Topic: Building Cheap Miners : My "Secret" - page 5. (Read 60234 times)
I just double-checked and I had these R815's updated to Ubuntu 18 LTS, that might be why I'm having issues.
Having some issues installing Turion for some reason and I can't seem to find the solution, any ideas?
Code:
g++ -O2 -MMD -MF obj/x86_64/.Brazos.d -MT obj/x86_64/Brazos.o -c -o obj/x86_64/Brazos.o Brazos.cpp
g++ -O2 -MMD -MF obj/x86_64/.Llano.d -MT obj/x86_64/Llano.o -c -o obj/x86_64/Llano.o Llano.cpp
g++ -O2 -MMD -MF obj/x86_64/.Interlagos.d -MT obj/x86_64/Interlagos.o -c -o obj/x86_64/Interlagos.o Interlagos.cpp
g++ -O2 -MMD -MF obj/x86_64/.MSRObject.d -MT obj/x86_64/MSRObject.o -c -o obj/x86_64/MSRObject.o MSRObject.cpp
g++ -O2 -MMD -MF obj/x86_64/.MSVC_Round.d -MT obj/x86_64/MSVC_Round.o -c -o obj/x86_64/MSVC_Round.o MSVC_Round.cpp
In file included from MSVC_Round.cpp:1:0:
MSVC_Round.h:5:17: error: ‘int round(float)’ conflicts with a previous declaration
int round (float);
^
In file included from /usr/include/c++/7/math.h:36:0,
from MSVC_Round.h:3,
from MSVC_Round.cpp:1:
/usr/include/c++/7/cmath:1771:3: note: previous declaration ‘constexpr float std::round(float)’
round(float __x)
^~~~~
Makefile:62: recipe for target 'obj/x86_64/MSVC_Round.o' failed
make: *** [obj/x86_64/MSVC_Round.o] Error 1
I have installed turionpowercontrol on all my R815's (both 6200 & 6300 Series) without issues. I am on Ubuntu 16.04 with all the current updates.
You may want check that you have the latest updates:
sudo apt update
sudo apt upgrade
This is what I do to install turionpowercontrol
Code:
sudo apt install ncurses-dev
sudo apt install build-essential
git clone https://github.com/mh0rst/turionpowercontrol.git
cd turionpowercontrol
make
After it compiles without errors this needs to be done before running it.
Code:
sudo apt install cpuid
sudo modprobe cpuid
sudo modprobe msr
Having some issues installing Turion for some reason and I can't seem to find the solution, any ideas?
Code:
g++ -O2 -MMD -MF obj/x86_64/.Brazos.d -MT obj/x86_64/Brazos.o -c -o obj/x86_64/Brazos.o Brazos.cpp
g++ -O2 -MMD -MF obj/x86_64/.Llano.d -MT obj/x86_64/Llano.o -c -o obj/x86_64/Llano.o Llano.cpp
g++ -O2 -MMD -MF obj/x86_64/.Interlagos.d -MT obj/x86_64/Interlagos.o -c -o obj/x86_64/Interlagos.o Interlagos.cpp
g++ -O2 -MMD -MF obj/x86_64/.MSRObject.d -MT obj/x86_64/MSRObject.o -c -o obj/x86_64/MSRObject.o MSRObject.cpp
g++ -O2 -MMD -MF obj/x86_64/.MSVC_Round.d -MT obj/x86_64/MSVC_Round.o -c -o obj/x86_64/MSVC_Round.o MSVC_Round.cpp
In file included from MSVC_Round.cpp:1:0:
MSVC_Round.h:5:17: error: ‘int round(float)’ conflicts with a previous declaration
int round (float);
^
In file included from /usr/include/c++/7/math.h:36:0,
from MSVC_Round.h:3,
from MSVC_Round.cpp:1:
/usr/include/c++/7/cmath:1771:3: note: previous declaration ‘constexpr float std::round(float)’
round(float __x)
^~~~~
Makefile:62: recipe for target 'obj/x86_64/MSVC_Round.o' failed
make: *** [obj/x86_64/MSVC_Round.o] Error 1
@SixLitr your results are really close to optimal hash rates.
