KYThrill's Norco 4220 Build


Recommended Posts

This is my third unRAID build.  I wish I had just went all out and built big the first time.  It would have saved me lots of time and money.  Anyone, so far operation has been sweet and trouble free.

 

OS at time of building: unRAID 4.7 Pro

CPU: AMD Sempron 145 (The second core unlocked for me, so with core unlocking it is actually an AMD Athlon II X2 4450e)

Motherboard: MSI NF750-G55 w/ Koutech USB Header Adapter

RAM: CORSAIR XMS 4GB (2 x 2GB) 240-Pin DDR3 SDRAM DDR3 1333 Ultra Stable Desktop Memory Model

Case: Norco 4220 w/ 120 mm fan plate and rails

Drive Cage(s): Included w/ Norco 4220

Power Supply: Seasonic X750 80 Plus Gold

SATA Expansion Card(s): 2X Supermicro AOC-SASLP-MV8 PCI Express x4 SAS Card

Cables: 1x Norco Reverse Breakout, 2x E-Bay's Finest SAS cable, 1X 3-Wire Fan extension cable

Fans: 2x Xigmatek XOF-F1251 120mm, 1x Scythe Slipstream SY1225SL12M 120 mm, 2x Enermax UC-8EB 80 mm

 

Build Cost (Excluding HDD's + unRAID license): $890 (shipping included)

 

Parity Drive: Hitachi 2TB 7K2000

Data Drives: 3x Western Digital 2TB WD20EARS, 1x WD 2TB EADS, 1x Western Digital 1TB WD10EAVS, 4x Samsung 2TB F4EG, 7x Hitachi 2TB 5K3000,  1x Samsung 1.5TB F2, 2x Seagate 2TB

Cache Drive: Samsung Spinpoint MP4 HM250HJ 250GB 7200 RPM 2.5" SATA

Backup Drives: 1x Western Digital 2TB Black, 1x Western Digital WD20EARS

Total Drive Capacity: 36.5 TB

 

Primary Use: Media Server

Likes: Fairly quiet.  Very low power consumption. Dim lights on drive trays (not too distracting in a home theater room).  Love the motherboard.

Dislikes: Drive trays aren't quite as high quality as my previous Icydock trays.  Norco fan plate is an inch or so too close to the backplanes.  4220 case is too long (could be 3-4" shorter). Soft power/reset buttons.

Add Ons Used: unMenu, apcupsd, bandwidth monitor NG, PCI utils, powerdown, Info Zip, Preclear, more to come

Future Plans: I would also like to see if I can get Lime to transfer my license to a Mushkin Mulholland flash drive from the cheapie Kingston I'm using now.  The current Kingston has write speeds at about 1.5 MB/s and the Mulholland easily does 12 MB/s.  The Mulholland has fast read speeds too, so I'm hoping to shave a few seconds off the boot/power down times.

 

I will also be adding a second SATA controller to the 2nd PCI-e slot once it is needed for more drives.  I haven't decided if that should be an Adaptec adapter or another Supermicro.  I know the GET_IDENTITY error on the Supermicro is harmless, but I just don't like seeing it in my syslog.  I still have a hodge podge of 1TB drives laying around, so some future drive expansions will use these drives.  Then once they are in use and my spare drives dry up, I'll start adding/replacing with 2TB green drives.  Eventually the array will have one 7200 rpm drive for parity, and all data drives will be green drives.

 

Boot (peak): 135W (9 drive config)

Idle (avg): 55W (9 drive config)

Active (avg): 99W (9 drive config)

Light use (avg): 89W (9 drive config)

 

4DFLW.jpg

 

ZBLY2.jpg

 

This was actually a very trouble free build.  The only problem I ran into was with the 80mm Enermax Tb Silence fans.  Their casing is a little larger than most 80 mm fans, so two of them wouldn't fit side-by-side in the back of the case.  I don't know that it matters though.  My motherboard reports a system temp of 36C, and processor temp of 50C (47C if I just run one core).  I don't know if a second fan would help much.  I may try two Nexus silent fans there someday though.

 

NOTE: It turns out that the onboard NIC does support Jumbo Frames.  However, the v.62 forcedeth driver built into the Linux kernel does not support changing the MTU through ifconfig.  There is a newer version of forcedeth, but it is built into a newer Linux kernel (so maybe unRAID 5 will have it).  A port of the v.64 forcedeth driver to OS X does now allow you to manually set the MTU, whereas the port of .62 did not.  So I'm assuming the newer .64 driver would now allow you to enable jumbo frames through ifconfig.  Probably not important to most, but could be important to some.

