Bitcoin Mining Operations Management Service

[Vladimir]

 _

[Cryptoman]

You:
- get hardware with as many GPU's as possible (ati 5970 please).

I'm confused as to how this works. First, are you providing the hardware, and if so, do WE access it remotely, or do you ship it to us?

You supply the hardware, vladimir supplies the expertise and remote monitoring capability.  Since he's recommending 5970s, you're looking at nearly 1000 USD by the time you put together a complete system with a quality power supply.  Mining isn't cheap any more.

[Stephen Gornick]

This is an awesome service!

What would the components list look like for such a system?

I created a typical configuration list on the Wiki for this.

Hopefully from such a parts list an individual willing to spend the money can order the components, and be mining by the following week, for instance.
  https://en.bitcoin.it/wiki/Mining_rig#Typical_Configuration

[edited]

[brocktice]

Currently an online service is in development. It will soon allow my customers to monitor performance of their mining cluster 24/7. Watch this thread for updates. Thank You.

This is interesting. Let me briefly describe what I have:

1) 3x 5970s, water cooled in a box with a Phenom II x3
2) My daemon.py, a custom front-end for m0mchil's miner designed for running multiple gpus on a linux box. Github repo is here: https://github.com/brocktice/poclbm. You may want to use this. It can be configured to email you when a block is found, for example. Let me know if you have questions.
3) A script that monitors GPU temps, chassis temps, and (for my water cooling) pump RPM, and will shut down the box if thresholds are exceeded
4) A script that checks every few minutes to make sure all miner processes are running, and emails me if not

I personally tweaked my GPU clocks and voltages. Currently I am running at 925-975 MHz (the individual cores vary) at 1.1625V, temps 58-68C.

One thing to watch out for is that when attempting to overclock the cores and tune them you may lock them up, requiring a reboot. If the machine doesn't come back up properly remotely you will have to involve your customers.

I've been thinking about setting up a simple reporting web server to show the temps and hashrates. What are you going to be offering in terms of web monitoring?

[barbarousrelic]

Just out of curiosity, how many hash/sec do you get, ballpark, on these miners with 1 or 3 ATI 5970s?

[Natanael]

Hi Brocktice,

  All valid points,  I'll have a look into your code, though python is not my cup of tea, I prefer erlang or perl. In terms of monitoring there will be real time charting of hash/sec and temperature for every core.

One thing to watch out for is that when attempting to overclock the cores and tune them you may lock them up, requiring a reboot. If the machine doesn't come back up properly remotely you will have to involve your customers.

Sure, the system  must be very resilient with ability to automatically recover in most scenarios, via reboot too if needed and as a last resort which should never happen email to a customer with manual powercycle request.

May I suggest making an Android app (or adapting an existing one) for the monitoring? A widget on your home screen that shows stats like average khash/sec last 10 minutes, temperature, uptime, uptime percentage, current system status, etc... That would be really useful. Too bad I can't afford setting up a rig!

[nelisky]

I personally tweaked my GPU clocks and voltages. Currently I am running at 925-975 MHz (the individual cores vary) at 1.1625V, temps 58-68C.

Can you tweak the voltage in linux? Please do share!

[Cdecker]

Hi Brocktice,

  All valid points,  I'll have a look into your code, though python is not my cup of tea, I prefer erlang or perl. In terms of monitoring there will be real time charting of hash/sec and temperature for every core.

One thing to watch out for is that when attempting to overclock the cores and tune them you may lock them up, requiring a reboot. If the machine doesn't come back up properly remotely you will have to involve your customers.

Sure, the system  must be very resilient with ability to automatically recover in most scenarios, via reboot too if needed and as a last resort which should never happen email to a customer with manual powercycle request.

May I suggest making an Android app (or adapting an existing one) for the monitoring? A widget on your home screen that shows stats like average khash/sec last 10 minutes, temperature, uptime, uptime percentage, current system status, etc... That would be really useful. Too bad I can't afford setting up a rig!

I managed to get everything I needed by adding JMX capabilities to the DiabloMiner and monitoring it with munin. That also works nicely to monitor temperatures and other stuff

[brocktice]

I personally tweaked my GPU clocks and voltages. Currently I am running at 925-975 MHz (the individual cores vary) at 1.1625V, temps 58-68C.

Can you tweak the voltage in linux? Please do share!

You have to flash the card BIOSes. I had to use ATIFLASH on a bootable dos USB stick, and I edit the BIOSes in RBE in a Windows XP vmware image.

[nelisky]

I personally tweaked my GPU clocks and voltages. Currently I am running at 925-975 MHz (the individual cores vary) at 1.1625V, temps 58-68C.

Can you tweak the voltage in linux? Please do share!

You have to flash the card BIOSes. I had to use ATIFLASH on a bootable dos USB stick, and I edit the BIOSes in RBE in a Windows XP vmware image.

Ahh, shucks. I can flash the cards, but not remotely, which is why I thought being able to do so using linux would rock. Thanks for sharing, though!