Mini-ITX NAS build

[MikeC]

and since there are two fans in the case, should I buy 2x Zalman Fan Mate and 2x adapter?

First, you don't need any adapters, as the fans come with both 4-pin molex and small 3-pin fan plugs.
Secondly, I might just try the fan headers on the motherboard first. I think we had limited control on the Zotac fan headers, but they do get changed in the BIOS (and sometimes in minor hardware revisions) so it might be worth trying. One fanmate can probably handle both fans -- with a splitter cable. Or maybe you can find something less cumbersome -- the fanmate has a very long cable. There might be some other convenient small 2-fan header controller at your shop.

As for testing w/ the Nexus fan, it's routine for us to try and have only one variable in the HS reviews -- just the HS itself. So if the fan can be swapped for our reference, then we use it. Same approach with bigger HS that require a bigger fan. This way you can see the actual HS performance -- not affected by a slightly better/worse fan than the competitors. (The Nexus 92 is a very smooth quiet fan that doesn't blow much air.)

And it's nice that you feel grateful for the help, but unfortunately no deals with any EU online sellers. Just spread the word I guess... and you can always donate $ via paypal.

[zprst]

ram: you wont need more than 2gb. ever. ddr3 wont make a noticeable improvement on a nas either. 1gb of ddr2 would probably be fine, but with prices you could consider 2gb just to have it.

i definitely would consider maxing out on RAM, as far as your budget allows you to: depending on your workload and software configuration, larger amounts of free RAM may result in a noticeable performance increase, both on read and asynchronous write operations, due to the aggressive filesystem caching mechanisms of most modern operating systems...

by the way, a nice way to use the 5.25" slot on PC-Q08: a simple cheap SATA hot-swap drive bay. quite handy for backups, for example.

[alfonso]

So, I decided also the last bit of my configuration. The HDs will be the new SEAGATE BARRACUDA GREEN ST2000DL003 for various reasons:

I found problems for both alternatives I had (Samsung HD204UI and SAMSUNG SPINPOINT F4 HD204UI)
- the Samsung HD204UI had some issues not so long ago: http://sourceforge.net/apps/trac/smartm ... GBadBlocks (Seriously? the patch does not change the version number? how lame is that!)
- the WD20EARS had a lot of problems: so much so that it's deprecated by Qnap: http://www.qnap.com/pro_compatibility.asp (click on the gray box to see the actual list)

Also the Seagate seems to do well in this review:
http://www.storagereview.com/seagate_ba ... t2000dl003

Plus I'm curious about ZFS and I might play with it (FreeNAS) and Seagate seems to have and edge on 4k sector thorny issue...
http://forums.storagereview.com/index.p ... -with-zfs/

I don't fool myself: the only reason why there are no complains on these disks online is that they are spank new. But I decided to trust "someone" who didn't screw up yet compare to the ones who did already (WD and Samsung).

So finally my shopping list is:

LIAN LI PC-Q08 81 eur (for some strange reason this is listed as discounted in the shop, so good discount)
COOLER MASTER SilentPro M600 95 eur
Zotac H55-ITX WIFI ( H55ITX-A-E ) 61 eur (it's not the latest H55ITX-C-E version with the USB 3.0 but it has a huge discount!)
UPDATE! the website where I bought the MB did not explain that for them end of stock and 2nd hand it's the same thing... so long story short I went to pick up the MB and inside the box there was some other MB, so I gave it back and ordered the H55ITX-C-E online for 122 eur
INTEL i3 550 119 eur
ARCTIC SILVER 5 5 eur
Scythe Samurai ZZ 25 eur
CORSAIR XMS3 DDR3 4 Gb 45 eur (a bit of an overkill? maybe...)
4x Seagate Barracuda Green SATA ST2000DL003 2 Tb 77 eur/each (I bought two disks in a shop and two in an other, to limit "bad batch" effect)
SATA cable (the MB comes with three, so I need one more) 4 eur

For which CPU to buy and size of the disks I chose based on best ratio (ratio of price per CPU mark and price per TB).

At the end, due to some discount in one shop I managed to pay 736 eur including shipping... (61 eur more due to the MB issue...)

next chapter... building it!

thank you for all the info...