 

NOTE: If you use the BIOS defaults with the RAM used in this build, for some reason Memtest reports that the MB is selecting 8-8-8-20 timing.  The RAM is only 8-8-8-24.  The RAM SPD reports a 9-9-9-24 setting.  With 8-8-8-20 and 9-9-9-24 memory settings, the RAM will fail Test #5 every time in my setup.  Oddly enough though, if you set it o 8-8-8-24 manually, it will pass (it passed three full tests).  I can understand errors with 8-8-8-20 because this is faster than the RAM rating.  9-9-9-24 is slower, so a little odd that it fails when set by the RAM's own SPD.  Just something to keep in mind.

 

NOTE: One stick of RAM failed a few months into this build.  So bad RAM could possibly had a role to play in some of the observations in the previous note.


 

 

Link to comment

I will also be adding a second SATA controller to the 2nd PCI-e slot once it is needed for more drives.  I haven't decided if that should be an Adaptec adapter or another Supermicro.  I know the GET_IDENTITY error on the Supermicro is harmless, but I just don't like seeing it in my syslog.  I still have a hodge podge of 1TB drives laying around, so some future drive expansions will use these drives.  Then once they are in use and my spare drives dry up, I'll start adding/replacing with 2TB green drives.  Eventually the array will have one 7200 rpm drive for parity, and all data drives will be green drives.

 

The Adaptec has half the ports of the Supermicro, but costs about the same.  Doesn't seem like a very tough decision to me.  Like you said, the HDIO_GET_IDENTITY errors are harmless.  I certainly would never spend more money just to clean up my syslog.

Link to comment

Not the 1430SA.  I can't find it now (of course), but I had found another unRAID build on another forum that had used some surplus SAS/SATA cards that were from old server pulls that were working for him in unRAID (they had an Adaptec chipset).  I think they were based off the 2805.  However, there were two part #'s that looked identical but one worked with unRAID and the other didn't.  They were PCI-e x8 (which is fine because the MB has two x8 slots).  Anyway, at the time some were for sale on E-Bay for $129.

 

I'll have to look around more for that post again (i saw it last fall when I bought my Supermicro).

 

http://lime-technology.com/forum/index.php?topic=10979.0

 

Or maybe I won't.  Here is a post from someone using a 2805 and it has problems too (no drive temps) and looses more features (spin up/spin down control) in the 5 beta.  I guess I don't want a solution that is less supported in future unRAID versions.  But the other post I found (AVS or some other place) swore up and down that their OEM card based on an Adaptec chipset worked perfectly in unRAID.

 

It will require more research.  But if I find a solution like that, which worked, and it was $129 vs $92, I might pay a little more for a card that was fully supported by unRAID and didn't generate extra errors.

Link to comment

It looks like the Enermax fans may not be cutting it.  All my drives are running 3C higher than they did in Icy Dock trays.  Unfortunately, the website where I bought the Enermax's had a typo on the CFM, which led me to buy them.

 

The four stock 80mm fans in the Norco 4220 are rated for 35 cfm each, so the four fans together are 140 cfm.  So three 120 mm fans should each move around 47 cfm.  I thought I was exceeding this, but it turns out the Enermax fans are only rated for 42 cfm.  So I'm actually shy about 15 cfm overall, plus only one rear fan because of fit.

 

So it looks like I may need to replace at least one of the fans.  I thought if I could find a 60+ cfm fan and put it in the middle, it may be enough to compensate and only swap out one fan.

 

I'll have to play around a bit.

 

UPDATE: I ordered a Scythe Medium Speed Slipstream.  Based on SilentPC's testing, it should only moderately add to noise (replacing a 16 dBA fan with a 21 dBA fan), but will up the CFM on that fan to 66.  I'll install it in the middle position, and the 120 mm fan plate should pull 150 CFM and exceed the 140 CFM of the stock fan plate.  It will be a net gain of 24 CFM over my current fan plate.

