Post your mining rigs

[funkywizard]

I'll start.

Shown is the hot aisle for most of my chia farm. Servers showing 12x drive bays are 36 bay servers. I have 3 more 36 bay servers hosted at a real datacenter for the moment. What is shown is in an evaporatively cooled warehouse.

Everything that's not in a 36 bay here is a plotter. Able to plot 150TB/day. Currently 2PB plotted, should be able to hit at least 3.5PB with what I have on hand.

 

[Rand__]

 

[tinfoil3d]

How much does that give?

 

[sovking]

How much is the ROI (Return of Investment) ?

 

[funkywizard]

How much is the ROI (Return of Investment) ?

Theoretically or actually?

Most of the hardware was "sitting around" -- bought before the massive chia-induced price increases, and needed for general customer use -- otherwise the whole thing would be way too risky.

For the last week or two, the "theoretical" earnings has stayed flat at $1k / day despite ramping our plotting at a tremendous rate -- total network growth is 10% per day so if you're not plotting faster than that your estimating earnings will drop every day.

So far for maybe 120k in hardware it's earned 5k, ignoring power costs and the value of people's time. About half the hardware value is the plotters which should be idle and available to rent to customers again in a couple weeks.

When we first started plotting, the estimated revenue of this hardware (if it had been fully plotted) was 9k / day, but 2 - 3 weeks in we've only made 5k total so... yeah.

Hoping to make maybe 50k over the next few months, then when chia revenue drops below 12 months ROI we can rent the hardware out to real customers again. Right now the only people who want this stuff would be mining with it, so we may as well do that ourselves instead.

 

[Jaket]

I don't see any Chia pets in there kind of disappointed.

It will be interesting to see what you end up making on everything after you quit doing it.
At least you had most of the equipment ahead of the boom otherwise more people here would be a little upset with you.

Makes sense not renting all of that out to miners if you can make more yourself doing the mining.
Good luck on the Chia pet mining!

 

[amalurk]

Any estimate KW/h per day you are using right now.

 

[Jaket]

Any estimate KW/h per day you are using right now.

I'd say a pretty good chunk between the three racks.

 

[funkywizard]

Any estimate KW/h per day you are using right now.

ballpark 12 - 15kw per hour.

 

[bash]

Not that far from you, we are on central and osborn in PHX. Only plotting around 25TB/day with 300TB plotted and 600TB installed in the 16U(4x4U) of servers we have on our single rack. Probably have enough space left to reach 1.3PB. The only real cost we have are the nvme drives we bought a couple of weeks back before prices went insane. Everything else is equipment we already had laying around. Some of the plotting servers are Xeon V2's lol.

 

[funkywizard]

Not that far from you, we are on central and osborn in PHX. Only plotting around 25TB/day with 300TB plotted and 600TB installed in the 16U(4x4U) of servers we have on our single rack. Probably have enough space left to reach 1.3PB. The only real cost we have are the nvme drives we bought a couple of weeks back before prices went insane. Everything else is equipment we already had laying around. Some of the plotting servers are Xeon V2's lol.

E5v2 should plot just fine depending how much disk and i/o you've got to work with.

Luckily I bought a good 100x 3.84tb nvme and 70x 1.6tb nvme, just to have stock on hand, a few weeks before everything went nuts. It had been getting harder to get these especially at good prices, and we had been pretty much down to nothing. Really pleased now we did so.

To make the most out of dual E5v4 I found that 3x 3.84tb nvme worked well (each rated for 3GB/s read and write) -- able to do nearly 7TB/day per server.

4x 1.6tb Intel P3605 nvme in the same server didn't have enough space for parellelization, nor enough speed to keep the threads busy, putting cpu utilization at around 50% on dual 2695v4.

 

[bash]

On the 4TB nvme's how many concurrent plots per and what kind of times are you getting with those. I have 2 x 4TB 4510's that start to hit 12+ hour plot times when ramped up to 14+ concurrent on each.

 

[funkywizard]

On the 4TB nvme's how many concurrent plots per and what kind of times are you getting with those. I have 2 x 4TB 4510's that start to hit 12+ hour plot times when ramped up to 14+ concurrent on each.

We budget 400gb storage per simultaneous plot just to make absolutely sure we dont run out of space, so I guess it would be 9 per, putting us at probably 27 simultaneous plotters on a 3x 3.84tb nvme / dual e5-2680v4 or 2695v4 server. Not sure the plot times but it can crank out just shy of 7tb / day, and keeps the cpus very busy most of the time.

With 18 core 2695 / 36 core / 72 thread per server -- this is reasonable concurrency. We set max threads per plotter to like, 8, so hypothetically we could use over 200 threads, but realistically the software struggles to use more than 4 threads even during the part of plotting that has the best multithreading.

Still, 4 threads in use times 27 is 108 maximum threads tied up. A lot of plotters will be single threaded at any given time as some parts of plotting are single threaded. This does leave spare cpu for those plotters who are able to use ~4 cores. So overall it keeps the CPUs busy.