I posted the above while you were posting so I didn't see your data until now.
Your results for Single and Dual CPU's is spot on and consistent.
The Quad CPU results are strange for the last CPU.
The results show around 2425 H/s for the first three CPU's but only 1545 H/s for the last CPU.
Is your system complaining about memory configuration for that CPU?
The other reason for that reduced result is that "HT assist" is enabled in the BIOS. If "HT assist" is enabled it reduces L3 cache by 2MB.
Check the BIOS settings under Processor settings and turn off "HT assist".
You should also gain hashes by doing the TurionPowerControl command:
sudo ./TurionPowerControl -psmax 1
I posted the above while you were posting so I didn't see your data until now.
Your results for Single and Dual CPU's is spot on and consistent.
The Quad CPU results are strange for the last CPU.
The results show around 2425 H/s for the first three CPU's but only 1545 H/s for the last CPU.
Is your system complaining about memory configuration for that CPU?
The other reason for that reduced result is that "HT assist" is enabled in the BIOS. If "HT assist" is enabled it reduces L3 cache by 2MB.
Check the BIOS settings under Processor settings and turn off "HT assist".
You should also gain hashes by doing the TurionPowerControl command:
sudo ./TurionPowerControl -psmax 1
Update:
I don’t believe my friend entered the command you provided correctly. Turns out he has an 815 with 4x6276. He jumped on Loki testnet using RandomX Loki algo and is getting 3400-3600 HS with xmrig for all four Opterons. Bummer...
I don’t believe my friend entered the command you provided correctly. Turns out he has an 815 with 4x6276. He jumped on Loki testnet using RandomX Loki algo and is getting 3400-3600 HS with xmrig for all four Opterons. Bummer...
His numbers "on Loki testnet using RandomX Loki algo" are still way to low.
I am getting 11225 H/s for a Dell R815 with quad 16-core Opteron 6378's or 2806 H/s per each Opteron 6378.
The Opteron 6276 is only 100 MHz slower (2600 MHz) than the Opteron 6378 (2700 MHz) and taking into consideration the architectural changes for the 6300 series (5.7% faster) the 6276 should only be about 9% slower or around 2556 H/s for each Opteron 6276.
----------------------------------------------------------------------------------------------------------------
Data on AMD Opteron's 6378, 6276 and 6300 architectural changes:
AMD Opteron 6378 specifications:
http://www.cpu-world.com/CPUs/Bulldozer/AMD-Opteron%206378%20-%20OS6378WKTGGHK.html
AMD Opteron 6276 specifications:
http://www.cpu-world.com/CPUs/Bulldozer/AMD-Opteron%206276.html
The Piledriver Architecture: Improving On Bulldozer:
https://www.tomshardware.com/reviews/fx-8350-vishera-review,3328-3.html
----------------------------------------------------------------------------------------------------------------
Here are a few pointers to help your friend properly test the RandomX on his quad Opteron 6276 server:
Huge pages need to be 2400 for each processor (1200 per node, 2 nodes per processor) so do this command to set the correct number of Huge pages:
sudo sysctl -w vm.nr_hugepages=9600
Do this command to verify above change took:
cat /proc/meminfo
Be sure numactl is installed:
sudo apt install numactl
This is the test I used:
seq 0 7 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 100000 --init 8 --threads 7
If his numbers are still to low there can be a couple of reasons.
#1 Memory needs to be in installed on each processor and needs to be optimally configured.
For example the Dell R815 needs four memory sticks installed for each Opteron.
#2 His system is not running at the peak 2600 MHz for all cores.
TurionPowerControl can check and fix that problem.
Multiple Processor G34 Checklist
https://hardforum.com/threads/multiple-processor-g34-checklist.1669698
Instructions to install TurionPowerControl:
Code:
sudo apt install ncurses-dev
git clone https://github.com/mh0rst/turionpowercontrol.git
cd turionpowercontrol
make
sudo apt install cpuid
sudo modprobe cpuid
sudo modprobe msr
Now run this command:
sudo ./TurionPowerControl -l
a bunch of data for for each Opteron will be shown.
