Monero Mining Performance

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Patrick

Patrick

Administrator
Staff member
Dec 21, 2010
12,513
5,802
113
zxt said:
Hello, I'm wondering how you can use Xeon Phi to mining Monero. Can you give some detail?
Click to expand...
Using the X200 series I loaded Wolf's cpuminer then used nproc-1 for threads. I also needed to set MCDRAM to cache mode in BIOS (this doubled performance).

The next step would be to compile with icc and see if that plus any of the AVX512 flags help performance.
 
  • Like
Reactions: BigDaddy and gigatexal
Patrick

Patrick

Administrator
Staff member
Dec 21, 2010
12,513
5,802
113
For those who want to start mining easily using Docker images, I just pushed a public image with instructions:
Monero CPU Mining in Docker

One line to turn a system into a miner. I have swapped over to minergate from Moneropool on most of our nodes already so this image is using minergate. You should be able to use the image to push over a Docker swarm cluster as well.
 
  • Like
Reactions: gigatexal
Patrick

Patrick

Administrator
Staff member
Dec 21, 2010
12,513
5,802
113
@jim Just tried non-dockerized on a dual E5 V4 machine:

[2017-01-22 21:25:46] : Result accepted by the pool.
HASHRATE REPORT
| ID | 2.5s | 60s | 15m | ID | 2.5s | 60s | 15m |
| 0 | 22.1 | 22.2 | (na) | 1 | 22.9 | 23.0 | (na) |
| 2 | 22.9 | 23.0 | (na) | 3 | 21.5 | 21.5 | (na) |
| 4 | 22.9 | 23.1 | (na) | 5 | 23.5 | 23.6 | (na) |
| 6 | 27.7 | 27.8 | (na) | 7 | 28.3 | 28.4 | (na) |
| 8 | 26.1 | 26.1 | (na) | 9 | 25.8 | 25.8 | (na) |
| 10 | 25.8 | 25.9 | (na) | 11 | 25.1 | 25.1 | (na) |
| 12 | 38.0 | 38.0 | (na) | 13 | 37.8 | 37.8 | (na) |
| 14 | 37.9 | 37.9 | (na) | 15 | 38.2 | 38.2 | (na) |
| 16 | 39.3 | 39.3 | (na) | 17 | 39.0 | 39.0 | (na) |
| 18 | 39.1 | 39.1 | (na) | 19 | 39.2 | 39.2 | (na) |
| 20 | 38.0 | 38.0 | (na) | 21 | 37.8 | 37.8 | (na) |
| 22 | 37.6 | 37.6 | (na) | 23 | 38.0 | 38.0 | (na) |
| 24 | 22.6 | 22.7 | (na) | 25 | 21.7 | 21.8 | (na) |
| 26 | 21.9 | 21.9 | (na) | 27 | 22.9 | 23.0 | (na) |
| 28 | 23.1 | 22.1 | (na) | 29 | 23.0 | 23.1 | (na) |
-----------------------------------------------------
Totals: 889.8 890.1 (na) H/s
Highest: 890.3 H/s


Here is the Docker CPU version results on the same machine:
[2017-01-23 05:27:48] accepted: 12375/12375 (100.00%), 890.41 H/s at diff 9244 (yay!!!)
[2017-01-23 05:27:53] accepted: 12376/12376 (100.00%), 887.44 H/s at diff 9244 (yay!!!)
[2017-01-23 05:28:03] accepted: 12377/12377 (100.00%), 895.22 H/s at diff 9244 (yay!!!)
[2017-01-23 05:28:06] accepted: 12378/12378 (100.00%), 903.99 H/s at diff 9244 (yay!!!)

Seems fairly close in terms of performance. I did all of your tweaks including setting CPU affinity for the stak version.
 
  • Like
Reactions: jim
M

Marsh

Moderator
May 12, 2013
2,645
1,496
113
I am using the docker version, where is the logfile of H/s that you are showing in the previous thread?
thanks
 
Patrick

Patrick

Administrator
Staff member
Dec 21, 2010
12,513
5,802
113
Marsh said:
I am using the docker version, where is the logfile of H/s that you are showing in the previous thread?
thanks
Click to expand...
If you are using the one I posted you can docker attach containerid and see the H/s rate.

