BUILD First NAS build

[depasseg]

Good point. That last picture was from the IPMI in my supermicro. So is this one. I don't see a DC 5v output though.

 

[marbus90]

that's because the PSUs handle only 12V and 5Vsb, other voltages are generated in the power distributor.

Gentlemen - now that the always difficult power supply conversation is out of the way, what about the MB, RAM, and CPU?

With OPs initial configuration there was no problem at all. only some people claimed that 360W wasn't enough for such a tiny system.

btw, the FreeNAS Mini uses a 250W PSU for 4 HDDs and the same mobo.

 

[mjws00]

There you go with real server gear that has only tagged 316w. Real SAS drives. Twice as many of them. Real Fans... Pretty cool that your sensors even show the efficiency of the PSU (watts in vs out). Even if it is smoothed and +-5% still shows watts up. (heh.)

 

[jgreco]

With OPs initial configuration there was no problem at all. only some people claimed that 360W wasn't enough for such a tiny system.

Who claimed that? It wasn't me. I merely said that it was cutting it close and suggested the 450W. It's a smarter choice. And I showed my reasoning.

btw, the FreeNAS Mini uses a 250W PSU for 4 HDDs and the same mobo.

And the AsRock Rack 1U12LW-C2750 also has a 250W PSU for 12 HDD's and that same mobo.

I'm sure that means it is perfectly fine to just go and trust that it will work out and not fry your drives. No, but seriously, just because someone designs something, doesn't mean it is engineered well. This whole discussion is about properly engineering things. The FreeNAS Mini is a poster child example of everything wrong with PCland. It's a product for a company that's trying to produce something that balances functionality with not being totally priced unreasonably out of the marketplace compared to similar units from other NAS vendors. That doesn't mean it is well engineered.

If you don't give a flying f about proper engineering, the original poster would be just fine putting a 150W power supply in there and just hoping that there's enough margin in the supply to handle the occasional spikes from drives starting up. I'd give it 90% odds on appearing to work just fine. But I'd also give it great odds that it'll cause damage to the drives eventually, and that the power supply will catastrophically fail by the time it hits five years.

It's the difference between "that's good enough to be likely to make it through the 1 year warranty period without a statistically unacceptable number of failures" and "this is engineered well enough that it'll probably still be just fine and dandy in 10 years."

We encourage people to buy ECC because we care about the integrity of their data and because we believe that they want to be able to reliably store data for the long term. Why would you cut corners on something as trivial as a power supply? It's a foolish thing to save ten bucks getting a smaller supply when your entire system's health is dependent on the proper operation of the supply.

 

[marbus90]

Ok. I am giving in. There's no point reasoning with you. We've listed enough examples to prove that a Seasonic, a proper brand, 360W PSU is sufficient for OPs system. We did not recommend <360W china cracker PSUs for OPs system. We did recommend Seasonic 360W because reasons, well explained reasons.

 

[mjws00]

The FreeNAS Mini is a poster child example of everything wrong with PCland.

I do like this one. ;)

 

[Bidule0hm]

It's the difference between "that's good enough to be likely to make it through the 1 year warranty period without a statistically unacceptable number of failures" and "this is engineered well enough that it'll probably still be just fine and dandy in 10 years."

Yep, I usually like to over-engineer what I build but here there's also the fact that a very lightly loaded PSU (let's say 5%) can be unstable (but again, a well designed PSU...) and will be less efficient. If this wasn't a problem you can throw in a 1 kW PSU and be safe forever (well, almost...).

Ok. I am giving in. There's no point reasoning with you. We've listed enough examples to prove that a Seasonic, a proper brand, 360W PSU is sufficient for OPs system. We did not recommend <360W china cracker PSUs for OPs system. We did recommend Seasonic 360W because reasons, well explained reasons.

Here we can be right that the 360 W PSU will be ok, but jgreco recommendations are right in a more general sense so I'd follow that personally ;)

 

[jgreco]

Ok. I am giving in. There's no point reasoning with you. We've listed enough examples to prove that a Seasonic, a proper brand, 360W PSU is sufficient for OPs system. We did not recommend <360W china cracker PSUs for OPs system. We did recommend Seasonic 360W because reasons, well explained reasons.

Yes, it's probably sufficient for the OP's system, but that's not what you came into this thread saying. You came in saying:

360W is enough for any array of 6 HDDs. that's even more than overplenty for that.

It is certainly not enough for an array of 6 HDD's on a Xeon system, therefore, it is not suitable for"any array".

It is also not "even more than overplenty" for an Avoton platform. Again, math provided.

While the 360W is probably sufficient for the OP's system, the 450W is still a better choice for the OP's system, a point the OP figured out before your first post in this thread.

As for your "reasons, well explained reasons", I have no idea where you held that discussion. I did see some relatively poor handwaving on your part, while I provided an actual breakdown and analysis in depth.

 

[jgreco]

Yep, I usually like to over-engineer what I build but here there's also the fact that a very lightly loaded PSU (let's say 5%) can be unstable (but again, a well designed PSU...) and will be less efficient. If this wasn't a problem you can throw in a 1 kW PSU and be safe forever (well, almost...).

Yes, this is why it is important to contemplate the minimum system load as well as the maximum. Going under 10% load on a PSU is getting into sketchy territory, and probably requires some consultation with the sales engineers for your PSU vendor. As the number of drives on a system increases, you end up with this ever-widening gully between the minimum and maximum system loads. Consider a 48 drive system, where the idle drive is consuming 6 watts, that's 384 watts, but a spinning-up drive is consuming about 35 watts (2.1A at 12V plus... ~~10 watts?) is 1680 watts. So many times the manufacturers start to rely on staggered spinup, which seems like a fantastic idea, as it means you could get by with a smaller supply. The problem with this is that if you employ drives that spin down, and then engage in some activity that causes a large number of them to spin up simultaneously, you risk the same problems. The flip side is almost all such systems have redundant power supplies, so you can take a risk that all PSU's are online when the spinup events occur, and as long as they are, you suddenly do have the ability to soak up greater-than-rated-load. Of course, that's both cheating and a bad idea. ;-)

 

[Bidule0hm]

I almost implemented staggered spin-up (DIY, as always...) but not for each drive, just for 2 halves, to lower the stress on the PSU but then I chose to not do it because the complexity overweighted the benefits and the more complex a system the more likely it'll fail... :)

I'd say a good old linear PSU would be great :D A transformer can handle a huge overload for many seconds so if you size the rest of the PSU correctly you can have a 200 W PSU who can handle 1 kW for a few seconds without a problem for example, let alone the other benefits. But then you have the one big problem of size and weight :P

 

[jgreco]

let alone the other benefits.

The mad scientist lab vibe alone would be worth it.

 

[Bidule0hm]

Come on... this one is ridiculously small, you want at least this one :P :D

 

[jgreco]

Come on... this one is ridiculously small, you want at least this one :p :D

55,000 watts, huh. So now we could spin up a server of 1600 drives! With 6TB drives that'd be almost 10PB. Wonder if I'll see that amount of storage in 1U before I die. I mean, it's only been 15 years and we reduced 5TB that filled a rack into a single 3.5" disk...

 

[Bidule0hm]

Well, if you live about 15-20 years more it's very probable you'll see PB drives :)

 

[depasseg]

All right, all right. I think we've beat this "how does this new system look" thread into glue.