is the post I was thinking of:
https://forums.freenas.org/index.ph...ni-itx-cases-quieter.25635/page-2#post-165599

 

[bestboy]

I have another related question: When you guys are connecting the chassis fans to the mainboard in order to PWM-control them, which of the 4 fan modes are you using with the SM mainboard? Usually there are 4 options with different fan thresholds to choose from via IPMI:

• standard speed
• optimal speed
  
• heavy IO speed
  
• full speed

Given that there are 7000 RPM fans in the chassis, I wonder how many RPM they are doing when throttled down by the different PWM profiles.

Code:

fan mode            speed when idle [RPM]
=========================================
full speed          7000
heavy IO speed      ?
standard speed      ?
optimal speed       ?

 

[jgreco]

I believe if you reverse lines 4 and 5, that's the correct order in order of aggressiveness.

 

[bestboy]

Ok, fixed. Switched the "heavy IO speed" and "standard speed" rows. Now, we just need some RPM values for the table :)

 

[SweetAndLow]

I have another related question: When you guys are connecting the chassis fans to the mainboard in order to PWM-control them, which of the 4 fan modes are you using with the SM mainboard? Usually there are 4 options with different fan thresholds to choose from via IPMI:

• standard speed
• optimal speed
  
• heavy IO speed
  
• full speed

Given that there are 7000 RPM fans in the chassis, I wonder how many RPM they are doing when throttled down by the different PWM profiles.

Code:

fan mode            speed when idle [RPM]
=========================================
full speed          7000
heavy IO speed      ?
standard speed      ?
optimal speed       ?

I think standard is 50% and optimal is 30%

 

[jgreco]

Ok, fixed. Switched the "heavy IO speed" and "standard speed" rows. Now, we just need some RPM values for the table :)

The RPM is going to vary based on the chassis and fans.

 

[tvsjr]

Ok, fixed. Switched the "heavy IO speed" and "standard speed" rows. Now, we just need some RPM values for the table :)

There aren't RPM values... there are target PWM duty cycles. Even the standard speed will run up to 7,000RPM if the system gets warm enough.

The details are in the manual. Full = 100%, standard = 50%, optimal = 30%. I don't know about heavy IO. Anything that isn't 100% duty cycle is temperature controlled.

 

[bestboy]

I think standard is 50% and optimal is 30%

Oh, I did not realize that the fan modes are just based off percentages. But this post I found via google seems to confirm that. Also I tried to evaluate it in my system with the CPU fan and it seems to be right.

2400 rpm @ full speed = 100%
1200 rpm @ standard speed = 50%
1000 rpm @ optimized speed = well, could be 30% with an enforced minimum
1000 rpm @ heavy IO speed = dunno. this one seems to be buggy and does nothing in my system. Considering the given pattern I'd have expected it at 80%... *shrug*

So inserting the 50% resp. 30% rates, the table would look like that:

Code:

fan mode            speed when idle [RPM]       rate [%]
===========================================================
full speed          7000                        100
heavy IO speed      ?                           ?
standard speed      3500                        50
optimal speed       2100                        30

 

[bestboy]

The RPM is going to vary based on the chassis and fans.

Well, yes I agree. I chose those 7k RPM fans, because the chassis I'm interested in (826BE16-R920LPB) is using the same fans as the 847 chassis. It's the FAN-0126L4 in the SuperMicro system fan matrix.

 

[bestboy]

The details are in the manual. Full = 100%, standard = 50%, optimal = 30%. I don't know about heavy IO. Anything that isn't 100% duty cycle is temperature controlled.

Yep. I did find a description of it in the BIOS section of X9SCM mainboard manual.

Fan Speed Control Mode
This feature allows the user to decide how the system controls the speeds of the
onboard fans. The CPU temperature and the fan speed are correlative. When the
CPU on-die temperature increases, the fan speed will also increase for effective
system cooling. Select "Full Speed" to allow the onboard fans to run at full speed
(of 100% Pulse Width Modulation Duty Cycle) for maximum cooling. This setting
is recommended for special system configuration or debugging. Select "Standard"
for the onboard fans to run at 50% of the Initial PWM Cycle in order to balance the
needs between system cooling and power saving. This setting is recommended
for regular systems with normal hardware configurations. Select "Optimal" for the
onboard fans to run at 30% of the Initial PWM Cycle for best power efficiency and
maximum quietness. The options are Full Speed (@100% of PWM Cycle), Standard
(@50% of PWM Cycle), and Optimal (@30% of PWM Cycle).

The mainboard manual of my X10SLM+-F did not contain such a description, tho. :/
Also the X9SCM manual does not explain the "heavy IO" mode which was introduced later on with the X10s AFAIK.

There aren't RPM values... there are target PWM duty cycles.

Is it wrong to assume that 50% initial PWM duty cycle is equal to 50% RPM? It seems to match up in my quick'n'dirty test and from looking it up quickly in Wikipedia I also don't see a contradiction. If a fan motor would get 50% of its input signal, it could run at 50% power, resulting in 50% of the RPMs...

 

[Ericloewe]

Is it wrong to assume that 50% initial PWM duty cycle is equal to 50% RPM?

Yes, but generally not by much. Each fan has an omega(PWM) curve. Lemme grab an example: https://www.sanyodenki.com/archive/document/product/cooling/catalog_pdf/San_Ace_120GV25_E.pdf

If you scroll down, you'll see that each variation of this family has its own (similar) curve. Different families will have different curves. Some fans will regulate down to zero, for instance (though it's rare).

 

[bestboy]

Great! Thanks.
That explains the RPM barrier I hit with my CPU fan test on "Optimal speed". At 30% initial PWM cycle the Intel stock fan should have been at ~720 RPM, but it did not go below 1000 (which is about 42%).