Link to comment

Well, I installed once Schythe fan in the center of my 120mm fan plate.  Temps did drop by 2C on my drives, but it still wasn't the gain I had hoped for.  The noise level on my case also went from about 20 dBA to about 27 dBA.  With all Enermax fans, I couldn't even tell the Norco was powered up.  It was dead silent.  I can clearly hear it drawing air now.  It's about as loud as my quad core desktop (also setup to be as silent as possible.  It is still a vast improvement over the stock fan setup, just not dead silent.

 

So now I can either go back to dead silent, but run my drives at 40C (probably hotter as I add more drives), or add more Scythes.  Each Enermax I swap for a Schythe will only add 1 dBA to the noise level.  So I can really up the cfm and cooling for minimal noise gain at this point.  Decisions...

 

I'm actually on the lookout for the new Cougar Vortex fan, CF-V12H.  I can't find anyplace that sells them though  (looks like they aren't slated to be released until the end of June).  They are about 60 cfm and 17.7 dBA.  They also create a vortex, which has show to cool better than other higher cfm fans that move air straight.  I could put three Cougars in and only be at about 22 dBA, which may take me back to dead silent status.  Fan plate airflow would be 180 cfm, vs 150 cfm in my current setup.  So I could increase my airflow by 20% and reduce my apparent noise level by around 25%.  Now if I could just find the stupid things without needing to import them from Japan.

Link to comment

Another option you might want to consider are the Noctua NF-P12 120mm Fans. They flow at 54CFM, and have a very nice noise signature (19.8dB). Even if you used them with the included fan silencers you'll get very good flow. 46 plus CFM with the LNA for example. Noctua calls the fan silencers LNA (16.9dB) and ULNA (12.8dB) adapters. A bit pricier, but the Noctuas work silently.

Link to comment

Another option you might want to consider are the Noctua NF-P12 120mm Fans. They flow at 54CFM, and have a very nice noise signature (19.8dB). Even if you used them with the included fan silencers you'll get very good flow. 46 plus CFM with the LNA for example. Noctua calls the fan silencers LNA (16.9dB) and ULNA (12.8dB) adapters. A bit pricier, but the Noctuas work silently.

 

I had already looked at the Noctua.  And while they are comparatively quiet, Noctua was one of the brands that misrepresents their sound levels.  For example, the NF-P12 running at the full 12V, actually produces 25 dBA.  With their "silencers" (just adapters that make them run at lower voltages at speeds), they ran 20 and 16 dBA respectively.  Their actual performance is basically identical to the Slipstreams (both are 9 blade designs running at similar or identical rpms).

 

I would have to run the Noctua's at the full 12V to achieve the min of 150 cfm cooling needed.  So my sound levels really wouldn't differ.  The pitch would probably change, and it could be more or less pleasing.

 

I have another fan that should be here today or tomorrow.  It is a 50 cfm fan that claims 18 dBA and only spins 900-1000 rpm.  About the max airflow you can normally get out of a 1000 rpm is 38-40 cfm.  It also had a high static pressure rating (which is good)  The novel thing about it is that it is supposed to get its performance because of a) a bullet shaped fan cone, special blade design (everyone has a special blade design though), and special housing design.  For $12, I had to give it a spin and see if it is just more false marketing hype or something worthwhile.  50 cfm was their highest flow though, so there would be no way to get more than 150 cfm out of my fan plate, if I ever needed more.

 

But my gut tells me this will end up being more marketing lies.  It will either be louder, spin faster than 1000 rpm, or produce much less than 50 cfm of airflow.  It is a 9 blade fan, and any other 9 blade fan (Noctua, Scythe, etc.) would move less than 40 cfm at speeds of 1000 rpm or less.  Can a bullet shaped cone and special housing design really improve airflow 25% or more?  Call me skeptical...

Link to comment

Another option you might want to consider are the Noctua NF-P12 120mm Fans. They flow at 54CFM, and have a very nice noise signature (19.8dB). Even if you used them with the included fan silencers you'll get very good flow. 46 plus CFM with the LNA for example. Noctua calls the fan silencers LNA (16.9dB) and ULNA (12.8dB) adapters. A bit pricier, but the Noctuas work silently.

 

I had already looked at the Noctua.  And while they are comparatively quiet, Noctua was one of the brands that misrepresents their sound levels.  For example, the NF-P12 running at the full 12V, actually produces 25 dBA.  With their "silencers" (just adapters that make them run at lower voltages at speeds), they ran 20 and 16 dBA respectively.  Their actual performance is basically identical to the Slipstreams (both are 9 blade designs running at similar or identical rpms).

