Interesting Storage Server - 4U 36 bay Atom 10Gb SFP+

[Patrick]

I recently saw this one on ebay and it struck me as a really interesting storage server. It has a SAS2308 add-in card, the expander 36-bay chassis and an Atom C2750 with 32GB RAM.

SUPERMICRO 4U STORAGE SERVER CSE-847 2.4GHZ 8C 32GB NO HDD A1SAM-2750F | eBay

Pricing at $1300 is slightly higher than getting good deals on lots of items separately, but then again, this is already assembled.

 

[Fritz]

That's the direction I going with my FreeNAS boxes.

 

[T_Minus]

same direction I went, just threw compute in there as well build for that in the next few weeks.!

 

[Black Ninja]

This is nice. Even has 10Gb nic , which won't work at full speed but still good to have.
I've been loking for 847 chassis like that but I got scared reading somebody post who said they are very loud.
Does anybody has 847 chassis to share insights about them ?

 

[BackupProphet]

They are not very loud. I would rather say, they are surprisingly quiet, but not completely silent.

 

[Stereodude]

Isn't that one of the Atom chips with the clock bug?

 

[Patrick]

Isn't that one of the Atom chips with the clock bug?

RMA-able at Supermicro, but reported failures are not high.

I have chatted with a lot of vendors and end users and after doing my research, I am still using a bunch of Atom gear.

 

[Stereodude]

RMA-able at Supermicro, but reported failures are not high.

I have chatted with a lot of vendors and end users and after doing my research, I am still using a bunch of Atom gear.

If you want to buy a ticking time bomb, more power to you I guess, but potential buyers should be aware of what they're getting.

 

[Patrick]

If you want to buy a ticking time bomb, more power to you I guess, but potential buyers should be aware of what they're getting.

All parts have failures. To give you an idea, I was talking to some IDC analysts and they said Atom C2000 series is tens of thousands of units a quarter.

I have spoken to folks that have sold/ or deployed hundreds of thousands of units combined and nobody has said it is like a 10% AFR after 18+ months.

I think there is a lot of FUD being spread by folks who hear bug and did not do research on how bad it is. The bigger impact is that since vendors have plans funded by Intel to remedy, actually servicing all of the old platforms in the field is the harder part.

On a new platform, I would likely just get the machine and RMA immediately to get the fix as a pro-active measure if I were concerned.

 

[Black Ninja]

They are not very loud. I would rather say, they are surprisingly quiet, but not completely silent.

Thanks for the input. I do expect some noise. I have few 836 chassis and I am hoping these won't be louder than that.

I also wonder how well the hdd on the back will be cooled, considering that the front will get fresh cold air , but the back ones will have to cool with warm air from the front ?

 

[BlueFox]

Thanks for the input. I do expect some noise. I have few 836 chassis and I am hoping these won't be louder than that.

I also wonder how well the hdd on the back will be cooled, considering that the front will get fresh cold air , but the back ones will have to cool with warm air from the front ?

The 847 chassis are considerably louder than the 836/846 ones in my experience.

As for temperature, thermodynamics don't really work like that. The temperature gradient between the front air intake and what is hitting the rear drives will be minimal given sufficient airflow.

 

[Black Ninja]

I wonder why would be louder ? It seems 847 uses the same fans that 836 has, so connected to the same MB they should should spin the same speed and therefore the similar in noise, no ?

 

[Black Ninja]

You put my thermodynamic knowledge in question But I bet the front drives will be cooler than the back and unless you have an actual data to prove me wrong.

 

[BlueFox]

847 has 2 layers of fans with 7 total while the 836 only has a single layer with 3 total (in the middle at least). The 2 rear fans on the 836 are smaller and slower.

There will indeed be a difference in temperature, however it's nothing notable. You shouldn't worry about 1-2 degrees.

 

[Black Ninja]

There are more fans (7) indeed , but considering the fact that they are the same type of fans like the front 3 fans of 836, wouldn't the noise be similar to 836 when they spin the same speed ? Right now in 836 my fans are spinning at 2400 Rpm connected to X9SCM-F. I mention the MB cause I notice that on one MB fans can spin twice faster than other MB changing the chassis from almost silent to most loud in the rack.(In both cases fans are not spinning at max speed).

The whole problem with home user is that I can put my hands on everything before I buy it , if I could there won't be questions or doubt what to get. But when operating on small budget it matter if you spend $1-$2K on chassis that may turn to be the wrong purchase.

 

[BlueFox]

I think generally one would hook up the front fans to the backplane at least. Having double the fans will make it considerably louder and you'd want to run them at a higher speed than the 836 chassis as you'll need the higher static pressure.

I wouldn't recommend this chassis for home use unless you can hide it in your basement/garage/etc. I moved mine to the datacenter and kept my 846 chassis around instead.

 

[Black Ninja]

I usually don't use the backplane for the fans, but I would like to know your opinion about it ? I just think that if fans are not connected to the MB, they won't adjust speed based on CPU and chipset temp, also will be invisible for the IPMI(you can monitor their speed and operation and can alert on email when fan fails , and I don't know if it will make them run on full speed when fain fails to help compensate to the loss like it would when connected to MB). I could be wrong ?!

I have a server rack in the basement , but I still need to control the noise in reasonable low limits, cause I need to be able to stay in the basement for other purposes too. I can use it as dedicated server room. I was thinking about 846 which should be where 836 is where the noise is, but I like the high density idea instead of dual 836 chassis.

 

[PigLover]

If you have the rught SM motherboard it will have an SMbus connector for the front and rear backplanes. If these are connected the fans will appear in the IPMI sensor data (as well as the temp and voltage sensors from the backplane). At that point they will be under IPMI speed control as part of Zone 1 (the IO zone).

With the fans spun down to about 3-5k rpm the noise levels are quite tolerable even with 7 fans spinning.

Without those connections the fans connected to the backplane will run full speed. Unfortunately the MB in this particular offer does not have the SMbus connectors for the backplanes...

Sent from my SM-G950U using Tapatalk

 

[Black Ninja]

That's very interesting to know. I know that the MB it does have a SMbus for Power Supply. I always use that connection to monitor PSU from ipmi, but I don't know much about the "SMbus connector for the front and rear backplanes" I want to find out more about this connection to the backplanes. If you are saying this MB don't have it , can you point a MB that has it so I can look at it closely to find out where on MB is this connector and where on the backplane it goes ?

"With the fans spun down to about 3-5k rpm the noise levels are quite tolerable even with 7 fans spinning." - Thats what I am hoping too, Full speed on fans all the time on server chassis is not an option on any chassis if you ask me.

 

[i386]

If you have the rught SM motherboard it will have an SMbus connector for the front and rear backplanes. If these are connected the fans will appear in the IPMI sensor data (as well as the temp and voltage sensors from the backplane). At that point they will be under IPMI speed control as part of Zone 1 (the IO zone).

You're talking about the i2c connectors (=External I2C Header on mainboard and I2C Connector#1-4 on the backplane), right?