Choosing the new hardware for TrueNAS SCALE Server

TrueNAS General
,
1

Hello,
at the moment im searching for new Hardware for my Server because the CPU is already pretty old and finally at a Point where everything gets laggy.

Im upgrading the RAM to 128GB ECC and i wanted to keep my 4-Port Network Card (PCIe x4).
In the Case i plan to use is Space for an ATX Board, 11 x 3.5" HDD and 4 x 2.5" SSD.
I wanted to plan everything futureproof, so that i can install these 15 Disks without Problems.

Transcoding isnt really a thing that happens often, when it happens then its just 1-2 Transcodes so i guess an iGPU should be enough for that?

I was reading that the Intel iGPUs are good for transcoding and the AMD iGPUs are bad at Transcoding.

For connecting all the HDDs i could easily buy an old HBA from the Company im working at but they are all PCIe x8 and the most Intel Boards have just x16 and x4/x1 PCIe slots.
So i researched a bit and found some PCIe x1 Cards with 2 SATA Ports or PCIe x4 Cards with 4/8/12 and so on SATA Ports.
So i wanted to ask if these “simple” Cards would work too or what the best to do is at this Point.
The x16 Port i wanted to keep free for an optional Upgrade to a GPU.

As CPUs i was till now looking at the following Intel and AMD CPUs:
Intel:
i5-12400, i5-13400, i5-12500
AMD:
5-5500, 5-5600, 7-3700X 7-5700X, 7-5700, 7-5700G

Maybe you can help me here out or give some additional advise.
Thanks :slight_smile:

2

First start with the documents section for TrueNAS Scale
https://www.truenas.com/docs/scale/24.04/gettingstarted/scalehardwareguide/

Also old forum hardware guide
https://www.truenas.com/community/resources/hardware-recommendations-guide.12/

You seem hung up on consumer, gamer gear instead of server choices. Look at used server gear and cases. You have planned 15 disks already and didn’t mention boot disk(s), add at least one, maybe two.

3

No - do not use these cards (the crappy PCIe 1x cards that is)
If you need extra ports use an LSI HBA 92xx or 93xx in a x8 (or better ) PCIe slot
Assuming your data is important to you

Put it in your “reserved” GPU slot and give up on the GPU which with the modern CPUs you are talking about and the number of transcodes you are talking about - you will never need an eGPU

2 Likes
4

Thanks i will have a look at this 2 topics.

I will have a look at outsourced hardware too.
Boot Disks are ofcourse already considered too, therefore i will take 2 of the planed HDDs

5

okay, then i will try my luck with that, we got some spare 9211, 9207 and 9201 laying around here.
Is there anything you could say about the CPUs?
I mean i am with the HBA tied to 1 with an iGPU but from the listed ones would you rather go with Intel or AMD?

6

Intel every time. You mentioned transcoding and it eliminates the need for a GPU. Those are all nice modern CPU’s.

2 * 12th Gen and 1 * 13th Gen. I am not aware of any issues with Scale - so not sure if it makes much of a difference

Just make sure that the HBA is flashed correctly (right firmware versions, in IT mode) and be certain its an HBA rather than a MegaRaid card (although some MegaRaid cards can be lobotomized to an HBA). I THINK all the ones you mention are all HBA cards.

The 92xx range is really for HDD’s only as they don’t have the guts for many SSD’s. Just use the on board SATA for any SSD’s

1 Like
7

Hey just wanted to come back at you to tell you guys that i now got this Hardware:
X11SPL-F
160GB DDR4 ECC RAM
Intel Xeon Silver 4110
and some SSDs and Harddrives.

for now im just testing around a bit and didnt build in the HBA for now

Are there any recommended settings for the Scrub and SMART test in terms of how often they should run?

1 Like
8

Just do a hardware burn in / stress test for CPU, RAM, storage, etc. Make sure you have enough air flow for all your planned devices.

9

I think way back when (can’t find original forum post), this was a (the?) recommended scrub & smart layout, which I am still using to this day:
Scrub:
image

Smart:
image

As others have stated, good idea to do some hardware tests to make sure everything actually works prior to deployment. Would also strongly suggest burning in the HDDs with some smart tests & a full run of badblocks (followed by more smart tests) if you can wait a few days before getting your data onto them.

…I also can’t find the original guide for using badblocks, but the very short version is to use multiple tmux instances to run - here are my short notes on the process from a few years back

HDD Burn in Guide:
(/dev/sdX on Scale)
1.	Smart Conveyance
•	smartctl –t conveyance /dev/adaX
2.	Short Test
•	smartctl –t short /dev/adaX
3.	Smart Long
•	smartctl –t long /dev/adaX

Read Smart readout (should be generally obvious).

Force badblocks *expect at least 4 days for 8TB drives
1.	Open tmux which will generally allow a new console per disk
•	tmux
2.	run badblocks per disk; following sets –c for multiple blocks tested at once (it won’t go past what drives allows bu may use more ram, give ssome speed benefits over default). –b for blocksize of the disk (4096 >2TB,  don’t use for for <2TB, high volume at like 20tb could be higher)
•	badblocks -c 2048 -b 4096 -wvs /dev/adaX
3.	new tmux windows by “ctrl+b” then push the quotation mark “
4.	repeat as needed
5.	to view in things like Truenas CLI go tmux attach