[HFat]

4G is overkill... unless you use ZFS. Maybe you'd be OK with half as much but a forum member said it was unsafe to run ZFS if you don't have a lot of spare RAM. I haven't tried it myself.

The slowest CPU of that generation is way overkill for a NAS. So there's no reason not to pick the cheapest. It will be branded "Pentium" and not "i3" but it's every bit good for a NAS.

[alfonso]

4G is overkill... unless you use ZFS. Maybe you'd be OK with half as much but a forum member said it was unsafe to run ZFS if you don't have a lot of spare RAM. I haven't tried it myself.

The slowest CPU of that generation is way overkill for a NAS. So there's no reason not to pick the cheapest. It will be branded "Pentium" and not "i3" but it's every bit good for a NAS.

thanks! but it won't be just a NAS, the box will also be connected to a big screen, so extra CPU (for such a small price difference) is welcome, especially if one day I will go crazy and buy and iPod and I will need to use this: http://www.inmethod.com/air-video/index.html (live streaming and real time conversion of videos for iPhone and iPad...)

about the memory: there is no logic... I just felt like "why not?"

[HFat]

Connecting it to a big screen and decoding videos doesn't require much CPU power. Transcoding for small devices isn't very taxing (unlike transcoding for high-resolution displays). The slowest Clarksdale is very powerful already.

Something else: you might want to mix drive models if you want to avoid not only the "bad batch" effect but also the "bad model" effect. I tend to mix Samsung, Hitachi and Seagate drives.

[alfonso]

Hi,

I got the case and I realized that I will have three fans (2 for the case + 1 for CPU) and the MB has only two connectors...

plus, I read here:

The H55-ITX has two PWM fan headers but unfortunately the one labeled SYS rans all fans connected to it, both 4-pin and 3-pin, at full speed. The board's fan control worked only with the CPU fan header and only if a PWM fan was used. The CPU fan behaved in accordance to the settings in the BIOS.

The two case fans of the Lian Li PC-Q08 are both 3 pin and come with adapter from 3 pin fan connector to 4 pins 5-1/4in IDE connector (something like this, not the pass-through kind, but you get my point).

My question is: does the fact that the fans have only three cable means they cannot be controlled based on temperature? I think so based on the wiring schema of the fan connector on the MB: http://en.wikipedia.org/wiki/Computer_fan#Fan_connector

Ok, I think I found a solution... newegg sells Y splitters:

http://www.newegg.com/Product/Product.a ... -_-Product
http://www.newegg.com/Product/Product.a ... -_-Product

and in the feedback of one of them I found this:

this item is used to operate 2 fans as if it were one fan.
you can use 2 or 3 linked y conectors and all fans will have there speed ajusted by the mother board at the same time.
im using 2 linked right now with 3 fans all have same speed given by mother board. mother board can only read one rpm per header thats why there's only one that has 4 pins while the other has 3

So this means that with two splitters, all fans will vary their speed according to the control provided by the Scythe SAMURAI-ZZ which supports PWM according to specs.

guys, do you agree with this analysis? Is there any reason why connecting all three fans to the same MB connector is a bad idea? I really wish to build a system that spins fans only when needed, and as slow as possible.

thanks!

[MikeC]

guys, do you agree with this analysis? Is there any reason why connecting all three fans to the same MB connector is a bad idea? I really wish to build a system that spins fans only when needed, and as slow as possible.

YES. This is a bad idea. I would not run more than 2 at most. Fan headers on most mobos are quite limited in their current capacity. When you turn on the power, the surge could burn out the circuit.

The only fan you really need to be thermally controlled (and not really even then) is the CPU heatsink fan. The others are best run at a low steady speed.

There is a really simply way to get 5V from the 4-pin Molex adapters that come with the case fans. An article on this topic is one of the earliest in the fan section of SPCR. Also described in the recommended fans page, iirc. It requires moving one of the pins from the 12V positions to the 5V position.

Or use the adapter to connect feed a Zalman fanmate directly to the PSU. This can easily power/control 2 fans w/ a Y at the fanmate output.