Here is a sniplet from my quad Opteron 6348 server:
Code:
Machine has 8 nodes
Processor has 6 cores
Processor has 8 p-states
Processor has 2 boost states
Power States table
-- Node: 0 Core 0
core 0 pstate 0 (pb0) - En:1 VID:24 FID:18 DID:0.00 Freq:3400 VCore:1.2500
core 0 pstate 1 (pb1) - En:1 VID:30 FID:15 DID:0.00 Freq:3100 VCore:1.1750
core 0 pstate 2 (p0) - En:1 VID:40 FID:12 DID:0.00 Freq:2800 VCore:1.0500
core 0 pstate 3 (p1) - En:1 VID:43 FID:9 DID:0.00 Freq:2500 VCore:1.0125
core 0 pstate 4 (p2) - En:1 VID:47 FID:5 DID:0.00 Freq:2100 VCore:0.9625
core 0 pstate 5 (p3) - En:1 VID:50 FID:2 DID:0.00 Freq:1800 VCore:0.9250
core 0 pstate 6 (p4) - En:1 VID:54 FID:12 DID:1.00 Freq:1400 VCore:0.8750
core 0 pstate 7 (p5) - En:0 VID:0 FID:0 DID:0.00 Freq:1600 VCore:1.5500
-- Node: x Core x : this is replicated for each core and each Opteron and is the same as above
at the bottom of the data dump you should this:
Code:
--- Node 0:
Processor Maximum PState: 6
Processor Startup PState: 6
Processor Maximum Operating Frequency: 3400 MHz
Minimum allowed VID: 123 (0.0125V) - Maximum allowed VID 24 (1.2500V)
again these are duplicated and repeated for each node.
The "Processor Maximum PState: 6" is the problem in not having the Opteron properly running at the All-core frequency.
This is my RandomX test run with "Processor Maximum PState: 6":
seq 0 7 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 100000 --init 6 --threads 6
Code:
Running benchmark (10000 nonces) ...
Calculated result: 3412fd710459cb4e710e2f72760f632f4f00a0463ab02dbdd7dae176aeb1ccc2
Performance: 1261.93 hashes per second
Calculated result: 3412fd710459cb4e710e2f72760f632f4f00a0463ab02dbdd7dae176aeb1ccc2
Performance: 1261.98 hashes per second
Or 2524 H/s for one 6348 Opteron
Now doing this command:
sudo ./TurionPowerControl -psmax 1
which sets "Processor Maximum PState: 1"
and retesting:
Code:
Running benchmark (10000 nonces) ...
Calculated result: 3412fd710459cb4e710e2f72760f632f4f00a0463ab02dbdd7dae176aeb1ccc2
Performance: 1371.44 hashes per second
Calculated result: 3412fd710459cb4e710e2f72760f632f4f00a0463ab02dbdd7dae176aeb1ccc2
Performance: 1371.41 hashes per second
Or 2743 H/s for one 6348 Opteron or 8.7% faster
Results from testing (running hugepages at 9600 if that matters). Some interesting results testing all 4 CPUs as I noticed the last couple virtual machines slowed down considerably, dropping to 7 threads didn't help and reduced the overall score.
Single CPU [2435h/s]
Dual CPU [4875h/s]
Quad CPU [8836h/s]
Single CPU [2435h/s]
Code:
seq 0 1 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 10000 --init 8 --threads 8
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
Memory initialized in 11.186 s
Initializing 8 virtual machine(s) ...
Memory initialized in 11.1881 s
Initializing 8 virtual machine(s) ...
Running benchmark (10000 nonces) ...
Running benchmark (10000 nonces) ...
Calculated result: 3412fd710459cb4e710e2f72760f632f4f00a0463ab02dbdd7dae176aeb1ccc2
Performance: 1218.28 hashes per second
Calculated result: 3412fd710459cb4e710e2f72760f632f4f00a0463ab02dbdd7dae176aeb1ccc2
Performance: 1217.59 hashes per second
Dual CPU [4875h/s]
Code:
seq 0 3 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 10000 --init 8 --threads 8
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
Memory initialized in 11.192 s
Initializing 8 virtual machine(s) ...
