All-in-one box: NAS HTPC RAID; RAID5 write speeds: linux, h67, PERC 5i

Built-in motherboard fake "RAID" is usually worse than software raid since you are at the mercy, stability-wise, of the hardware. All of the parity calculations for these setups is still done by the CPU, so they offer no real performance benefits.

Cooling of the PERC in a small case is going to be challenging, but if you had the money for the card and space in the case, my vote personally would be for a true HW RAID if you need performance.

Ultimately, it breaks down to the constraints of the system and the needs you have for the storage solution. Work those out and the solution should be clear.

So from the research i've done, transfer speeds through the gigabit lan is about 90-120mbits (which is what my network is built on).

As far as i can tell the PERC 5i will do about 450mbits (which is clearly overkill other than when i need to rebuild the array).

What i'm really unclear on is the speed of the LINUX RAID 5 write speeds. I've seen anywhere from 1mbits-10mbits to 170mbits.


If a LINUX software raid 5 can achieve close to or better than gigbit lan speeds, then the perc 5 is over kill. I will still probably get the h67 to get the 6sata ports, and buy a expansion sata card when i run out of bays (i can have 7 hd with this case)



SO IT BOILS DOWN TO...what RAID5 solution would i need to take advantage of the gigabit lan without overkill?

THanks for any help you guys can drop on me.

Considering the low-power nature of the board, I would tend to agree with you that trying to get SW raid to feed a Gb pipe may lead to disappointment, but I would imagine that the on-board RAID is not going to be much better (if at all)

I'd say at this point I would look into lower-profile cards (and do some PSU budgeting calculations) or accept that the disk system may be the bottleneck (you should examine some of the various stripe size and read-ahead tweaks to try to tease extra performance out of the system before giving up). I suppose there's no harm in a graduated approach (trying software, if that doesn't meet your needs, give the mobo fakeraid a try, if that doesn't meet your needs, look for a card)

Thanks a lot for the information.

Do you really thing my 300w PSU is gonna cause an issue with my system? I tried to choose the lowest power proc and components.


In terms of a gradual approach, i was gonna use SW in RAID1 until the HD prices come down, but once i migrate to a RAID5, i dont think i can easily switch between SW, on board, and PERC5.


I guess i was hoping someone on this board was more intimate with the PERC5 and possibly the write speeds of a linux Raid5 (but still thanks for all the help)

A system with that CPU and no discrete graphics is going to use less than 100W under full load  

... yes, until you toss a RAID card and 4+ hdds into the mix

I believe the Samsung F4 spinpoints shouldn't generate that many w?

Am i wrong to assume this...i figure if they built the thing with a 300w psu, i should be ok.

Actually a hardware raid that I use only consume 6W and Green 3TB @ 6W per drives so RAID5 of 15TB raw cost less 40W and transfer about 210MB/sec. It's way over the Gb NIC

Aye, but you need to be aware that those ratings are usually rotational, whereas motors suck the most juice when at stall (read: spinup) and if all disks are spinning up at once the system may fail to boot simply due to startup draw...

..that said, 300W should be fine for your system

I did built a 40TB NAS/iSCSI system that only consume 180W of power, base on Atom CPU, Dual Gb Ethernet bonding in Load Balance mode... I can transfer to/from my Win7 @ 75~95MB/sec.

If you really read the spec you will see the power consumption is way over spec, which costs money unnecessary

Thanks a lot everyone, i've pretty much put this build on hold while the hdd prices are killer.

Looks like a late 2012 build for me =\!