 

I would have to run the Noctua's at the full 12V to achieve the min of 150 cfm cooling needed.  So my sound levels really wouldn't differ.  The pitch would probably change, and it could be more or less pleasing.

 

I have another fan that should be here today or tomorrow.  It is a 50 cfm fan that claims 18 dBA and only spins 900-1000 rpm.  About the max airflow you can normally get out of a 1000 rpm is 38-40 cfm.  It also had a high static pressure rating (which is good)  The novel thing about it is that it is supposed to get its performance because of a) a bullet shaped fan cone, special blade design (everyone has a special blade design though), and special housing design.  For $12, I had to give it a spin and see if it is just more false marketing hype or something worthwhile.  50 cfm was their highest flow though, so there would be no way to get more than 150 cfm out of my fan plate, if I ever needed more.

 

But my gut tells me this will end up being more marketing lies.  It will either be louder, spin faster than 1000 rpm, or produce much less than 50 cfm of airflow.  It is a 9 blade fan, and any other 9 blade fan (Noctua, Scythe, etc.) would move less than 40 cfm at speeds of 1000 rpm or less.  Can a bullet shaped cone and special housing design really improve airflow 25% or more?  Call me skeptical...

 

Misrepresents their sound levels?

 

All I can say, for my needs, best 120mm silent fan out there is the Nexus real Silent, but you are looking for a lot more airflow than these produce. My experience with the Noctua fans has been positive, and I'd agree with what SPCR reports about these-that they become inaudible at the higher voltage, but to each his own. They're a bit pricey for just a test, though.

 

Best of luck with your fan selection. I hope the $12 ones work as you're hoping for, or you find your ideal setup.

 

Link to comment

Another option you might want to consider are the Noctua NF-P12 120mm Fans. They flow at 54CFM, and have a very nice noise signature (19.8dB). Even if you used them with the included fan silencers you'll get very good flow. 46 plus CFM with the LNA for example. Noctua calls the fan silencers LNA (16.9dB) and ULNA (12.8dB) adapters. A bit pricier, but the Noctuas work silently.

 

I had already looked at the Noctua.  And while they are comparatively quiet, Noctua was one of the brands that misrepresents their sound levels.  For example, the NF-P12 running at the full 12V, actually produces 25 dBA.  With their "silencers" (just adapters that make them run at lower voltages at speeds), they ran 20 and 16 dBA respectively.  Their actual performance is basically identical to the Slipstreams (both are 9 blade designs running at similar or identical rpms).

 

I would have to run the Noctua's at the full 12V to achieve the min of 150 cfm cooling needed.  So my sound levels really wouldn't differ.  The pitch would probably change, and it could be more or less pleasing.

 

I have another fan that should be here today or tomorrow.  It is a 50 cfm fan that claims 18 dBA and only spins 900-1000 rpm.  About the max airflow you can normally get out of a 1000 rpm is 38-40 cfm.  It also had a high static pressure rating (which is good)  The novel thing about it is that it is supposed to get its performance because of a) a bullet shaped fan cone, special blade design (everyone has a special blade design though), and special housing design.  For $12, I had to give it a spin and see if it is just more false marketing hype or something worthwhile.  50 cfm was their highest flow though, so there would be no way to get more than 150 cfm out of my fan plate, if I ever needed more.

 

But my gut tells me this will end up being more marketing lies.  It will either be louder, spin faster than 1000 rpm, or produce much less than 50 cfm of airflow.  It is a 9 blade fan, and any other 9 blade fan (Noctua, Scythe, etc.) would move less than 40 cfm at speeds of 1000 rpm or less.  Can a bullet shaped cone and special housing design really improve airflow 25% or more?  Call me skeptical...

 

Misrepresents their sound levels?

 

All I can say, for my needs, best 120mm silent fan out there is the Nexus real Silent, but you are looking for a lot more airflow than these produce. My experience with the Noctua fans has been positive, and I'd agree with what SPCR reports about these-that they become inaudible at the higher voltage, but to each his own. They're a bit pricey for just a test, though.

 

Best of luck with your fan selection. I hope the $12 ones work as you're hoping for, or you find your ideal setup.