Running benchmark (10000 nonces) ...
Memory initialized in 11.2044 s
Initializing 8 virtual machine(s) ...
Running benchmark (10000 nonces) ...
Memory initialized in 11.2342 s
Initializing 8 virtual machine(s) ...
Running benchmark (10000 nonces) ...
Memory initialized in 11.2376 s
Initializing 8 virtual machine(s) ...
Running benchmark (10000 nonces) ...
Calculated result: 3412fd710459cb4e710e2f72760f632f4f00a0463ab02dbdd7dae176aeb1ccc2
Performance: 1222.54 hashes per second
Calculated result: 3412fd710459cb4e710e2f72760f632f4f00a0463ab02dbdd7dae176aeb1ccc2
Performance: 1219.92 hashes per second
Calculated result: 3412fd710459cb4e710e2f72760f632f4f00a0463ab02dbdd7dae176aeb1ccc2
Performance: 1221.46 hashes per second
Calculated result: 3412fd710459cb4e710e2f72760f632f4f00a0463ab02dbdd7dae176aeb1ccc2
Performance: 1213.3 hashes per second
Quad CPU [8836h/s]
Code:
seq 0 7 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 100000 --init 8 --threads 8
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (8 threads) ...
Memory initialized in 11.181 s
Initializing 8 virtual machine(s) ...
Running benchmark (100000 nonces) ...
Memory initialized in 11.2132 s
Initializing 8 virtual machine(s) ...
Running benchmark (100000 nonces) ...
Memory initialized in 11.2208 s
Initializing 8 virtual machine(s) ...
Memory initialized in 11.2227 s
Initializing 8 virtual machine(s) ...
Memory initialized in 11.2278 s
Initializing 8 virtual machine(s) ...
Running benchmark (100000 nonces) ...
Running benchmark (100000 nonces) ...
Running benchmark (100000 nonces) ...
Memory initialized in 11.2378 s
Initializing 8 virtual machine(s) ...
Running benchmark (100000 nonces) ...
Memory initialized in 19.4071 s
Initializing 8 virtual machine(s) ...
Running benchmark (100000 nonces) ...
Memory initialized in 19.4136 s
Initializing 8 virtual machine(s) ...
Running benchmark (100000 nonces) ...
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1219.7 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1217.48 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1216.08 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1215.29 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1213.56 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1212.34 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 773.804 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 771.625 hashes per second
I am "the friend" - and after reading everything I will be performing some more tests to help provide a more accurate representation of the servers I have ( 2 x R815's each with 64GB of RAM and 4 x Opteron 6276 running Ubuntu 16).
Quick question, as I am learning my way around "server CPU" mining and relatively new to this (been GPU mining for years). How do I test all 4 CPU's together, in addition to a single and dual for comparison? (Just posting the commands is fine, but I prefer to also learn what they do and why/how to use them). Thanks!
Quick question, as I am learning my way around "server CPU" mining and relatively new to this (been GPU mining for years). How do I test all 4 CPU's together, in addition to a single and dual for comparison? (Just posting the commands is fine, but I prefer to also learn what they do and why/how to use them). Thanks!
Update:
I don’t believe my friend entered the command you provided correctly. Turns out he has an 815 with 4x6276. He jumped on Loki testnet using RandomX Loki algo and is getting 3400-3600 HS with xmrig for all four Opterons. Bummer...
I don’t believe my friend entered the command you provided correctly. Turns out he has an 815 with 4x6276. He jumped on Loki testnet using RandomX Loki algo and is getting 3400-3600 HS with xmrig for all four Opterons. Bummer...
With the single cpu command you posted above, he got 1894 hashes.
Running the original command (16 threads) for 100k nonces, his hash actually went up to 3600 from 3300.
Running the original command (16 threads) for 100k nonces, his hash actually went up to 3600 from 3300.
