This is Part 2 of the Mine Ravencoin (AMD) guide – To see the first part of this guide you can click here for Part 1.
We left off on Part 1 with just setting up our aceneun sgminer -k x16r mining software and doing some initial test runs, after going through some common setup problems.
In this part of the guide I will go over further testing the aceneun sgminer -k x16r boundaries, as well as testing another AMD sgminer fork.
In the image above we can see the miner display after 12 hours of run-time. We can also see the hashrate is actually somewhat lower than out 1 hour results. So getting an accurate display of the actual hashrate over-time is proving to be a bit difficult.
As a further hindrance to my testing, it seems MiningPanda pool has decided to remove the individual miner hashrate graph on the wallet stats page. It still displays the point in time hashrate, but the graph was useful to see how it acted over time. I will need to look into this and perhaps move to another pool if I can find one that displays the information in a graphic format.
The one thing I forgot to mention in part one was the stats of the hardware. As I mentioned, these rigs were setup for Ethereum mining, so all the cards have the BIOS mods used with Ethereum. In addition the test rig is using AMD drivers version 17.12.2 as shown below:
Radeon Software Version – 17.12.2
Radeon Software Edition – Adrenalin
Graphics Chipset – Radeon (TM) RX 470 Graphics
GPU Memory Size – 4096 MB
GPU Memory Type – GDDR5
Core Clock – 1242 MHz
Windows Version – Windows 10 (64 bit)
System Memory – 4 GB
CPU Type – Intel(R) Celeron(R) CPU G3930 @ 2.90GHz
I left the core and memory clocks at default setting (1242/1750) as x16r prefers more core clock than it does memory, so if you are moving over from Ethereum mining you will need to keep that in mind. With Ethereum this rig would have run at 1100 MHz core and 1950 MHz memory. The other thing you may need to adjust is the power settings, as again with Ethereum able to run at lower core settings I could get away with 920mv core voltage, with z16r, since some algorithms are a bit more aggressive, I found I needed to adjust the voltage up to 950-960mv range.
So while it seem like I will need to puzzle a bit harder over how to get meaningful run-run comparisons, we can move on to testing another mining versions.
The other AMD miner currently available is from user brain112358 and is also based upon sgminer. Brian was second place winner in the Ravencoin AMD miner bounty and his miner is also linked from their homepage.
The latest version can be found on the brian112358-sgminerx16r GitHib page here.
Since this miner fork is also based upon sgminer, this makes switching between the two simple as we can simply copy paste the previous Miner.bat file we used with the aceneun fork.
I have modified the bat file a bit and it now contains the following commands which will be used on both versions.
sgminer.exe -k x16r -o stratum+tcp://miningpanda.site:3636 -u RGvCLDzUAYPoryFBxd8zwUjJCi4bckQZgg -p c=RVN -I 19 --gpu-platform 1 -L log.txt
The same as applies with the aceneum fork, you may need to change the gpu-platform to 0 (zero) if you do not have a on-board video device active.
In the image above we have the 1 hour results from running the brian112358-sgminer-x16r fork. Since I used the exact same batch file to launch the program as well as kept all the other settings the same, the first thing I noticed is the markedly lower performance.
Still, having no good way to know what algorithms were running over time, or during the image capture, I decided to try a different approach. I am going to break up the rig into a 3 and 3 configuration, running both forks simultaneously and using just 3 GPUs with each. Since all the settings as well as the pool will be the same, it should hopefully offer us some points for comparison.
I am using the following launch strings to make my comparisons:
aceneum-sgminer-x16r (using GPUs 1, 2 and 3)
sgminer.exe -k x16r -o stratum+tcp://miningpanda.site:3636 -u RGvCLDzUAYPoryFBxd8zwUjJCi4bckQZgg -p c=RVN -I 19 -d 0,1,2 --gpu-platform 1 -L log.txt
brian112358-sgminer-x16r (using GPUs 4, 5 and 6)
sgminer.exe -k x16r -o stratum+tcp://miningpanda.site:3636 -u RGvCLDzUAYPoryFBxd8zwUjJCi4bckQZgg -p c=RVN -I 19 -d 3,4,5 --gpu-platform 1 -L log.txt
The added command -d specifies the devices (GPUs) you want to use, this makes it convenient to shift specific GPUs to each version. For this first test run I assigned GPUs 0-2 (1,2,3) to the aceneum fork and GPUs 3-5 (4,5,6) to the brians112358 fork.
I have placed both windows side by side in the above image to directly compare the results to one another. On the left is the aceneum version and on the right is the brian112358 version. They were both started at the same time (within 1 second) and ran for just over 2 hours.
The initial results seems to imply that the aceneum fork offers better performance for now. Keep in mind that both of these forks are initial release so changes may be int he works, but for right now it appears the aceneum fork gives better results in terms of hashrate and WU. This assumption is further bolstered by looking at the number of accepted shares, 7224 shares with aceneum, versus 5036 with brain112358. Rejects are also a bit lower with the aceneum fork.
I suppose if I were to be completely thorough, I would run the test again swapping the cards each fork is running, but I have done some other testing behind the scenes with other rigs that also bear similar results. I will keep an eye on both versions if they implement any changes, but for right now I would recommend the aceneum fork based solely on the results I have obtained so far.
Update (03/21/2018) – Brian112358 released an update to his mining program brian112358-sgminer-x16r-v0.3.1-dev yesterday (02/20/2018). I downloaded and ran it on the same rig used through-out the rest of this guide to see if there are any improvements from the previous version. After running it for 14 hours here is what I found:
The long-term hashrate has improved somewhat over the previous version and the issues of hardware (HW) errors has been addressed. The update provided by the developer states that:
I fixed a bug with the whirlpool kernel which was causing HW errors due to uninitialized memory. This should increase your poolside hashrate from earlier versions.
After reviewing the results of the 14 hour test run, running with the exact same settings and configuration as the previous tests, I can confirm that the HW errors are now gone. While the hashrate with the latest version may have improved slightly, it still seems the aceneum fork still performs a bit better over the longer term and for now I will stick to using that.
For now that concludes my testing, but hopefully this two part guide gives you enough information to get your own Ravencoin mining rig setup and running.
I plan to try and get a more accurate method of comparison results, and I will look into parsing log files to look for minimum and maximum readings over a several hour period and then averaging them out.
As always, if you have any questions or comments please leave them below.