 

 

Yes.  Many fan manufacturers put false cfm and spl data on their fans in order to market them appear to be a better fan that what they are.  In some cases the numbers are just false and made up.  In other instances, a manufacture will monkey with their testing rig to produce better results, even though those results can never be repeated in the real world.  I'm a Test Engineer and favorable test biasing is unfortunately par for the course in any industry.  I typically design an unbiased test that represents real world conditions.  As soon as a product doesn't perform favorably in that setup, marketing will demand that the setup be modified to produce favorable results, which can usually be done using a series of erroneous assumptions.  Sad but true.

 

Here is an example of an actual Noctura fan test.

 

http://www.silentpcreview.com/article1040-page3.html

 

As you can see, their test of a NF-S12-1200 produces a 12VDC sound level of 21 dBA.  The manufacturers published rating on that fan at that voltage was 17 dBA.  Like wise, all the slower speeds were in error too.  At five volts, the publisher spec says <6 dBA.  The test results show that at 5.1 VDC, the sound level was 11 dBA.  Noctura simply provides false numbers on their fans.

 

But if you look at the Nexus fan from that test, they publish 18 dBA at 1000 rpm and than fan tested 16 dBA, better than the rating.  Or the Scythe from that test publishes 24 dBA at 1200 rpm, and the fan tested was 21 dBA at 1400 rpm.  So again, actual results were better than the results provided on the datasheet.

 

Also, some manufacturers provide datasheets that have values that they feel they can guarantee.  So they may test 20 fans, and publish the values of the worst one.  So you will normally get better performance than what the datasheet suggests, but at worst, you will get what is on the datasheet.

 

Other manufacturers will test 20 fans, and publish a datasheet with the values from the best performing fan out of the 20.  Then you will normally get a fan that performs worse than the datasheet suggests, but its possible you may get lucky and get one that will do what the datasheet says.

 

There are lots of ways to manipulate data, and from everything I've seen personally and read elsewhere, Noctura is one of the companies that manipulates data to make their products look more favorable.

 

Link to comment

I'm still trying to solve the mystery of cooling the drives in this box.

 

The factory 80 mm fan plate moves 141 cfm.  I have replaced it with a 120mm fan plate that can move 150 cfm and was still getting drive temps hotter than the stock setup.  I then upped the fan plate cfm to 160 and was still getting temps 2C hotter than the stock setup.

 

So what this tells me is that the two rear fans must account for some of the air movement that cools the drives.  The stock mid plate is 141 cfm and the two stock rear fans are 160 cfm, more than the mid plate.

 

In my case, I only have 24 cfm at the back of the case, because my cpu is low wattage and produces little heat.  But I'm thinking that to cool the HDD's in the front, I will need to up my rear cfm.  So I think I am going back to a 150 cfm mid plate configuration and then I'll take a look at what I can do at the back, without significantly increasing the noise.  I think I can get at least 60 cfm at the rear without adding any significant noise, but I don't think there is an easy way to get 160 cfm quietly.

Link to comment

Are you also checking static pressure?

 

think of it like torque to have the power to move the air in the tight space between the drives.

 

It's almost impossible to determine static pressure without actually instrumenting your server chasis.  CFM is basically addative, so three 50 cfm fans move ~150 cfm airflow (maybe a little less if they create turbulence that interferes with other fans).  But if you have three fans with a static pressure rating of 2.2, your resulting static pressure is not going to be 6.6.  It isn't linear or scalable.  When measuring it, you can highly bias your results based on your test rig.

 

Here is a good article on why static pressure can't really be used to select a fan.  It is geared towards HVAC, but applies here as well.

 

http://www.esmagazine.com/Articles/Feature_Article/87328157de43a010VgnVCM100000f932a8c0____

 

Static pressure only paints part of the picture, and it is possible to have a fan with high static pressure that performs more poorly than a fan with a lower static pressure.  Their really is not enough data given with a PC fan to determine its performance based on pressure.

Link to comment

There is an old axiom with cars, fast, cheap, reliable any two from three is easy.

 

Same thing with PC

fast,quiet,cheap any two from three is easy.

 

First thing you need to do is fix your exhaust fan situation. You have four 120mm fans (three hdd pulling fans and PSU) blowing 140-200cfm into the MB cavity and a single 80mm exhaust fan (20cfm if we are being generous). This is going to give you a high pressure in the MB chamber which will result in poor flow rate through the chamber.

 

Trying removing the PSU, reconnect it out the back of the case and removing the unused PCI slot protectors. If this drops the temps as expected then you need better exhausts. Simples.