I don’t know much about Opterons or Linux, but this is what a buddy of mine shared with me for a 6276.
sixlitr@SixLitr-R815-1:~/RandomX/build$ ./randomx-benchmark --largePages --mine --jit --threads 16
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (1 thread) ...
Memory initialized in 70.2612 s
Initializing 16 virtual machine(s) ...
Running benchmark (1000 nonces) ...
Calculated result: 38d47ea494480bff8d621189e8e92747288bb1da6c75dc401f2ab4b6807b6010
Reference result: 38d47ea494480bff8d621189e8e92747288bb1da6c75dc401f2ab4b6807b6010
Performance: 3310.21 hashes per second
sixlitr@SixLitr-R815-1:~/RandomX/build$ ./randomx-benchmark --largePages --mine --jit --threads 16
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (1 thread) ...
Memory initialized in 70.2612 s
Initializing 16 virtual machine(s) ...
Running benchmark (1000 nonces) ...
Calculated result: 38d47ea494480bff8d621189e8e92747288bb1da6c75dc401f2ab4b6807b6010
Reference result: 38d47ea494480bff8d621189e8e92747288bb1da6c75dc401f2ab4b6807b6010
Performance: 3310.21 hashes per second
Thanks for sharing.
I do want to point out a couple of issues with the above data. You might want to send this post to your friend for retesting.
First Running the benchmark with only 1000 nonces is way too small. This causes too high results and is not realistic if mining at 100% over an extended period.
Quote
Select N (number of nonces) equal to 10000, 100000 or 1000000 depending on the performance of your system. Aim for at least a 60-second mining period for accurate results.
https://github.com/tevador/RandomX/issues/25
I use 10000 nonces for quick testing and 100000 nonces for actual benchmarking tests of around 60 seconds. By using a too small nonces number the processor never even heats up so any issues with that are not seen and any power measurements would be invalid.
Second the "Performance: 3310.21 hashes per second" is too high for a single 6276. More than likely the system is a Dual Opteron 6276 system and the benchmark test from above is running on both processors and thus causing the too high number.
These Opterons are actually two dies in a single package (NUMA node 0 and 1) so to test properly you need to run two benchmarks one assigned to each node.
To test a single 6276 Opteron do this:
seq 0 1 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 10000 --init 8 --threads 8
This will run 16 threads (the same as the initial test from above) but does so by assigning 8 threads to each of the two dies inside a single Opteron.
For testing a Dual Opteron 6276 system do this:
seq 0 3 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 10000 --init 8 --threads 8
I don’t know much about Opterons or Linux, but this is what a buddy of mine shared with me for a 6276.
sixlitr@SixLitr-R815-1:~/RandomX/build$ ./randomx-benchmark --largePages --mine --jit --threads 16
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (1 thread) ...
Memory initialized in 70.2612 s
Initializing 16 virtual machine(s) ...
Running benchmark (1000 nonces) ...
Calculated result: 38d47ea494480bff8d621189e8e92747288bb1da6c75dc401f2ab4b6807b6010
Reference result: 38d47ea494480bff8d621189e8e92747288bb1da6c75dc401f2ab4b6807b6010
Performance: 3310.21 hashes per second
sixlitr@SixLitr-R815-1:~/RandomX/build$ ./randomx-benchmark --largePages --mine --jit --threads 16
RandomX benchmark v1.0.4
- full memory mode (2080 MiB)
- JIT compiled mode
- hardware AES mode
- large pages mode
Initializing (1 thread) ...
Memory initialized in 70.2612 s
Initializing 16 virtual machine(s) ...
Running benchmark (1000 nonces) ...
Calculated result: 38d47ea494480bff8d621189e8e92747288bb1da6c75dc401f2ab4b6807b6010
Reference result: 38d47ea494480bff8d621189e8e92747288bb1da6c75dc401f2ab4b6807b6010
Performance: 3310.21 hashes per second
I just upgraded one Dell R815 to quad 16-core Opteron 6378's and tested RandomX on it.
Results were not as I expected.