[alfonso]

YES. This is a bad idea. I would not run more than 2 at most. Fan headers on most mobos are quite limited in their current capacity. When you turn on the power, the surge could burn out the circuit.

The only fan you really need to be thermally controlled (and not really even then) is the CPU heatsink fan. The others are best run at a low steady speed.

There is a really simply way to get 5V from the 4-pin Molex adapters that come with the case fans. An article on this topic is one of the earliest in the fan section of SPCR. Also described in the recommended fans page, iirc. It requires moving one of the pins from the 12V positions to the 5V position.

Or use the adapter to connect feed a Zalman fanmate directly to the PSU. This can easily power/control 2 fans w/ a Y at the fanmate output.

Mike, thank you!

I believe you refer to this aticle: http://www.silentpcreview.com/article6-page1.html

I understand and respect your feedback but your suggestion is to manually set the speed slower (either by under-volting from 12V to 5V or having the same result using the Zalman Fan Mate).

Wouldn't this create problems if more airflow is needed (i.e. under load)?

Is there any off the shelf solution that includes a temperature-based regulation?

thanks!

UPDATE!

what do you think about this? (AKASA AK-CB002 PWM Splitter)

seems to me the perfect answer to my wish and your concern about the power surge!

[MikeC]

I believe you refer to this aticle: http://www.silentpcreview.com/article6-page1.html

Wouldn't this create problems if more airflow is needed (i.e. under load)?

Yes, that is the article.

Only if you were to stress the cpu at 100% in hot weather for something like >half an hour. This is so unlikely in a NAS. I've been running this type of cooling setup in my PCs for years. Never a problem w/ overheating anything.

what do you think about this? (AKASA PWM Splitter)

seems to me the perfect answer to my wish and your concern about the power surge!

Yes, that looks clever -- but you will need all the fans to be pwm. Your case fans are not.

[alfonso]

Yes, that looks clever -- but you will need all the fans to be pwm. Your case fans are not.

Hi Mike,

actually I think that's not the case.

If you look here you can see than it's meant for only on PWM fan (the CPU one).

Actually after thinking a bit about it I found the problem with this setup: if you have a hot GPU and many disks this can be a problem since the fans will spin only according to the CPU temperature without regard for the temp inside the case, but this is fine with me since my system doesn't have an external GPU and has only three green disks with very low power consumption (so I guess they won't heat too much).

[lodestar]

Actually after thinking a bit about it I found the problem with this setup: if you have a hot GPU and many disks this can be a problem since the fans will spin only according to the CPU temperature without regard for the temp inside the case

This isn't quite how it works in practice. The CPU temperature will be affected by both ambient temperature and heat being generated inside the case. High performance gaming cards do put quite an additional load on the CPU as well as stressing themselves, so going from idle to intense gaming with such a GPU might well see a rapid increase in CPU temperature of around 15 to 20C. That's why gamers typically are the most common users of PWM fan chains using splitters, particularly for the CPU plus exhaust fan combination.

[azazel1024]

If I was that concerned about ramped up temperature under heavy server load, I'd just spring for some thermally controlled fans. Gelid makes really great silent fans with good air flow. They have PWM controlled fans, a similar spec fan that is 3 wire that is thermally controlled on its lonesome, and plain 3 wire fans that run a little slower than the other two do at full speed.

For $10 a fan, pretty cheap to replace them.

You could also look at something like this http://www.frozencpu.com/products/5300/ ... g47c17s422, but frankly buying a new thermally controlled fan would be cheaper.

You are unlikely to generate a lot of heat in that case. You could probably get away with a single 120mm intake and 120mm exhaust fan + PSU fan. If not less than that, especially if you'll tolerate a bit more heat in the case.

For price and what you want to do with it, I'd go with a basic Clarksdale Pentium chip. You really just don't need much computing power in there, even for HTPC work and smaller transcoding unless you need to get stuff transcoded fast. I transcode using handbrake DVDs up to about 840 pixels on the long end with pretty high quality settings and can usually manage about 28-35fps using my Core 2 Duo E7500@3ghz. That processor is probably about the equivelent to a clarksdale pentium or maybe even a hair slower.