 

If it doesnt then you arent drawing enough air over the HDDs. First make sure you have spread your HDDs so they have decent air flow around them. Buy some decent fans, you've spent major bucks on the system. Don't cut costs on fans especially if you are expecting a quiet system. Artic, Nexus, Sharkoon, Noctua or Zalman all do excellent 80cm and 120cm quiet fans than can shift decent amounts of air at low noise levels.   

 

Lastly decide what you want and work with that. You have brought a server case, designed for data centres, not designed to be quiet. As witnesssed by the stock fans (I know I've heard them, I laughed and I'm used to data centre noisy but those things are ridiculous!). Sure they flow 140cfm but at 80db.

 

You are using 7200rpm drives, potentially not good for heat and noise (not easy to change).

Do you spin them down when not in use?

What do they sit at ambient?

You've turned on extra core on CPU, not needed turn it off.

Do you need a cache drive? Remove it. Less heat, less noise.

Look at putting a vertical CPU cooler with 120MM fan blowing toward exhaust fans (CPU should be low 40's).

Zalman or Artic Cooling again all do suitable designs.

Look at fitting the case with some accoustic absorbtion materials once you have the heat under control.

Link to comment

I broke down yesterday and ordered a 120mm fan plate and 3 Noctua NF-P12-1300's

 

more than i had originally wanted to spend, but it seems like the logical move. Most people are swearing by these for the norco's and i am a lemming and i will jump on the bandwagon.

 

the funny part is, this time of year i have a box fan by the server that drowns out the norco stock fans anyways.

 

i had considered buying a 4th and mounting it in the area that my unused PCI brackets are.i was thinking that might help with pressure/airflow and help keep my MV8's a tad cooler. i noticed they are getting a bit warm.

 

Link to comment

Hi Johnm,

 

Glad to see you decided to give the Noctuas a try. Hope they don't disappoint. I bought a pair of Noctua 80mm fans some time ago for my Zalman HTPC, and haven't looked back. All temps stabilized, and I usually run those Noctuas at 50% in winter and full in Summer without Aircon. All I can hear from the media pc is the Scythe fan facing my hard drives (although it's usually at 7 volts).  My  gamer rig (was based on SPCRs 22dBA design) runs only Nexus via fan controllers. All my systems have soundproofing material (Accoustipak or similar).

 

For my Supermicro SAS card, I put an Xfan silent pci fan in the next empty pci slot, and card temps went down dramatically (sensor I slapped on it went from 50 plus degrees to mid 30s). Maybe a similar solution works for you.

 

Only a full bag of tricks can get any rig close to being silent.

 

I designed my server's cooling solution for the lower hard drive cage based on the limitations of my Fractal XL case. Simple, cheap, and effective (those drives went from mid-40s to lower 30s), but that won't work in a Norco with your drive bays.  

 

For these larger systems, IMHO, best bet remains testing overall flow and direction of airflow, improve exhaust, increase overall flow through the case, reduce static pressure around MB chamber, and control air leaks and direction of airflow.  

 

Link to comment

I wouldnt normally recommend those PCI/VGA slot coolers but I had good results with an Antec one and a SAS card. Dont last more than 6 months though before the bearings wear out (really noisy). On plus side they are really cheap so disposable.  

 

 

Link to comment

Totally agree that most pci coolers are too noisy, and usually not worth the effort.

 

This XFan little version that I opted for, was so cheap and spec'd as virtually silent, that I thought worth testing. It has delivered up to now.  Another option that comes to mind, since this is heat that concentrates around a small heatsink, would be to try an Antec spot fan. Although I've immediately replaced every Antec fan I've owned, I've heard good things about their spot fan.

Link to comment

KyThrill, I'm not sure if you're interested in trying the Noctua's or not, but they are sale for $16.99 ATM each at directron. (or their everyday price is $10 less then everyone else)

 

They only charged shipping for one ($5.50). That's a far cry cheaper then newegg.

 

 

As far as the PCI cooler. i had the top off my server today and and just placed a 120 silent fan sitting on top of the cards blowing down onto them. .. it's a 80lb rack mounted server. I'm not going to move it.. LOL.

 

I think it is ok like for now.

 

I am starting to think about that guy that wanted to mod his norco by cutting fan holes into it.. I am starting to see merit in that. one in the top for pci card cooling. a 120mm in the side opposite the power supply for better pressure release  in motherboard bay created by midfans. that might assist in HDD cooling.

