Worth upgrading? Threadripper ( DDR3->DDR4, NVME) ?

[schoolpost]

I'm contemplating an upgrade of an existing FreeNAS build. ( FreeNAS-11.3-U5 )

I've fallen a little out of the loop with FreeNAS/TrueNAS over the past couple years since building it. So I'm looking for some opinions / feedback on the idea of an upgrade.

The Hardware for my current build is as follows:
-2x Intel Xeon X5675 ( 12c/24t )
-64GB DDR3 ECC
-Supermicro X8DTH-6F
-LSI HBA's ( IT MODE )
-10x 4TB HDD's ( WD RED's ) ( Z3 Array )

As time goes on, I subscribe more and more to the idea of: "Don't mess mess with something that works." and in this case, everything has been working just fine.

Looking forward my current hardware platform is quite limited, no NVME; stuck at DDR3, diminishing performance per watt at time goes on, etc...

However... In the wake of a recent system upgrade of my workstation, I now have access to a fairly nice AMD based platform that I am considering to use as an upgrade for my FreeNAS Build.
-Threadripper 1950x ( 16c/32t )
-Asrock X399 Taichi

Which addresses many of the future proofing concerns I have with my existing build. ( NVME for L2ARC, DDR4 ECC support, modern architecture... )

I don't have a enterprise use case for my NAS, it's mostly a data store / archive that get's accessed infrequently throughout a day. Very rarely is it ever pegged to the max either.

-Is this upgrade worth pursuing? ( I will need to get some addition hardware to make it work, namely ECC DDR4 )
-What is consensus with using threadripper with FreeNAS/TrueNAS? anything to look out for?
I have concerns with power consumption of a threadripper based system, opposite of overclocking I would be looking to run the CPU undervolted or in ECO Mode if the motherboard supports such feature.

Looking for any general feedback that may tip the scale, one way or another.

 

[ChrisRJ]

Well, if so far you did not feel the need to upgrade their likely isn't one. In terms of power consumption, if that is a consideration in your case, I would think about moving to at least one CPU generation newer, plus fewer and bigger HDDs. I have an X9 board since 2020 and feel no need to upgrade.

Future proofing has been a myth, in my view, since at least the late 1980s, when CPU board and passive backplanes were "fashionable". And for at least the last 10 years my impression was that improvements went towards incremental. With your short description of the current situation, I don't see a need to upgrade.

But I admit that having something newer and better is always nice. In the other hand, the hardware will be at the very least need to be 50% more powerful before you even see a noticeable change.

 

[Etorix]

I don't have a enterprise use case for my NAS, it's mostly a data store / archive that get's accessed infrequently throughout a day. Very rarely is it ever pegged to the max either.

For this use case, a 16-core Threadripper looks like massive overkill.

I have concerns with power consumption of a threadripper based system, opposite of overclocking I would be looking to run the CPU undervolted or in ECO Mode if the motherboard supports such feature.

I understand that you already own the motherboard, so you should know its power use.
If low power is the main reason to upgrade (and it would certainly make for a very good reason coming from X8…), I would look to an actual low power platform such as Atom C3000 or Xeon-D, selling the X399 board and Threadripper to offset the costs.

 

[ChrisRJ]

The biggest jump in terms of idle power consumption will be from X8 to X9 (using Supermicro boards as reference). My system (X9 and 8 Seagate Exos 16 TB) uses about 85-90 watts idle with all drives spinning plus two beefy high-pressure fans. So the board (X9SRi-F) and CPU (E5 1620) use 20 watts or less with BIOS settings for power saving.

 

[joeschmuck]

The Hardware for my current build is as follows:
-2x Intel Xeon X5675 ( 12c/24t )
-64GB DDR3 ECC
-Supermicro X8DTH-6F
-LSI HBA's ( IT MODE )
-10x 4TB HDD's ( WD RED's ) ( Z3 Array )

If you do not have any major overhead needs and you want to save a little power, could you remove one of the two CPU's?

As time goes on, I subscribe more and more to the idea of: "Don't mess mess with something that works." and in this case, everything has been working just fine.

Unless there is a real benefit which I don't see at this time, then I would recommend you leave well enough alone.

Find some other purpose for your motherboard and CPU or possibly sell it for a little money to offset the money you just spent upgrading your personal computer.

 

[schoolpost]

When dealing with technology it's easy to get caught up in doing something simply because one can.

I appreciate the feedback, even if it wasn't where I was initially leaning with my decision.

My current FreeNAS build has been working fine for my use case a simple network share / data store; However I've always wanted to explore the possibilities of the more advanced configurations that are possible with L2ARC, setting up applications and VM's; Maybe if anything I was most set on exploring a hardware upgrade out of the desire to upgrade things on the OS/Software side. Feel's right to perform both in tandem, instead of just one or the other.

Also like to repurpose things where I can, I have an extensive knowledge of the history of the hardware being the sole user of it; same can't be said come upgrade time and seeing what is available on eBay, can be kind of hit and miss.

Maybe more impactful would be looking into a better network link, since I built the machine I've always desired to eventually have a 10Gbe link, but found at the time that hardware was expensive and options were limited, how has state of 10Gbe changed in the past couple years? What are some of the go to "inexpensive" options out there? My FreeNAS machine is separated from my workstation by no short distance, however I do have a CAT 6A link between the two.

There a lots of cheap 3rd party NIC's I've seen pop up from Asus and other vendors, but from what I've gathered those either are not supported on FreeNAS or do not perform very well?

 

[Etorix]

Maybe more impactful would be looking into a better network link, since I built the machine I've always desired to eventually have a 10Gbe link, but found at the time that hardware was expensive and options were limited, how has state of 10Gbe changed in the past couple years? What are some of the go to "inexpensive" options out there? My FreeNAS machine is separated from my workstation by no short distance, however I do have a CAT 6A link between the two.

See the "10 GbE primer" in @ChrisRJ's recommended readings.
For a "no short" distance, the best option is to go optical (SFP+).

There a lots of cheap 3rd party NIC's I've seen pop up from Asus and other vendors, but from what I've gathered those either are not supported on FreeNAS or do not perform very well?

Second-hand/refurbished/"clean pull" Chelsio or Solarflare NICs should come even cheaper than consumer cards based Marvell/Aquantia controllers and will indeed come with better support.

 

[joeschmuck]

I've always wanted to explore the possibilities of the more advanced configurations that are possible with L2ARC, setting up applications and VM's;

You have been here for a few years now and have hopefully seen the threads about L2ARC's, mostly when people install them and they do not provide the desired effect. 64GB RAM is probably the minimum amount to have for installing an L2ARC, but do you have specific files that you frequently load? Or in other words, if you are using this as a Movie Server then the files are not loaded often. But if you are accessing a database file frequently then the L2ARC could load it in advance and provide it a little faster. So the use case matters a lot.

As for VM's, I prefer to use a very mature hypervisor such as VMware ESXi, which is free and works great, well for me and for many others. All those cores would come in handy on ESXi. This allows you to let TrueNAS run in a VM and then you could play with other VM's independently.

Just my two cents.