How much RAM I really need for eth mining with 19 GPUs ? (5700 XT 8GB)

Practically all mining tutorials boast that only 4 GB is enough ram for mining. But phoenixminer in hiveOS wont mine and gives "Allocating buffers failed with: clCreateBuffer (-4) " error.

My goal is to mine with 19 GPu with my B250 motherboard and my friend says (he is mining with 19 GPU with same mobo) that I need 16 GB RAM in order to mine efficiently.

could someone verify how much RAM I really need for 19 GPU ethereum mining?

For setup with 19 GPU 8Gb will be recommended
But I’m not sure that motherboard can even start (power on) with 19 AMD GPU on B250. Maybe BIOS update will be needed

I’m runnin 13 GPUs in Asrock H110 Pro BTC+ motherboard. My RAM is 4 GB and only 1.5 GB is used, 2.5 GB is free. probably 4 GB will be enough for 19 GPU. But no harm if you use 8 GB.

Yea, I’m not getting more than 14 GPU’s to work on my B250 mining expert, and Have 2.3Gb free ram so I have similar experience as you.

Have you been satisfied on myös Asrock H110 Pro BTC+ ? What benefits does it have over mining expert? Does it display slot detection somehow? Im thinking to build second rig but no sure which mobo to pick

Yes, I’m satisfied with Asrock H110 Pro BTC+ even if it’s not easy to handle. It took a time for a while to understand how to keep it stable. But once you configured everything correctly and set your rig prorperly (enough power supplies, distribution of power cables, stable internet connection, a good SSD) it works quite stable.

What i did not like about it:

  • it requires continuously monitor connection if you use more than 9 GPU. If it restarts for any reason(power cut, watchdog etc), and if it’s not connected to a monitor it does not come up and in BIOS, CSM status changes to enabled. When CSM enabled, it’s not possible to run more than 9 GPU. You have to configure BIOS again. For this reason I had to buy a second hand monitor (for $10, its screen broken) just to keep monitor connection alive. I use monitor in stand-by state. It works this way.

Statistics from last 24 hours. 13xGPU
It earns 0.04 ETH each day.
Each GPU brings $1.25 everyday by today’s ETH price.

What are your BIOS settings? I can’t run stable at 10 GPU

HiveOS version 171
4 sapphire pulse 5700 XT
6 gigabyte 5700 XT (5 non OC, 1 is used with OC in model name)

PCIE set to auto
CSM disabled

HivesOS: 5.6.0-200909-hiveos-beta

2 x Sapphire 5700 XT Nitro+
11 x Gigabyte 5700 XT
(No mod. All GPUs are with stock rom)
BIOS:

OC Tweaker\Voltage Conf:
PCH VoH: 1200 mV
VCCIO: 1200 mV

Advanced\Chipset Conf:
Above 4GB mmio: Enabled
VT-d: Disabled
All PCIE link speeds: Gen2
All ASPM Supports: Disabled
IOAPIC 24-119: Enabled
IGPU: Enabled
HD Audio: Disbaled
Reset Power Loss: power ON

Boot\CSM Disabled

Which version HiveOS do you use?

So you have a monitor always connected to GPU that’s connected to PCIE x16 slot?

0.6-157@200908

Yes. Actually it works without monitor. But when restarted, if there is no connected monitor, it does not boot.

Do you think my USB 32gb is the issue of not running stable, you said you run on SSD? What size? I’m running into out of memory error kill process and says my total-vm is 18GB

Don’t use USB!
It was my mistake at the beginning too, that I was trying to run it with USB.
USB is not fast enough to run 13 GPUs work. OS will be overloaded with queued jobs, some GPUs will go offline and Load Average will increase very fastly and rig will restart frequently because of high LA.

Definetly use SSD!
I use 120 GB SSD.

And another issue is internet connection. This was my another mistake at the beginning.
I was using wifi connection and then I realized it was not good. I connected rig to router with ethernet cable. That increased stability but not enough. Because there were too much devices at home using internet > 1 desktop, 3 laptop, 4 mobile phone, 1 smart tv, 1 tablet. So I decided to do some client limitations. In router, I divided traffic in two parts. One part was dedicated to rig. All other devices can use only other half part of bandwidth. Thus rig has a constant and undisturbed connection speed. After that arrangement, reboots because of high LA are ceased.

My recommendation is:
1- Use SSD
2- Don’t use wifi connection
3- Provide constant internet connection to your rig

Which make and model SSD did you use? Thanks for your help!

What are your OC settings? Here are mine.

Don’t use USB!
It was my mistake at the beginning too, that I was trying to run it with USB.
USB is not fast enough to run 13 GPUs work. OS will be overloaded with queued jobs, some GPUs will go offline and Load Average will increase very fastly and rig will restart frequently because of high LA.

Definetly use SSD!
I use 120 GB SSD.

Im using from USB3.0 without any problems; its stable for weeks. How can I create image from my USB drive and transfer it to SSD ? I have made some tweaks and need to use this particular “build” to SSD…

How many GPU you have on there and what’s your BIOS settings (Asrock H110 right?)

No, I have B250 mining expert. How do I create image from my USB disk?

I cannot get the white pci-e slots to work…any idea?