Dd test vast transfer speed differences between same drive model

Hello guys,

my system is using a Supermicro X11SCA-F board with 64GB ECC RAM and a Xeon E-2124G. I have four Seagate IronWolf 4TB drives (ST4000VN006) and I am having slow transfer speeds so I did some testing of the drives. Currently I am running only two in a mirror. One of them performs at 168mb/s and the other at 204mb/s. Which is already not great. I wanted to go for striped mirror so I bought two more drives. Both are relatively new. One performs at 168mb/s but the other new drive only reaches maximum 138mb/s when using dd if=/dev/xx of=/dev/null bs=1M count=5000.
Can a faulty cable cause the lower speed here or is there something else I am missing?

Easy enough to determine. Swap drives between cables, see if it follows the drive or cable or port.

I did. It wasn’t the cable unfortunately

On my drives easily there is a 15mbps difference read/write between slowest and fastest.

Is normal from what I understand.

But 138mb/s and 204mb/s is a lot. 168mb/s seems to be the average speed

Whay bios does the board have? Firmware Resources | Support

Oh sorry, I thought this was solely a comparison between 138 & 168. That is my bad. Yeah 138 vs 204 is massive.

Either you got 1 rotten apple, or 1 golden sample there… Or worse; both at the same time.

Really annoying questions - are you certain they’re the same model? Any difference in terms of how much data are on the drives? Any irregularities on smart tests?

@Okedokey Is there a way to check that within Truenas?

I think the one that reaches 204mb/s is a golden sample since the specs say max 202mb/s. 168mb/s is what I have on two other drives. The 138mb/s seems to be a real dud. The only thing I noticed is that it has an all red label compared to the other 3 drives but the model number, firmware etc. is identical still. Also no SMART issues. I am really at a loss. I will probably try to get it returned. I bought a new one to see if that will perform more in line with specs. I doubt the BIOS is the issue here.

This makes the most sense.

I’m wondering if the all red label is some voodoo where they changed something physically without bothering to make a new model number or updating the spec. Regardless of it showing the same model/firmware reported even in software… wouldn’t be the first time such nonsense has happened from manufacturers.

Hopefully the replacement doesn’t work even worse - that’d be sad.

Oh yeah but I have a two week return window if it turns out it’s just as bad. But yeah shenanigans like that has happened before.

Little update. New drive arrived. Speed is 215mb/s. So the old drive literally was a dud.

1 Like

Now I have another question. Two of my drives perform at 165mb/s. Two perform at 202 and 215mb/s. Would it help to defrag the slower drives in Windows since I am planning to redo my pool anyways. The performance differences are still a little too large for my taste and I wonder if that would help in any way.

No.

Defragging a drive in Windows works with a Windows native file system such as NTFS. When you use it in TrueNAS, a ZFS file system is created on the drive.

So it is just luck of the draw on those drive performances? I guess I have two really good samples here at 200+ mb/s. Would it help if you use those in a striped mirror or use one 200mb/s drive once per mirror?

I wouldn’t read too much into the figures that you get from a simple dd command. The drive’s data sheet shows a maximum sustained write speed of 202MB/s, but this is not a promise that the same speed will be achieved regardless of where on the disk surface the data is being written. It certainly won’t be constant.
As long as the drives work, just press them into service and use them.

1 Like