Link to comment

KyThrill, I'm not sure if you're interested in trying the Noctua's or not, but they are sale for $16.99 ATM each at directron. (or their everyday price is $10 less then everyone else)

 

They only charged shipping for one ($5.50). That's a far cry cheaper then newegg.

 

 

As far as the PCI cooler. i had the top off my server today and and just placed a 120 silent fan sitting on top of the cards blowing down onto them. .. it's a 80lb rack mounted server. I'm not going to move it.. LOL.

 

I think it is ok like for now.

 

I am starting to think about that guy that wanted to mod his norco by cutting fan holes into it.. I am starting to see merit in that. one in the top for pci card cooling. a 120mm in the side opposite the power supply for better pressure release  in motherboard bay created by midfans. that might assist in HDD cooling.

 

Excellent solution of the 120 facing the SAS cards. A 120 should definitely prove more durable and directed than a pci slot fan. Maybe test with a lid in place.

I would concur that given that these cases are built/designed for a server room, strategically cutting a new hole for a 120mm fan to improve flow and exhaust in general, could work. I closed the factory top fan slot on my Antec P182, but flow was improved elsewhere, gaining the overall temp control, and a virtually silent rig in the process.

Given the expected environment a server case should normally live in, testing a few of these solutions with mock up lids/flow direction vanes, etc. might be worth trying.

Link to comment

KyThrill,

 

After Johnm's comments I had another look at your build specs. Although you've mounted the PSU with the fan facing the MB to help exhaust air from the case, Seasonic PSU fans are controlled by PSU internal temperature. These PSU fan do not kick in until the PSU temperature rises. So the PSU should not be considered as helping active case exhaust in this system. You actually have a single 80mm fan as the active hot air exhaust with a set of fans pumping in 150CFM. Case exhaust needs to be improved.

Link to comment

There is an old axiom with cars, fast, cheap, reliable any two from three is easy.

 

Same thing with PC

fast,quiet,cheap any two from three is easy.

 

First thing you need to do is fix your exhaust fan situation. You have four 120mm fans (three hdd pulling fans and PSU) blowing 140-200cfm into the MB cavity and a single 80mm exhaust fan (20cfm if we are being generous). This is going to give you a high pressure in the MB chamber which will result in poor flow rate through the chamber.

 

Trying removing the PSU, reconnect it out the back of the case and removing the unused PCI slot protectors. If this drops the temps as expected then you need better exhausts. Simples.

 

If it doesnt then you arent drawing enough air over the HDDs. First make sure you have spread your HDDs so they have decent air flow around them. Buy some decent fans, you've spent major bucks on the system. Don't cut costs on fans especially if you are expecting a quiet system. Artic, Nexus, Sharkoon, Noctua or Zalman all do excellent 80cm and 120cm quiet fans than can shift decent amounts of air at low noise levels.    

 

Lastly decide what you want and work with that. You have brought a server case, designed for data centres, not designed to be quiet. As witnesssed by the stock fans (I know I've heard them, I laughed and I'm used to data centre noisy but those things are ridiculous!). Sure they flow 140cfm but at 80db.

 

You are using 7200rpm drives, potentially not good for heat and noise (not easy to change).

Do you spin them down when not in use?

What do they sit at ambient?

You've turned on extra core on CPU, not needed turn it off.

Do you need a cache drive? Remove it. Less heat, less noise.

Look at putting a vertical CPU cooler with 120MM fan blowing toward exhaust fans (CPU should be low 40's).

Zalman or Artic Cooling again all do suitable designs.

Look at fitting the case with some accoustic absorbtion materials once you have the heat under control.

 

I thought I had mentioned it, but I guess I didn't.  I did disable the second core.  I was only utilizing the first core 20-30% most of the time, so the second core was unneeded at this point.  I don't have any addons installed yet that even support the second core.  It dropped the CPU temp 2-3C.  CPU temp with a single core already maxes at 47C, so I think the stock cooler is doing an adequate job.

 

Drives spin down after 30 minutes.  Usually run 26-28C in a 25C ambient when not in use.  

 

The basic theory is this.  The 120mm fans create a low pressure in front of the mid plate (the HDD cavity).  Pressure outside the box is higher, so air flows through the front, over the drives (cooling them), and into the HDD cavity.

