NICE! 
Does your case come with those slick external hard drive holders?? 
I’m just about to take the plunge and am using your system as a reference one. Hence a few last minute questions!
- Any thoughts now you’ve been running this rig a little while? Anything you’d do different?
- What was your logic on running all the 980 pro consumer grade drives, when all the guides say to go for enterprise grade?
- What is the purpose of the plot dump drives?
- Why 6 X 980 Pro Drives? For 16 cores I thought that max disk space needed would be 16 x 239GiB = 3,824 GiB. You’ve 3 X this requirement.
2 Likes
Beefier water cooler, mainly, and install all the 120mm intake fans (2 on bottom, 1 on side). Other than that AMD+ASUS is {chef's kiss} such an amazing fit for this!! I actually ordered another one and it’ll arrive tomorrow. I am excited! I’ve never been particularly excited about having “m0ar c0res” until now* because I have actual need for a LOT OF CORES! So AMD to the rescue!
I’ve done the research and Samsung drives last forever. Check out the endurance studies people did on real hardware, writing endless terabytes until drives actually died. And my testing shows that CPU speed is a major factor, once disk is “fast enough” (good NVME) you really don’t squeeze that much extra juice…
Yes, I may have gone a little overboard on the drives lol. You could use fewer drives. But hey, these will last even longer now, right, since they’re being used less… 
* plus the way Qualcomm uses “lots of crappy slow cores” to hide their absolutely ABYSMAL mobile SoC single threaded perf. A brand new Android device is about as fast as an iPhone 7, and this performance gap is gonna be with us for another decade… my god… if I could burn down the Qualcomm factory, and plow the ground underneath it with salt so nothing will ever grow there again, I would…
6 Likes
Cheers @codinghorror , i’ve basically replicated the system, but with 4 plotting drives. Should be able to build it up on Thursday!
1 Like
I love the disk party idea!
You neee to place them a.disco ball in the center!!
1 Like
heh thanks! makes it extra sweet to find 2 CHIA an hour after posting this! SHEEEEESSSHHHHH
6 Likes
Did you run out of orange zip ties OR were the blue ones intentional??
2 Likes
Using a NAS server running FreeNAS for farming. It’s also used as a Plex server and Zoneminder rig. PC is currently plotting. Mostly 4TB drives but 4x12 on the way. I have a 5950x in the mail, thinking about going the same route as @codinghorror with that chip.
Nice to see you here Jeff- I’ve been a big fan for a long time! 
What purpose do the M.2 drives serve in the Chia process?
Edit: Chia process, not Chia farming process.
2 Likes
They are used for plotting, not farming 
2 Likes
Ah, got it 
.
So is there a difference between the 2tb M.2 drives for plotting and the 12tb plot dump drives? Or do they serve the same purpose?
Plotting is very IO / CPU intensive and should be done on fastest drives / CPUs.
Farming is very low IO and should be done on the cheapest (but reliable) storage you can get your hands on.
3 Likes
that’s some nice setup !!!
Looks nice. Are those a raid array or simply hdd storage?
May have been the wrong choice, but it’s set up as a storage pool, 1 big drive, “thick volume” with no redundancy/protection.
This is my precious farming rig:
8TB of the most pristine plots the algorithm can generate 
Rejoice!!
2 Likes
Thats good. So if one drive fails you still have the rest? My understanding is if its set to Raid 0 if any drive fails you loose the lot?
Yeah, I might lose everything if a drive dies… fingers crossed that doesn’t happen. If I actually win some coins before I fill these up, I’ll get a different storage system and move everything, and then run individual drives.
Losing a drive in RAID0 you are gonna be SOL. If you span a volume, you can do surgery with mdadm and a few tools and get the volume back online, minus the data from the failed drive and any partial data that spanned across disks at either end of the failed drive. You might also consider mergerfs if you just want to deal with one big volume. It doesn’t solve the “lost space” due to badly packed plots, but each volume is effectively independent but still appears as one larger block of storage.