Setting up Ethereum mining with over 2+GH/s capacity in India
At Jalan Technology Consulting, we believe in the potentials of blockchain technology. Among many, Ethereum has shown a great potential to be a platform to build meaningful applications that can leverage blockchain technology. In this article, we talk about our efforts to setup Ethereum mining operation in India.
To achieve our goal, we setup 16 machines running on Windows 10. Each machine had a Motherboard, CPU, RAM, SSD, PSU, 6 GPU’s attached to motherboard with USB raisers. We also developed a custom rack in Stainless Steel to hold all them together and provide enough air circulation at the same time.
It is well known that Sapphire Rx4xx & Rx5xx are best for Ethereum mining. We were lucky to get hands on few of them. Here is what we ended up using:
Motherboard: AsRock H81 Pro BTC. It has 6 PCI slots to attach 6 GPU’s.
CPU: Intel G1840/G3250. Any basic CPU does as most of the mining is done by GPU’s
RAM: 4GB/8GB DDR3 (4 GB is sufficient)
HardDisk: 64/120 GB SSD.
PSU: 1200 Watt Corsair or Seasonic
GPU: Asus 4GB Rx 470 Strix , Zotac 3GB GTX1060 AMP, Gigabyte 3GB GTX1060 IXOC, Sapphire 4GB RX570 Nitro+ OC, Sapphire 4GB Rx 570 OC PULSE
Once a machine is assembled, we made following configuration changes:
Ensure that system auto logs in when and if system reboots.
Ensure that system never sleeps by changing power options.
Few misc changes such as changing page size setting and optimizing the system for performance.
Select right drivers for mining. For Rx4xx & Rx5xx graphics card, AMD came out with special driver.
Use a mining pool. We ended up using Claymore — https://bitcointalk.org/index.php?topic=1433925.0
Once all the system are running, we observed about 1.7–1.8 GH/s mining capacity. To improve performance, we change settings for each GPU (modding & overclocking). The goal is to lower the core clock speed as low as possible without impacting performance, increase the memory clock speed as high as possible without impacting stability and lower both the core and memory voltage as low as possible without impacting stability.
After we were done, we saw an improvement of 15–20% in hashing rate.
Few challenges we faced during this:
Motherboard would not detect all the GPU’s.
Some time overclocking may results in to system instability : artifacts on the screen, a GPU crash, Windows blue screen.
Controlling temperature and dust (the mining rigs produces a lot of heat).