For HTPC work you need even less processing power (though not exactly scads less).

For a file server you need very little. You can get away with an underclocked single core Sempron (just how much underclocked I intend to find out with the build I am just finishing) with a fraction of the computing power of that i3.

I would personally seperate file serving and HTPC duties, but that is me.

Current build of a file server is running to $250 including the 1tb Samsung F3 in it. Granted you've got a lot more storage in yours...but mine, when you convert for currency, is running about 1/4 the price and can max out gigabit ethernet just fine (sempron based system, cheap board, cheap case, good PSU, single 2GB stick of memory, single HDD to start with).

In the end there'll probably end up being another $300-400 invested in the setup with a dual disk eSATA enclosure (for backups) and several drives for both the enclosure and the file server itself. That still comes well under your current build budget. That is all down the road as it just expands the basic file server, the server itself is less than $300 in the end though.

[alfonso]

Hi azazel1024,

and thanks for the links!

I would personally seperate file serving and HTPC duties, but that is me.

If I had a bigger house, I would agree. but living in a small apartment it doesn't make sense to have two boxes.

The only thing I "worry" is that it would be cheaper (in terms of electricity) to run a 24x7 NAS with a much greener CPU and then start the HTPC only when needed. This might be the 2nd step, if I ever reach it.

I'll monitor the bills and see if it makes sense to invest in an other box to save on electricity.

[azazel1024]

You would probably save some on bills.

I think the most cost effective thing to do is get the most power efficient PSU you can, and one rated at no more than what you think the computer will ever use. Don't over spec it, that just leads to spending more and it makes it less efficient. You could almost deffinitely get away with a 300w PSU in what you are talking about. Could be a little iffy with all that drive spin-up, but AHCI will spin them up one at a time to prevent too much surge current on the PSU.

The Pentium clarksdale processors will probably draw a couple of watts less than the i3.

The big thing is, just make sure that the motherboard supports wake on LAN and make sure you use resonable drive sleep and system sleep settings. If it is being used as a file server and is only accessed a handful of times a day, it makes no sense for it to be idling 23hrs and 15 minutes out of every 24, it should be sleeping most of that time.

A high spec hexa core machine with a couple of 480s in SLI is going to draw just as much power as an Atom with integrated graphics is when asleep. Okay, about the same. Its really the LAN chipset and the PSU that is going to dictate how much power a system uses when asleep waiting for a WOL signal. Typically you should see between 3-7w of power use in S3 waiting for a WOL.

If you use appropriate sleep settings, that machine is probably going to spend 18-23hrs a day in S3 instead of idling. You are going to save relatively little money between say an Atom and what you are spec'ing out with the relatively idle power consumptions. You might save $10-20 a year, so long as both machines are sleeping most of the time. Idle 24/7, and yeah an atom system might save you $30-50 a year in power bills.

1 watt 24/7/365 is about $1 a year in power consumption. An idling machine using 25w (atom, give or take a little) compared to one idling at 70w (about an i3 without a big video card) is maybe going to be $45 a year in power bill differences (a little higher since it is going to not be at idle some of the time, and the power differences are going to be even higher then). Sleep it 80% of the time, and you have 20% of the power cost difference.

Just my 2 cents. I am a big greenie...but sometimes the impact really is not worth either the trade-offs you'd face or the $ savings you'd net. If it can't pay for itself within the expected useful life of the system, or at least can't pay for 50% of it (in my case I care a lot more about the environment than simple economic sense)...then it isn't worth it to me.

In my case, eventually building an HTPC and having the seperate file server I just built is more than saving some electrons and the environment (because heck, building two computers uses a lot of power and resources to make the components for two of them...though maybe not as much as the lower power would save). It my case its because an HTPC I want in as small of a case as I can resonably get away with and as bare bones as is possible and still work well as an HTPC.

A file server I care a lot more about low power consumption, but I don't need a video card (slow integrated graphics is fine for Ubuntu, thanks very much) and don't need much processing power, but I do need a fairly big case, lots of SATA ports and expansion slots for additional GbE NICs or RAID cards later for expansion.

