Is there a better solution instead of building this from scratch?
I recently purchased a Skylake setup (see sig.). I cheaped out on RAM, starting at 16gb. That was not nearly enough for smooth operation. Still I cannot validate performance gains from my jump from 16->48gb (memtest still in action for hardware validation). With this said, I've been a bit humbled by some posts discussing the 'future strategy' of <any> FreeNAS build. I'll share some of my thoughts on the topic.
Too many people seem to build a small form factor box, cram in a bunch of drives, and call it the day. Then the day comes when that build no longer is sufficient and questions emerge on what to do next. Some add their older box as a 2nd line backup solution of the more important files. Others... ..simply remake the same mistake once again by purchasing a similar setup (yet with higher performance parts aka a bit more ram, a bit more cpu and a bit juicier drives with a bit larger capacity).
Since you are about to enter this scenario for real, you still have the option to pose the question to yourself <what do I want from this hardware, when it is time to upgrade next?>. My own goal was to slowly build a monster as I could afford parts.
To put that reasoning into context, if your current AMD build would've supported 32gb ram, the likely hood of you swapping out the motherboard/cpu/ram for <ram reasons> at this point in time - would be sitting in a completely different light, wouldn't it?
My point is to setup a <strategy> of what should stay and what should go on your upcoming upgrades.
Some points that I considered before my own purchase.
- HBA - I took the hit and got a 4x SFF8087 16x drive 'out of the box' compatible card. Should set me up for at least one additional upgrade (adding 7x drives)
- CPU - I went the chicken route. I got the middle road of an i3-6100. It is hugely overpowered for what I'm doing. I could've gone for the G4400..
- Case - Getting a case that allows for at least one 'added vdev' to the main zpool is a must in my opinion.
- RAM - I discarded the X10 series due to the cap at 32gb. Only got myself 16gb initially due to budget constraints and easiness for future upgrades.
- vdev size - I decided 7x drives would be enough 'value for the cash up front' and decently upgradeable (7drives are steep to purchase at once but better than 12x..)
- PSU - Better slightly overpowered at the beginning than in dire need for replacing once a vdev has been added.
Then I became interested in ESXi and virtalization of my utility servers... number of cpu cores, PCIe slots and ram capacity have suddenly become my immediate points of restriction.....
So if I could've redone my choices here?
I'm completely happy with PSU/HBA/vdev
On the other hand, the parts which are limiting at this point are first lack of CPU cores. Yes, an Xeon 1230 could still be purchased. Beyond that...
I would've taken the hit up to E5... Here is what a comparison looks like with my local prices:
X11SSL-F (Max 64gb ram)
Intel Xeon E3-1230v5 3.4/3.8GHz QuadCore HT, 8MB,
2x16gb ECC UDIMM
7728SEK = 940USD
X10SRL-F (Max 512gb ram)
Intel Xeon E5-1620v3, 4-Core, 3.5GHz 10MB cache
2x16gb ECC REG
9173SEK = 1100USD
At least - that means less likelihood to be stuck in a spot forcing RAM to be the reason for upgrading.
The CPU's are on sockets that does not seem to change quite as fast as the consumer grade stuff I'm used to. That holds for future possibilites to upgrade (ie, in 3 years when current monster cpu's are hitting the market 2nd hand).. The 20 cores cpu are ...nonomonom ...just out there within reach.
To wrap this rant up:
It is difficult to purchase a system that fits when the use case may change. Yet, there are still some pieces of hardware that will be valid if plans change for use case and/or hardware requirements change for the 'more intensive'.
Hope this helped you somewhere in reasoning :)
Cheers,