You can also run -it instead of -d when you launch
 
M

Marsh

Moderator
May 12, 2013
2,645
1,496
113
My little Gigabyte D-1521 system running the docker version

2017-01-23 06:35:21] accepted: 799/799 (100.00%), 129.51 H/s at diff 1338 (yay!!!)
[2017-01-23 06:35:21] accepted: 800/800 (100.00%), 127.73 H/s at diff 1338 (yay!!!)
[2017-01-23 06:35:35] accepted: 801/801 (100.00%), 129.47 H/s at diff 1338 (yay!!!)
[2017-01-23 06:35:40] accepted: 802/802 (100.00%), 129.49 H/s at diff 1338 (yay!!!)

Tomorrow, if the weather is bad , and rain. I'll install it on the ASUS 4nodes server.

Thanks Patrick
 
  • Like
Reactions: Patrick
Jeggs101

Jeggs101

Well-Known Member
Dec 29, 2010
1,529
241
63
Signed up using your code. This was stupid easy but I still messed up the example@example.com part ;)

One command ha!

It also made me realize my nvdocker and docker 1.13 weren't compatible. I had to re-install NVidia-docker.

for the NVidia image you'd be better off not using GPU0 in naming since it means when you launch a second time it uses the same name. Problem if you've got multiple GPUs and want to script the initialization.

4 GPU containers initialization took around 20s only
 
OBasel

OBasel

Active Member
Dec 28, 2010
494
62
28
I've now mined 3.03 XMR. What can I do with them? Can I buy stuff? I've only found bitcoin accepted. How do I get USD?
 
J

jim

New Member
Jul 22, 2016
7
5
3
32
Patrick said:
@jim Just tried non-dockerized on a dual E5 V4 machine:

[2017-01-22 21:25:46] : Result accepted by the pool.
HASHRATE REPORT
| ID | 2.5s | 60s | 15m | ID | 2.5s | 60s | 15m |
| 0 | 22.1 | 22.2 | (na) | 1 | 22.9 | 23.0 | (na) |
| 2 | 22.9 | 23.0 | (na) | 3 | 21.5 | 21.5 | (na) |
| 4 | 22.9 | 23.1 | (na) | 5 | 23.5 | 23.6 | (na) |
| 6 | 27.7 | 27.8 | (na) | 7 | 28.3 | 28.4 | (na) |
| 8 | 26.1 | 26.1 | (na) | 9 | 25.8 | 25.8 | (na) |
| 10 | 25.8 | 25.9 | (na) | 11 | 25.1 | 25.1 | (na) |
| 12 | 38.0 | 38.0 | (na) | 13 | 37.8 | 37.8 | (na) |
| 14 | 37.9 | 37.9 | (na) | 15 | 38.2 | 38.2 | (na) |
| 16 | 39.3 | 39.3 | (na) | 17 | 39.0 | 39.0 | (na) |
| 18 | 39.1 | 39.1 | (na) | 19 | 39.2 | 39.2 | (na) |
| 20 | 38.0 | 38.0 | (na) | 21 | 37.8 | 37.8 | (na) |
| 22 | 37.6 | 37.6 | (na) | 23 | 38.0 | 38.0 | (na) |
| 24 | 22.6 | 22.7 | (na) | 25 | 21.7 | 21.8 | (na) |
| 26 | 21.9 | 21.9 | (na) | 27 | 22.9 | 23.0 | (na) |
| 28 | 23.1 | 22.1 | (na) | 29 | 23.0 | 23.1 | (na) |
-----------------------------------------------------
Totals: 889.8 890.1 (na) H/s
Highest: 890.3 H/s


Here is the Docker CPU version results on the same machine:
[2017-01-23 05:27:48] accepted: 12375/12375 (100.00%), 890.41 H/s at diff 9244 (yay!!!)
[2017-01-23 05:27:53] accepted: 12376/12376 (100.00%), 887.44 H/s at diff 9244 (yay!!!)
[2017-01-23 05:28:03] accepted: 12377/12377 (100.00%), 895.22 H/s at diff 9244 (yay!!!)
[2017-01-23 05:28:06] accepted: 12378/12378 (100.00%), 903.99 H/s at diff 9244 (yay!!!)

