Monero Mining Performance

im10er

New Member
Feb 11, 2018
8
0
1
41
Code:
Architecture:          x86_64
CPU op-mode(s):        32-bit, 64-bit
Byte Order:            Little Endian
CPU(s):                1
On-line CPU(s) list:   0
Thread(s) per core:    1
Core(s) per socket:    1
Socket(s):             1
NUMA node(s):          1
Vendor ID:             GenuineIntel
CPU family:            6
Model:                 63
Model name:            Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz
Stepping:              2
CPU MHz:               2399.998
BogoMIPS:              4799.99
Virtualization:        VT-x
Hypervisor vendor:     KVM
Virtualization type:   full
L1d cache:             32K
L1i cache:             32K
L2 cache:              256K
L3 cache:              20480K
NUMA node0 CPU(s):     0
Flags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon rep_good nopl eagerfpu pni pclmulqdq vmx ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer xsave avx f16c rdrand hypervisor lahf_lm abm tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid xsaveopt
 

DrStein99

If it does not exist ? I am probably building it.
Feb 3, 2018
104
4
18
47
New Jersey, USA
I just realized that crypto-1 is E5-2630 not E5-2630L I checked the logs and it says

monero.0.o3c2ts4e28qd@crypto-2 | * CPU: Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz (1) x64 AES-NI
monero.0.f2o6eqhfcz79@crypto-3 | * CPU: Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz (1) x64 AES-NI
monero.0.u43o90yv3qi0@crypto-1 | * CPU: Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz (1) x64 -AES-NI
Is "monero.0.o3c2ts4e28qd@crypto-2" a bench-mark result? Can you help me decipher what this is?
 

sharpnel

New Member
Aug 22, 2016
11
5
3
26
just bought some cpu to test.
All of them are diskless
Hashrates are from 30mins average.

2x E5-2470 + 1x 4GB 2Rx8 1333 + Intel S2400SC + 350W Gold PSU
= 890H/s using XMRig 20 Threads
10 Cores Total Enabled. (5+5 with HT)
155W (150W with fan removed)
~30C ambient temp
5.5+ Hash per watt.


2x E5-2430 v2 + 1x 4GB 2Rx8 1333 + Intel S2400SC + 350W Gold PSU
= 670H/s using XMRig 14 Threads
8 Cores Total Enabled. (4+4 with HT)
90W (85W with fan removed)
~30C ambient temp
6.7+ Hash per watt.


2x E5-2680 v2 + 1x 4GB 2Rx8 1333 + Quanta Windmill
= 1250H/s using XMRig 24 Threads
12 Cores Total Enabled. (6+6 with HT)
135W (125W with fan removed)
~30C ambient temp
8.5+ Hash per watt.

**I didn't test this properly.
2x E5-2448L v2 + 1x 4GB 2Rx8 1333 + Intel S2400SC + 350W Gold PSU
= 820H/s using XMRig 24 Threads
12 Cores Total Enabled. (6+6 with HT)
xxxW
~30C ambient temp
~8.0+ Hash per watt. (estimated.)



 
Last edited:
  • Like
Reactions: Joel and Marsh

Joel

Active Member
Jan 30, 2015
815
166
43
39
2x E5-2680 v2 + 1x 4GB 2Rx8 1333 + Quanta Windmill
= 1250H/s using XMRig 24 Threads
12 Cores Total Enabled. (6+6 with HT)
135W (125W with fan removed)
~30C ambient temp
8.5+ Hash per watt.
That's pretty impressive. Which OS are you running to get this?

That's actually pretty close to Vega in h/w metrics.
 
Last edited:

sharpnel

New Member
Aug 22, 2016
11
5
3
26
That's pretty impressive. Which OS are you running to get this?

That's actually pretty close to Vega in h/w metrics.
Win 10 pro.

@sharpnel are those 1x 4GB per CPU or per system?
4gb per system.

PS. I don't know why I got 10+H/s more hashrate from Intel motherboard, maybe the bios setting is different. Testing E5-2470 on ASUS Z9NA got around 875-880H/s avg, same config.
 

mantis

Member
Nov 17, 2017
38
6
8
49
I could not get over 600h/s without enabling all the cores. Using all the 10 cores per cpu, its around 637h/s per CPU. On E5-2680v2.

Are the results you got average speeds or top speeds?
 
Last edited:

Joel

Active Member
Jan 30, 2015
815
166
43
39
So I duplicated your config on Ubuntu Server with Patrick's universal xmrig docker image.

Code:
sudo docker run -itd -e username=$wallet.$worker /
-e numthreads=24 --cpuset-cpus="0-23" -e pool=cryptonight.usa.nicehash.com /
-e startport=3355 servethehome/universal_cryptonight:latest
On my dual node Supermicro with 4x 2680v2s onboard, I'm saving 100w at a cost of 100h/s (1120h/s per node instead of 1170h/s) vs. all cores online. Now that the weather's getting a bit warmer the reduced heat output is also welcomed.