Dropping form 2695v4 to 2680v4 we saw no meaningful difference in TB plotted per day. Multipyling cores times mhz, the 95 should be at best 15% faster than the 80v4 despite having 18 cores vs 14. The faster cores in the 80v4 are much better for the single threaded part of plotting, so, much of this 15% hypothetical gain from 95v4 is lost to the fact that, fewer faster cores means less parellelism is necessary to keep those cores busy.

So especially if you are limited on disk space or ram, faster cores will give you a little boost vs lots of cores. Though ideally you want it all -- tons of space on absurdly fast nvme drives connected to high clock speed high core count CPUs. Multiply by as many servers as you can get : P

 

[Marsh]

If you are cranking out 7TB from a dual CPU highend server.
I think I could stop tweaking my plotter.

4 machines , each machine has same config , except size of Fusion SSD
Single E5-4667 v3 ( 16c/32t) , all-core-turbo to 3.0ghz
3 x 1.2TB or 1.6TB Fusion-IO iodrive2 ( older pcie 2.0 model)
4 thread , 8gb ram per plot,
3 plot per Fusion Card , 16 max jobs
Daily plot output is around 32-38 plot per day , average 3.8TB per day per machine.

I brought all the parts long time ago. CPU was $85 each , Fusion SSD were dirt cheap.

For now, I am saving my box of Intel Pxxxx nvme SSD,
because not sure there is a huge advantage over older Fusion SSD.
Actually, I am looking forward to wear out the Fusion SSD ( 17PB write each )

 

[BobTB]

So far for maybe 120k in hardware it's earned 5k, ignoring power costs and the value of people's time.

So you got like some 8 XCH from this monstrosity of a machinery? OK

And as the netspace grows, you will get less and less... I believe this Chia thing is just waste of equipment and time... sadly... is sounded great.. but now with everyone plotting and waiting to "get rich" with their 2Tb drives and making netspace grow as crazy... its just nothing anyone can do...

I am sad to see some people waste 10k or so for the equipment, and they will have to duplicate the capacity every week or so, to just keep getting their small rewards, even if official pools become a reality, or it will dwindle down to nothing in months.

You are good only because you had the 120k in hardware idle, and it can just earn something in the meantime

 

[funkywizard]

So you got like some 8 XCH from this monstrosity of a machinery? OK

And as the netspace grows, you will get less and less... I believe this Chia thing is just waste of equipment and time... sadly... is sounded great.. but now with everyone plotting and waiting to "get rich" with their 2Tb drives and making netspace grow as crazy... its just nothing anyone can do...

I am sad to see some people waste 10k or so for the equipment, and they will have to duplicate the capacity every week or so, to just keep getting their small rewards, even if official pools become a reality, or it will dwindle down to nothing in months.

You are good only because you had the 120k in hardware idle, and it can just earn something in the meantime

6 XCH lol.

My estimated block find time is 0.75 blocks per day but I sure as heck are not finding that many.

 

[Iaroslav]

@funkywizard what a setup!
I'm on 1500 plots, 4 coins. Golds and some v3-v4 machines with 6-8TB nvme. Have 500 more TBs to fill, didn't buy much for mining, I agree it's too risky and not a fair play with guys like... Like us)

 

[msg7086]

My farming machine that's still WIP. I use a plotting service and I'm downloading the plots home. I have a Supermicro 2U that is used to fill the drives, and once filled the drives are moved to this chassis for farming. I made custom SATA power cables to connect all the drives. I chose 16 AWG wires so they can handle at least 20 amps, that's about 20 drives per cable sustained, or 8-10 drives per cable at peak. Each will be handling 4-5 drives for load balancing.

Specs:
Motherboard: B85M Pro4 (old rig)
CPU: i7-4770 (old rig)
HBA: H310 (old rig)
Chassis: Meshify 2 no glass window ($142) with 4 drive trays ($28)
PSU: Seasonic FOCUS PX-850 Platinum ($135)
OS drive: ADATA SU800 1TB (old rig)
HDD: 11x Exos 16TB ($3280)
UPS: APC BX1500M ($165, arriving today)

 

[Rychek]

This is my "hobby farm":

Plotter Specs (and Etherium miner):
Motherboard: Asrock Rack RomeD8-2T

Chassis: Chenbro RM42300-F
CPU: EPYC 7272
RAM Kingston DDR4 3200 Registered ECC 128 GB
OS: Windows 10
OS drive: Kingston DC1000B
HDD: Assortment of drives ranging from 5400 RPM HDDs to SATA SSDs an NVMe drive and one 1.2 TB Intel U.2 drive.
UPS: Cyberpower OR 1500L-CDRM1U
GPU1: GTX 1070 TI
GPU2: RTX 3060 (full speed ETH mining)

NAS Harvester/Full Node

NAS Specs (Farmer is a Win10 VM):
Motherboard: SUPERMICRO MBD-X9SCM-F-O
CPU: Intel Xeon E3-1230 V2
Chassis: Norco RPC-4220
OS drive: Some Sata SSD (can't remember at the moment)
HDD: Assortment of drives ranging from 5400 RPM and 7200 RPM HDDs (84 TB total. 43 TB for personal data/41 TB for chia farming).
UPS: Cyberpower OR700

The wood rack is home made, but not by me.

 

[bash]

Need to organize rack