Seems fairly close in terms of performance. I did all of your tweaks including setting CPU affinity for the stak version.
Click to expand...
Looks strange, it's slower per core than E5 2670 V1.... I tried wolf's cpuminer some time ago on my dual 2670, but even after manually binding with numactl and setting cpu affinity the best I could get is ~ 800h/s; I haven't tried the dockerized image though.
I'm currently running the stak miner only @ 19 threads out of the optimal 20, as I'm using it for other stuff :)
Here's the result:

HASHRATE REPORT
| ID | 2.5s | 60s | 15m | ID | 2.5s | 60s | 15m |
| 0 | 44.5 | 45.9 | 45.9 | 1 | 43.9 | 45.5 | 45.6 |
| 2 | 49.4 | 51.1 | 51.2 | 3 | 49.0 | 50.8 | 50.8 |
| 4 | 49.4 | 50.8 | 50.8 | 5 | 50.1 | 51.2 | 51.3 |
| 6 | 50.2 | 51.6 | 51.7 | 7 | 50.6 | 52.0 | 52.1 |
| 8 | 48.0 | 48.5 | 48.6 | 9 | 53.6 | 54.0 | 54.1 |
| 10 | 53.0 | 53.7 | 53.8 | 11 | 52.6 | 53.3 | 53.4 |
| 12 | 52.6 | 53.3 | 53.3 | 13 | 53.0 | 53.8 | 53.8 |
| 14 | 53.1 | 54.1 | 54.2 | 15 | 53.5 | 54.5 | 54.7 |
| 16 | 44.5 | 45.9 | 45.9 | 17 | 43.9 | 45.5 | 45.6 |
| 18 | 48.0 | 48.5 | 48.6 |
-----------------------------------------------------
Totals: 942.8 964.1 965.4 H/s
Highest: 967.6 H/s

My guess would probably be a difference with the hugepages setups, or the AVX2 extension on the E5 V4...
For hugepages, as I'm running 19 threads, I can verify that it's indeed locking exactly 19 hugepages of 2MB each...

$ cat /proc/meminfo |grep Huge
AnonHugePages: 0 kB
ShmemHugePages: 0 kB
HugePages_Total: 128
HugePages_Free: 109
HugePages_Rsvd: 0
HugePages_Surp: 0
Hugepagesize: 2048 kB
 
Patrick

Patrick

Administrator
Staff member
Dec 21, 2010
12,513
5,802
113
@jim, I think the E5-2670 V1 does well from a clock speed perspective.

One part I am not liking is that it seems like stak is much harder to setup for each system. I would love to get an extra 10-15% hash rate but I would want to automate the setup. Stak seems to have a bunch of items that still require manual tuning.
 
Patrick

Patrick

Administrator
Staff member
Dec 21, 2010
12,513
5,802
113
@Bilbo I saw that one, when Wolf's gets updated, the miners will get updated. If I can get Stak working across platforms in an automated way, may move there. Still a few thousand in just fees to launch that attack.
 
Patrick

Patrick

Administrator
Staff member
Dec 21, 2010
12,513
5,802
113
@Marsh Comments on heatsink seems somewhat concerning. But interesting. Need to do some research into the BIOS mods miners seem to be using. I also saw this one which looks intriguing Computer Parts, PC Components, Laptop Computers, LED LCD TV, Digital Cameras and more - Newegg.com

@Ned Kelly The Xeon Phi 7210 has several different MCDRAM and fabric settings so consider that a WIP. It takes a BIOS change for each combination and the differences are that big. 1117 is the highest I have gotten thus far.
 
Last edited:
N

Ned Kelly

New Member
Jan 24, 2017
6
0
1
44
@Patrick
Thanks for reply.
So, you used nproc-1 in both cases just changed DRAM settings, is it correct ? Is it possible to test with -t=L2 / 2

Also have you tried Claymore CryptoNote CPU Miner v3.5 with -lowcpu 2 switch using some high cores count Xeons? It use 1 thread, but much faster than -t 1 mode. I guess it utilized L1/L2 cache across different cores not L3. However this miner run under Win only, I guess.
 
You must log in or register to reply here.
Top Bottom