Edit: Running two threads affined to each NUMA node brought me back to 1170. I don't know how that affected power consumption, but I'm sure there's still some savings.
 
Last edited:

DrStein99

If it does not exist ? I am probably building it.
Feb 3, 2018
104
4
18
47
New Jersey, USA
How do you have 4 e5 CPU in one machine? Am so reading that right? Or is a "supermicro 4x" the make & model of the machine?
 

jims2321

Active Member
Jul 7, 2013
184
44
28
anybody done benchmarking with the new universal image on opteron's yet that can share their results. I have not had time to try the new image due to work.
 

jims2321

Active Member
Jul 7, 2013
184
44
28
generally you match the miners to the physical cores. so each miner is running essentially on a individual cpu, despite there been to cpu's on each motherboard and shared ram.
 

Joel

Active Member
Jan 30, 2015
815
166
43
39
How does that work? Are you just running 2 miners at the same time?
Clicking on the link I posted above would be helpful if you're curious. But maybe that's too much trouble...

"Two hot-pluggable systems (nodes) in a 2U form factor. Each node supports the following:
1. Dual socket R (LGA 2011) supports
Intel® Xeon® processor E5-2600
and E5-2600 v2 family†"


SYS-6027TR-D71QRF.jpg
 
  • Like
Reactions: DrStein99

DrStein99

If it does not exist ? I am probably building it.
Feb 3, 2018
104
4
18
47
New Jersey, USA
"Two hot-pluggable systems (nodes) in a 2U form factor.
Ok, so it is a rack frame housing, to swap the whole system. Like a blade server without the bulky PCI(e/x) bus. I thought you were booting them up to merge the whole resources of the machines into one operating system over a fibre backplane. Crazy me, I guess I watch too much science fiction, my imagination drifts off into the future where that could actually be possible.
 

onsit

Member
Jan 5, 2018
98
26
18
30
Strange, I get 650H/s on a Dell R520 with 2x E5-2450L (120 watts). I picked up a bunch of Supermicro SSG-2027B-DE2R24L (2 node LGA1356 system). Running just one node atm, the system seems to run at (190 watts) - identical cpu settings as the Dell. (6 cores active, DDR3L running at 800 mhz).

This is with nothing plugged into the backplate, just using the fancy 64gb SSD that seems to usually come with these types of supermicros.

Is it normal for supermicro machines to be more powerhungry compared to HP or Dell?
 

Joel

Active Member
Jan 30, 2015
815
166
43
39
Strange, I get 650H/s on a Dell R520 with 2x E5-2450L (120 watts). I picked up a bunch of Supermicro SSG-2027B-DE2R24L (2 node LGA1356 system). Running just one node atm, the system seems to run at (190 watts) - identical cpu settings as the Dell. (6 cores active, DDR3L running at 800 mhz).

This is with nothing plugged into the backplate, just using the fancy 64gb SSD that seems to usually come with these types of supermicros.

Is it normal for supermicro machines to be more powerhungry compared to HP or Dell?
I only have experience with the 6027 setup, but I suspect some of the common hardware has to be powered on regardless of how many nodes are online. When I was doing tests, a single 2x2680v2 node ran at every bit of 250w hashing with a 1050 onboard as well. Imagine my surprise when 4 nodes "only" pulled ~920w.

Upgrading to latest IPMI firmware and setting fans to "Optimized" mode might reduce power use (and noise too!) as well.
 

onsit

Member
Jan 5, 2018
98
26
18
30
I only have experience with the 6027 setup, but I suspect some of the common hardware has to be powered on regardless of how many nodes are online. When I was doing tests, a single 2x2680v2 node ran at every bit of 250w hashing with a 1050 onboard as well. Imagine my surprise when 4 nodes "only" pulled ~920w.

Upgrading to latest IPMI firmware and setting fans to "Optimized" mode might reduce power use (and noise too!) as well.
Yeah you seem to be correct, i'd imagine the backplate has a lot of electronics to provide 12v and 5v so it takes more idle power. Removing the Mezzanine sas / raid cards seemed to have dropped power consumption. Running 2 nodes at idle around 200watts now, so it's about 100 watts of overhead for the fans / backplate.

Haven't had time to setup xmrig yet, but i'll be happy if i can get it to 1200H/s @ 250-260watts 2 node server all-in.

This unit also came with 1U fans instead of 2U fans - I should have done my homework, but I assumed all 2U supermicros came with 80mm fans.

Need to figure out how to use ipmiutil to drop fan speed to 10%, these 1U howlers are LOUD at 50% idle. I might just pull all of them out and stick a household fan infront of the rack.

E5-2450L are a hidden gem @ $30 a cpu for dual cpu 650H/s. Finding hardware to run LGA 1356 is the only issue.