SOLVED Ditching the HBA and connecting HDD to Mobo

[mistermanko]

In my newly found hobby of reducing power usage of stuff in my household, because energy costs skyrocketed in Germany, I thought about ditching my mighty x-flashed Dell H310 and connect everything to SATA.
My motherboard has enough slots, so I should be good right? RIGHT?

Jokes aside, I didn't found any exact power usage of RAID cards, ServeTheHome lists some cards with values between 6 to 12 watts. The max wattage makes about 1/10th of my total consumption, so it should save some money, in the long run.

Anything I didn't thought about? Supermicro specific BIOS settings to deactivate hardware RAID maybe...

 

[ChrisRJ]

It's what I've been doing for over 10 years with various boards. While that in itself is purely anecdotal, I am not aware of any limitations from this forum.

Well, perhaps if you want to run TrueNAS in a VM, a separate HBA for pass-through is needed. I even seem to remember that the HBA introduces some latency (albeit only a small amount) relative to the ports on the motherboard.

 

[mistermanko]

Thanks Chris,

perhaps if you want to run TrueNAS in a VM,

that I did have in mind too, but the Grinch scared me away :lol:
I'll shelve it for now.

 

[jgreco]

some cards with values between 6 to 12 watts.

Usually a bit tighter than that, figure about 8-10, for older 8i cards like the 9211 and 9240. But still ballpark correct. This just follows my usual testing technique of taking a large E-ATX system, finding out if it's watt-stable when redirected into BIOS using the bench power meter, then cramming like half a dozen of the card in question in and seeing where the watts end up. Then divide the difference by the number of cards.

I've had some people all shocked and in horror that I don't have some special magic testing jig measuring the exact wattage at the PCIe connector. You can do that but damn it's a lotta work to set up. My questions are usually along the lines of "Is the Solarflare SFN6122 more efficient than the Intel X520-SR2 (answer: yes by about 50%)" and this can be measured easily by the technique above.

 

[jgreco]

Well, perhaps if you want to run TrueNAS in a VM, a separate HBA for pass-through is needed.

This is not necessarily true. You may be able to pass thru an entire SATA controller on the mainboard, as long as it isn't needed for anything else (like hypervisor boot). This works out fine if you are using something like a RAID controller to host your hypervisor's boot and VM's.

I even seem to remember that the HBA introduces some latency (albeit only a small amount) relative to the ports on the motherboard.

The mainboard ports are optimized for their job and work well especially with SSD's. HBA's, especially older ones such as the SAS2008, may already be at a disadvantage of being PCIe 2.0, but you also have to remember that there's an onboard MIPS CPU shoveling the traffic around, which is an extra handling of the data. Especially the older 6Gbps ones (2008/2308) are going to be a bit underpowered for the job compared to the modern (12Gbps) ones, because they were designed in the era before SSD's were "a thing."

 

[mistermanko]

A thought just came up...
If I just shift the work from the LSI card to the integrated SATA controller. How am I saving energy? Or is the SATA stuff always pulling the same wattage as long as the mobo is powered...?

an onboard MIPS CPU shoveling the traffic around, which is an extra handling of the data.

Ok, I think you may have answered the question with the post above.

 

[jgreco]

The SATA hardware takes a fixed amount of power because it is built in to the mainboard. If it can be disabled with a jumper, this might reduce power consumption a bit.

Adding in an LSI HBA adds electronics for eight SAS lanes, and also a MIPS CPU that is constantly running code in order to service requests, on top of the SATA hardware on the mainboard. It is generally believed that the MIPS CPU is the majority of the watt burn on the HBA. Whether or not this is true is hard to know, as it is integrated into a "RAID on Chip" controller chip. Anyways, the big frickin' heatsink that typically sits atop the RoC chip should make it clear that it burns watts.

 

[mistermanko]

the big frickin' heatsink

on which I mounted a tiny little Noctua fan because I was afraid the card could overheat. I'm still not sure if that did anything. But yes, it was warm to the touch and dissipated heat means power consumption.

 

[mistermanko]

In my newly found hobby of reducing power usage of stuff in my household, because energy costs skyrocketed in Germany, I thought about ditching my mighty x-flashed Dell H310 and connect everything to SATA.
My motherboard has enough slots, so I should be good right? RIGHT?

Jokes aside, I didn't found any exact power usage of RAID cards, ServeTheHome lists some cards with values between 6 to 12 watts. The max wattage makes about 1/10th of my total consumption, so it should save some money, in the long run.

Anything I didn't thought about? Supermicro specific BIOS settings to deactivate hardware RAID maybe...

Replying to myself, just wanted to share that switching from HBA to MOBO has saved me 30 watts in idle power consumption.
At the current rates in Germany that's about 100 Euros saved per year.

 

[Morris]

You did not say how your system is used. SSDs use les power than hard drives. Hard drives can be spun down and it might be worth laying out your pools so that data that is frequently accessed is on SSDs or on a small number of hard drives. Then spin down less frequently accessed drives.

Some people will be worried about drives waring out from spin down and back up. Modern drives are designed to do this.

A lot of people build there NAS using old components. If you did this, a newer motherboard and processor may save you power yet it will take quite some time to make up the cost unless your processor is usually at high CPU usage.

 

[mistermanko]

You did not say how your system is used.

A glance over my signature would give you a hint on how my pools and usage is laid out. But to help your point, I'll elaborate happily:

RAIDZ2 6 x 4 TB HDD for file storage of any kind.
MIRROR 2 x 1 TB NVME for apps, VMs, databases.

On the Z2 pool resides user data as well as my media library and bunch of other stuff that needs frequent access through apps (Plex, nextcloud, downloads) which are served via the NVME pool. This is the reason I never ever would use spin down. LATENCY.
It may have benefits on power usage but will introduce a heavy amount of latency to many of the used apps, which I can't tolerate. On whether modern drives are less subject to wear-and-tear through spin down, I've read mixed opinions here on the forum.
Yes you can use spin down, but don't do it if your TrueNAS is more than just a networked attached storage.

A lot of people build there NAS using old components. If you did this, a newer motherboard and processor may save you power yet it will take quite some time to make up the cost unless your processor is usually at high CPU usage.

Cpu and Motherboard are actually a smaller fraction of the power usage of your system. If you don't go crazy with high TDP processors or multi-cpu-boards, it really doesn't matter if you pick a X9 family Supermicro board or a recent Ryzen system.
I've seen only one application here on the forum that is excessively hogging your cpu which will result in extreme power usage and that is software transcode - be it for plex, zoneminder or anything other media related. Don't do it unless you pay no power bills or have money like hay.
The rest of the total power usage consists of storage related stuff. spinning drives, HBAs, a well sized PSU, even the active cooling fans.

 

[ChrisRJ]

Some people will be worried about drives waring out from spin down and back up. Modern drives are designed to do this.

You have said so in the past, presenting your personal opinion as a general fact. However, when @joeschmuck asked you for a reference to this claim, you did not provide one. It is of course ok to have an opinion that differs from common consensus. But presenting it as if it were an objective fact and generally agreed-upon is potentially harmful to those who make decisions based on it.

Would you care to elaborate how you come to your conclusion in this matter?