In recent years, the price of bitcoin has continued to decline, but
the computing power has increased significantly. No matter what the
willingness behind it is, this environment has brought great
pressure to bitcoin miners.
In order to reduce the proportion of electricity expenditure,
persistent miners have turned their attention to a new generation
of 28nm process mining machines. Today, we want to evaluate the
28nm fourth generation miner from Avalon``
Our Avalon 4 is strictly an engineering prototype, so all test
results are for reference only.
The size of the miner body is 395 mm x 160 mm x 165 mm, which is
about one-third of the size of the standard ATX chassis. The net
weight is 8.5Kg. The weight is mainly derived from aluminum fins.
The mining machine adopts honeycomb air inlet and outlet, which can
not only ensure the smooth flow of air, but also obtain better
electromagnetic shielding effect. There are also power interface,
data line interface and LED indicator at the air inlet.
The diameter of the cooling fan is 12cm, which is manufactured by
Nippon Power Co., Ltd. we will test the noise level of this fan
later.
There are two force calculation boards in the mining machine, a
total of 40 a3222 force calculation chips, and the nominal force of
a single chip is 25g, a total of 1t. It was troublesome to
dismantle the machine, so we only dismantled the fan and the
controller of the miner. This controller was developed by Avalon
and put online as an open source, which was adopted by many mining
machine manufacturers.
There is not much to be picky about the hardware design of this
miner. Next, we will enter the installation and configuration
stage.
The installation is very simple. Because the interfaces are
different, there is basically no possibility of wrong insertion.
The only thing to note is that the four 8-pin power supply
interfaces on the miner are also compatible with the 6-pin power
supply interface. Power supply we use the Great Wall Julong 1250W
gold medal power supply, with a conversion efficiency of more than
90%. After installation, power on and prepare to start
configuration.
Avalon officially provides a mining system modified based on
openwrt and sets a static IP, which belongs to a relatively less
common network segment, so we need to set it first.
Open the setting interface of the router through the browser, find
a place similar to "Intranet IP configuration" or "LAN port IP
configuration", modify the IP address to 192.168.0.1, save it, and
the router will prompt to restart. After restarting, you can
generally find the raspberry pie in the router interface. Here we
can see that the IP address of raspberry pie is 192.168.0.100. We
can access this IP address through the browser to enter the
management interface of the mining system on raspberry pie. If you
want to change the router's intranet IP address back to the
original, you need to change XXX to DHCP, and then go to the router
to change the intranet IP address back to the original.
Back to the mining system management interface, we can see some
basic information of the mining machine and modify the mining
configuration of the mining machine, which can basically meet the
needs of ordinary users.
However, in order to facilitate the test and observation, I decided
to use version 4.9 of cgminer, that is, the latest version, for the
test. The system is an old friend arch Linux. We have already made
a complete introduction to the installation of arch Linux and
cgminer in the previous video on evaluating Avalon nano, which will
not be repeated here( Video here, using Pacman - SYU)
First run through the official default settings. Here we can see
that there are three values in the default frequency. It is said
that it can be used to stabilize computing power and optimize power
consumption. The initial value of the voltage is 0.8125v, but in
the test, we found that the system will automatically adjust the
voltage value to reduce power consumption. In our test, about 30
minutes after mining, the voltage automatically decreased to
0.7625v, and the power decreased to about 585w. During this period,
the calculation force fluctuates up and down at 915g. At this time,
the energy consumption ratio is 0.64w/g.
It is gratifying that the miner power at this time is much lower
than the nominal power, but the calculation performance is not very
satisfactory, although it has reached the official nominal standard
of 10% fluctuation of 1t. So we decided to try overclocking.
After many tests, the parameters of frequency 495:435:420 and
voltage 0.825v are matched, and satisfactory results are obtained.
After stabilization, the voltage automatically drops to 0.7875. At
this time, the computing power reaches 1.03t, and the power
consumption is 693w, which is close to the official 700W. At the
same time, the average computing power also slightly exceeds 1t.