Also its the OS thing for me, I want a Linux based OS for the file server...but for an HTPC I'd want Windows...and I don't want to have to run VMware or other virtualization software to accomplish this, so two physically and logically seperated machines is what I need.

Might not matter or make sense in your case, and it doesn't sound like it does, to have two physically or logically seperated machines.

I still advocate, don't over spec anything in the system in terms of performance unless you actually think you'll need the extra performance later. It just costs more money and uses more power.

[alfonso]

I agree, but now I made up my mind and all HW is on its way.

Yes, the server will run few hours a day, so hopefully mostly as you say I'll pay for idle, provided I configure it correctly!

[HFat]

(in my case I care a lot more about the environment than simple economic sense)

In that case the most rational choice will depend a lot on how your electricity is provided. Buying gear to save electricity makes little sense in Iceland. But it might make sense in Germany.
In many locales it remains mostly harmless to consume electricity for much of the day and especially at night. This is also something to take into account in your "green calculus". So buying new gear as opposed to second-hand often doesn't make green sense.

Even if you only look at the dollars, there are more variations in electricity prices than electronics prices around the world. So you can't really make statements like a watt will cost you 1$ over this or that time period.

[azazel1024]

True, but if you are paying your power bill in US dollars, I assume you are also living in the US. Even taking in to account varying electrical rates, 1 watt/year of electricity runs from about $.80-1.40 in the US based on the various electrical generation and transmission rates that I am aware of from low to high.

Prices certainly vary more across the world, but European electrical prices don't vary that far from that range that you find US prices or at times are worse. Looking at the European exchange, prices range from .09 Euro in Estonia to about .18 Euro in Germany for electrical rates. Converting to US dollars that comes out to $1.10-2.20 per watt/year of power consumption. So actually a little worse than the US range of typical prices.

More incentive to save power than if you live in the US/Canada.

[HFat]

There's a larger range than that in the US even if you consider only the continental United States (it's probably even more expensive in Hawaii). Someone posted their rates in NYC recently and they were a good bit higher than your figures.

The US dollar is used in some other countries like Ecuador and of course in US territories around the world.

[azazel1024]

Obviously a pretty old chart, but I found one from 2007 that lists the lowest rate in the US as being in WV at 6.2c per kw/hr and HI as 20.7c per kw/hr. It listed MD as 8.5c per kw/hr. Considering my current rates, I'd say all those on the chart are probably about 20% higher today than in 2007.

That means the range should be about 7.5c per kw/hr to 24.8c per kw/hr. That doesn't include transmission costs, which vary, but tend to be about 3c per kw/hr most places. So the actual range would be about 10.5-28c per kw/hr. Annual costs for 1w would then be $.92 to $2.46 a year.

So, you are right, more variability that I had estimated before, but really it is from a floor of about $1 per year, up to well over $2 per year, but less than $3 per year per 1w of constant draw.

The worse you'd see in a system consuming, say, 10w more than another system might be $30 a year. Something to keep in mind, but probably not a deal breaker. Now a difference of say 30 or 40w might make a pretty big impact, 1 or 2w is pretty much just forget about it unless your setup is power constrained in some way, such as penalties from your utility company by going over a certain peak draw (IE running an IT company), there is a large number of systems in play (so multiplying the effect of a small extra power draw), it is cheaper or the same price to use lower power draw components, the pay back is well within the expected component use life (I figure 3 years personally for most stuff), you care about the environment so price is less of an issue, or you are using your setup in a heat/power constrained environment such as a poorly ventilated enclosure, high temperature environment or have low generation capacity or storage (using the system in an RV, boat, etc).

[HFat]

In the low-cost market the electricity consumption might make a big difference because the price differences can be quite small.

There is also the price risk to consider if your electricity is generated by burning oil or natgas. Eletricity prices must have risen by more than 20% since 2007 in some locales.

But as far as consumers are concerned I think the main reason they'd look at electricity costs is not so much to inform a purchase but rather to decide if they're going to shut the computer down when they're not using it.

[justice99]

What about these motherbard , they contain an GPU+CPU, and the price is not that bad.

One of these + Lian Li Q08 , sound good no?

I also found these case mod , its usefull?