All testing were done with: TurionPowerControl -psmax 1
Testing one Opteron 6378:
seq 0 1 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 100000 --init 8 --threads 8
gave good results of 2995 H/s
but when I tested all four Opteron 6378's at the same time:
seq 0 7 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 100000 --init 8 --threads 8
the H/s dropped down to 2777 H/s per Opteron 6378 or a reduction of 7.3%
It seems that some thermal slowdown is happening with thread count 8 (16 per cpu) when all processors are active. I tested inside under AC so the ambient temp was not a factor.
I tried with 7 threads and it slightly improved 1% to 2806 H/s per Opteron 6378 or 11225 H/s for a Dell R815 with quad 16-core Opteron 6378's.
seq 0 7 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 100000 --init 8 --threads 7
6 threads had a large drop off so 7 threads is optimal.
The 2806 H/s per Opteron 6378 is only 5.8% better than an Opteron 6348 which produces 2653 H/s not the 16.1% improvement I was expecting.
Measured power during mining was 815 watts (idle 265 watts) or 13.77 Hashes per Watt.
Results were not as I expected.
All testing were done with: TurionPowerControl -psmax 1
Testing one Opteron 6378:
seq 0 1 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 100000 --init 8 --threads 8
gave good results of 2995 H/s
Code:
Running benchmark (10000 nonces) ...
Calculated result: 3412fd710459cb4e710e2f72760f632f4f00a0463ab02dbdd7dae176aeb1ccc2
Performance: 1497.22 hashes per second
Calculated result: 3412fd710459cb4e710e2f72760f632f4f00a0463ab02dbdd7dae176aeb1ccc2
Performance: 1497.49 hashes per second
but when I tested all four Opteron 6378's at the same time:
seq 0 7 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 100000 --init 8 --threads 8
the H/s dropped down to 2777 H/s per Opteron 6378 or a reduction of 7.3%
Code:
Running benchmark (100000 nonces) ...
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1398.73 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1394.46 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1393.11 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1390.36 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1386.95 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1386.33 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1383.43 hashes per second
Calculated result: d6660144e9a2e68bf47d7cc8afc206672e72f82dfff69fe0d974531e85f7504f
Performance: 1375.52 hashes per second
It seems that some thermal slowdown is happening with thread count 8 (16 per cpu) when all processors are active. I tested inside under AC so the ambient temp was not a factor.
I tried with 7 threads and it slightly improved 1% to 2806 H/s per Opteron 6378 or 11225 H/s for a Dell R815 with quad 16-core Opteron 6378's.
seq 0 7 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 100000 --init 8 --threads 7
6 threads had a large drop off so 7 threads is optimal.
The 2806 H/s per Opteron 6378 is only 5.8% better than an Opteron 6348 which produces 2653 H/s not the 16.1% improvement I was expecting.
Measured power during mining was 815 watts (idle 265 watts) or 13.77 Hashes per Watt.
Guys,
I am playing around with Proliant G7 4x E7-4870 and I cannot figure out the cpu threads config for XMR-STAK.
Getting either thread 0 error or very low hashrate.
I am playing around with Proliant G7 4x E7-4870 and I cannot figure out the cpu threads config for XMR-STAK.
Getting either thread 0 error or very low hashrate.
What is your OS (Linux or Windows) version number?
Post your Nvidia.txt config file here so I can look at it and give you some help.
Hi. I booted HiveOS on this machine from the USB drive.
I no longer mine using HiveOS so I suggest going to the HiveOS Forum and asking them for help.
https://forum.hiveos.farm
Ar you using plain Ubuntu?
Guys,
I am playing around with Proliant G7 4x E7-4870 and I cannot figure out the cpu threads config for XMR-STAK.
Getting either thread 0 error or very low hashrate.
I am playing around with Proliant G7 4x E7-4870 and I cannot figure out the cpu threads config for XMR-STAK.
Getting either thread 0 error or very low hashrate.
What is your OS (Linux or Windows) version number?
Post your Nvidia.txt config file here so I can look at it and give you some help.
Hi. I booted HiveOS on this machine from the USB drive.