 

Likewise, the midplate fans create high pressure in the  MB cavity.  The outside pressure is lower, so air moves out through the rear exhaust.  Adding fans at the rear of the case increases the pressure differential across the MB cavity causing more air to move across the MB cavity, improving cooling on the PSU, addon cards, CPU, etc.  Personally I think everything in the MB cavity is being cooled adequately, so airflow through that cavity isn't a concern.

 

However, what I believe is happening with stock fans is that the pressure at the rear of the case is actually lower than the pressure at the midplate.  Therefore, in the stock setup, the rear fans are actually responsible for moving some additional air across the HDD's.  So potentially, no amount of fans at the rear of the case will help with cooling the HDD's until you install powerful enough fans to lower the pressure at the rear of the case below the low pressure side (HDD side) of the midplate.  The only way the back fans will help is if the pressure in the MB cavity is high enough that it restricts the airflow through the midplate, causing the pressure to rise in the HDD cavity.  Extra fans at the rear will lower the pressure in the MB cavity, so if you are choked, cooling would improve (and cooling of the MB cavity components would improve).

 

Enermax claims the fan moves 24 cfm and independent testing has shown real world results of 22-25 cfm.  So close enough.  I have two Nexus 80 mm silent fans that are more noisy than the Enermax fans and move 20 cfm.  As soon as time allows, they will go in, at least temporarily.  This will increase rear fan airflow from 24 cfm to 40 cfm.  Pressure at the rear of the case should drop just slightly,  If the midplate airflow was choked, this should open it up a little and cool the drives at least a little more.  If I do see a change, I can start searching for the right pair of 80mm fans to balance noise and air flow.  If there is no improvement at all, it points back to rear fans not helping with HDD cooling until they can move enough air to create a lower pressure at the exhaust than the low pressure at the midplate.

 

In my current configuration (two Xigmateks and one Scythe, with a single 80 mm Enermax, my 7200 rpm drives run 37-39C when heavily accessed.  My 5400 rpm green drives are typically running 33-36F under heavy access.  I can almost live with these temps.  And these are temps after adding a tenth drive so the server is over half full.  I'm still going to try swapping in a third Xigmatech and replacing the Scythe.  The Xigmatech fans are much quieter (but lower airflow).  If the lack of exhaust is choking the midplate airflow, then reducing the airflow of the fan plate will make no difference in cooling at this point.

Link to comment
  • 1 month later...

Well, one of the sticks or Corsair RAM went belly-up, so I have to RMA.  Currently running just fine on 2GB of RAM.

 

While I had it open, I also removed the single 80mm Enermax TB Silence, and replaced it with two standard Enermax 80 mm fans.  To be honest, the new ones are cheaper than the TB Silence, they fit correctly, and I can't tell a bit of difference in sound or airflow.  The fan leads are shorter and I did have to use an extension on one of them to reach the fan header on the MB.

Link to comment

Thanks for the write-up on your build. I just finished a build with a new Norco 4220. I installed 1 Discrete SATA to SFF-8087 Mini SAS Reverse breakout cable to run the top row of 4 drives. My only surprise was the numbering of the breakout cables. Drive 0 is on the right.

 

I also installed  a Supermicro AOC-SASLP-MV8 8-Port SAS/SATA Add-on Card and 2 SFF-8087 to SFF-8087 Internal Multilane SAS Cables. Drive 0 was on the right for these too.

Link to comment

Thanks for the write-up on your build. I just finished a build with a new Norco 4220. I installed 1 Discrete SATA to SFF-8087 Mini SAS Reverse breakout cable to run the top row of 4 drives. My only surprise was the numbering of the breakout cables. Drive 0 is on the right.

 

I also installed  a Supermicro AOC-SASLP-MV8 8-Port SAS/SATA Add-on Card and 2 SFF-8087 to SFF-8087 Internal Multilane SAS Cables. Drive 0 was on the right for these too.

 

I am about to undertake a 4224 build, and I need to clarify this numbering as well.  For the purposes of the reverse breakout cable, how exactly are the drives numbered?  Since I have 6 SATA ports on my MB, one reverse cable will be only using 2 of the drives.  When you say Drive 0 is in the right, are you talking from the front or rear of the RPC case?

Link to comment

If you are using the OEM norco cables, drive zero is the right when facing the front of case.

 

this is true of both the reverse breakout cable and the SAS cable.

 

I belive i pointed this out in the norco thread.

 

PS. the Norco cables are just rebranded molex.

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.