Bitcoin Forum

Hi there

It's just a theoretical question since I don't have the money to do it ;-) However if you would built bigger mining rigs (let's say 30 - 60 cards), is it better to run small systems with 6-8 cards (Asus Prime z270, MSI Z170A, Gigabyte GA-H110 etc) or 13 gpu systems with a ASrock H110 Pro / Asus B250? You're a probably more flexible with smaller rigs regarding cards and psu but need more time to maintain the system. So what would be overall better would you say?

use search bro, there was thread on this 2 days ago...

GPU's are not individual, if one gpu crashes all crashes, if each gpu was a single restart if error occurs entity then you could build mining rigs with many gpu's then the only real upgrade would be adding a very good cpu and ram to withstand their load. It's impractical right now till they make gpus a hotplug product.

I own the MO B250 mining expert it contains 19 GPU lines and as of now it works with the celeron processor for my mining rig. Even 12 cards GPU MO also a fine to start a small mining rig. Please ensure that you are adding the Rx series cards such as Rx 470/580 both 4GB and 8GB are good.

This will produce almost 22 MH for mining eth with the 470 cards and 580 may give upto 30 MH.


I see above some suggested to have the best RAM and CPU. I wanna say you RAM need to have the DDR4 and above but processor does not matter to mine bro. Have that in your mind first.

There is an obvious advantage and disadvantage on each setup, you can decide this once you have already your budgets, you can go for mining motherboards if you are comfortable and use to it, as some users encountered challenges getting to work all PCIe slots and GPU's, another things is the centralized failure compared to boards with 6 GPUs. also the resale value, Z series boards are hot on gamers also.

there isn't any MB that can handle 60 gpu...

Please stop posting garbage. Not only is it painful to read but also incorrect.

no but 13 (asrock h110 pro) or 19 (asus b250), so it makes a big difference if you got 1 rig or 3 with the same gpu power.

It depends. 

Grouping as many GPUs as possible onto a single PC helps with maintenance, cost and energy (powering 1 MB/CPU/HDD instead of 60).

Taking the other extreme, if each PC only used one GPU, then one or two crashing isn't going to significantly impact the hash rate of your farm.  Plus, you can service them separately.

I average about 3-4 GPU per rig, mainly to exploit prices of mid-level hardware.  I want to fill them into cabinets but sourcing affordable rackmount cases for GPUs has not been easy. 

First, it's more common to refer to one "rig" as one CPU/mobo/RAM set/etc. and all associated cards hosted on that individual system. I'm going to interpret your 30-60 number as the overall "size" of your theoretical mining operation, regardless of how many rigs are used to host that number of graphics cards.


The "best" distribution is a matter of available hardware (not everyone can buy a 19-slot board and 6x P106 cards right now) and personal preference.

Running more GPUs on each rig may give you a decent advantage in price and operating costs. Keep in mind that riser costs almost always scale exactly with the number of cards you have, and PSU costs do to a similar extent (except for extremes, such as 1-2 cards per rig, since each rig always requires a minimum of 1 PSU), so those will not be the main areas for savings.

However, in costs for the motherboard, memory, CPU, and storage device, you can save. (The motherboard slightly less so, since motherboards that can support more GPUs are generally more expensive.)


On the other hand, having more cards per rig can start to put too many eggs in one basket. If any part of the rig "base" fails, for example, a CPU, then you have more cards down until you can repair it (whether that requires buying a new replacement board, or just finding the time to swap out for an available spare, although if you have spares readily available for all rig base components, we'd have to ask why you don't make some extra rigs, since purchasing spares defeats the point of saving on the initial investment, and not using them only saves a negligible amount of power, compared to what a 30-60 card operation would require).

Metroid's post had a little truth to it. If one card throws an error, it can bring down the mining software/driver/OS/etc. This definitely isn't always the case, but it can happen. (The most common example is a dual-PSU rig, where the secondary PSU (which only runs some of the GPUs) fails and manages to down the entire rig.)


Setup headaches are also something to consider. With a good motherboard and a uniform set of GPUs, I generally haven't had scaling problems between 6-12 GPUs on Ubuntu. In both cases, setting up drivers has been an all/nothing thing, rather than an issue of getting all 12 GPUs recognized. However, others have reported much bigger headaches, and I can definitely imagine the 19-slot Asus board being a pain, especially since it still requires 6x P106 "mining-oriented" cards in order to make full use of the board's potential.

In general, as you add more GPUs per rig, things get harder at different rates. For example, going from 5 to 6 might not be as hard of a jump as getting from 13 to 14. (This also depends on the motherboards you use, setup will generally get more difficult once you try exceeding the number of PCIe slots by throwing in PCIe splitters and/or M.2 adapters, but YMMV all the same.)

How you weight/prioritize the above factors is a matter of personal preference.

The card requirements to actually use all of the slots on some boards is a dealbreaker for me personally (P104/106). With more cards comes less stability in many cases as well, which is why I just prefer to have a few more motherboards with less cards rather than a few with a lot. In the end I usually just go for deals on boards that I personally trust for stability and have positive reviews. Plenty of people run rigs on boards that support less than 10 cards and they run just fine.

I agree that processor does not matter much.  But if I can spend $10usd more to get a Pentium over a Celeron, I'm going to do it.  I started all of my 6 gpu rigs with 4gb ddr4 and they ran just fine most of the time ... I've since upgraded them all to 8gb ddr4, I don't notice a lot of difference.  I think I see a bit of a difference when I remote into them.  The system as a whole is more responsive to command.  Those little upgrades only cost me a couple days of mining, so I do it.

Majority of my issues have been gpu driver crashes.  Most of that came from core and mem overclocking a little too much on a specific gpu.  Causes the entire rig to crash.  Even if they are the same brand and model, they won't all be the same as far as how much you can overclock.

My solution to that is a switched pdu.  Although it is not very necessary, if your gpu's are clocked appropriately, it gives me piece of mind.  I can shut down the power to one rig and then turn it back on from my computer at work, my phone, with a laptop from anywhere that I have a connection to the internet.

All above is based on Win10 OS.