I no longer mine using HiveOS so I suggest going to the HiveOS Forum and asking them for help.
https://forum.hiveos.farm
Guys,
I am playing around with Proliant G7 4x E7-4870 and I cannot figure out the cpu threads config for XMR-STAK.
Getting either thread 0 error or very low hashrate.
I am playing around with Proliant G7 4x E7-4870 and I cannot figure out the cpu threads config for XMR-STAK.
Getting either thread 0 error or very low hashrate.
What is your OS (Linux or Windows) version number?
Post your Nvidia.txt config file here so I can look at it and give you some help.
Hi. I booted HiveOS on this machine from the USB drive.
My personal opinion is XMR is gonna fuck something up, RandomX isn't going to be the magic bullet they think it will and they will eventually be forced to embrace ASICs.
Well they have put in a lot of work in RandomX and have made some changes to it to further harden it to be ASIC resistant until the code freeze in April.
Specification: https://github.com/tevador/RandomX/blob/master/doc/specs.md
Design: https://github.com/tevador/RandomX/blob/master/doc/design.md
Possible ASIC design: https://github.com/tevador/RandomX/issues/11
Single-chip ASIC design: https://github.com/tevador/RandomX/issues/31
Quote from: PharmEcis
New AMD chips are also putting out 1/2 the hash of a full R815 on a single chip for a 1/3 of the juice.
And for those miners they will have to purchase new AMD CPU's and motherboards. Those are costs we don't have to experience as we already have the hardware.
I am doing some minor (and cheap) upgrades to the Dell R815's by purchasing some Opteron Upgrades that will pay out for higher Hash rates and lower power.
Purchased 10x AMD 12-core 6348 Opteron's for a total of $64.10. These along with 2x 12-core 6348 Opteron's from a Supermicro server will going in three of my Dell R815 servers.
Purchased 8x AMD 16-core 6378 Opteron's for a total of $69.29. These will be going into my other two Dell R815 servers.
Purchased 2x AMD 16-core 6376 Opteron's for a total of $24.67. These will be going into the Supermicro server that had the 6348 Opteron's in it.
As for power used. Yes it will be higher power and "Hashes per watt" will be lower but if the profit is high enough then that works for me.
This is what I am measuring for a Dell R815 Server with 4x AMD 12-core 6348 Opterons:
Idle Power: 265 watts
Mining Power: 865 watts - That is after the warmup period. It starts at 800 watts and slowly climbs and peaks at the 865 watt number.
RandomX Hash rate: 10629 H/s overall or 2657 H/s for each AMD 12-core 6348 Opteron.
Dell R815 Server RandomX "Hashes per watt": 12.3
Quote from: PharmEcis
I'm hoping they will make something, but I really feel like these servers along with a lot of other things I bought over a year ago were a fools folly.
We should make some profit mining RandomX on the Dell R815's. Not so much (or at all on the HP DL580 G7's).
Also remember that no one should be mining RandomX with GPU's and that botnet mining should be also a thing in the past and that a lot of current miners have Celerons for their processors driving large GPU mining rigs. Those Celerons are pretty useless for RandomX mining.
Guys,
I am playing around with Proliant G7 4x E7-4870 and I cannot figure out the cpu threads config for XMR-STAK.
Getting either thread 0 error or very low hashrate.
I am playing around with Proliant G7 4x E7-4870 and I cannot figure out the cpu threads config for XMR-STAK.
Getting either thread 0 error or very low hashrate.
What is your OS (Linux or Windows) version number?
Post your Nvidia.txt config file here so I can look at it and give you some help.
What coin are you mining?
XMR on CnR
What coin are you mining?
Guys,
I am playing around with Proliant G7 4x E7-4870 and I cannot figure out the cpu threads config for XMR-STAK.
Getting either thread 0 error or very low hashrate.
Can you please share the config? Thanks in advance
I am playing around with Proliant G7 4x E7-4870 and I cannot figure out the cpu threads config for XMR-STAK.
Getting either thread 0 error or very low hashrate.
Can you please share the config? Thanks in advance
Jump to: