Monero Mining Performance

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

Spartus

Active Member
Mar 28, 2012
323
121
43
Toronto, Canada
it seems to me monero mining doesn't scale linearly with clock speed. it is more related to cache and AES acceleration. Honestly though I haven't thoroughly investigated.
 

Marsh

Moderator
May 12, 2013
2,645
1,496
113
@MiniKnight
I published the I7-3770K result on this thread, hoping other people could confirm my finding.

I spend an hour scratching my head , and repeated the test many times by rebooting and restarting it.
Still gets 300H/s.

I have 3 x E3-1240 v2 ( similar to I7-3770K without video ). Later today, I'll perform similar test ( not the same motherboard and PS, I may use a Intel server board to confirm the hashrate ).

I have a I7-4770s in a Shuttle SH87 tiny PC, not exactly the same.
I7-4770S ( 4 threads ) idle at 18w ( 2gb ram ) , 283H/s , 54w = 5.24H/s per watt

I think there is something special about energy efficient of HP 8300 SFF with I7-3770K
Here is a similar test with a I5 in HP 8300 SFF
I5-3470 ( 3 threads ) idle at 11w ( 8gb ram ) , 208H/s , 39w = 5.33H/s per watt
 

Spartus

Active Member
Mar 28, 2012
323
121
43
Toronto, Canada
Upgraded main pc to ryzen 1700, thought I might share some numbers

Ryzen 1700 Stock clocks: 515 H/S

2x R9 290 @1025/1400: 1640 H/S

System total: ~2155 H/S

power at wall: 442 W with platinum power supply. (120 V)

4.87 H/W
 
  • Like
Reactions: Patrick

MiniKnight

Well-Known Member
Mar 30, 2012
3,072
973
113
NYC
@Marsh at what point do you think smaller systems are better/ worse than big ones? More power strip, switch ports and everything.

I'm also considering that it may be better to use no BMC boards since they'll save 5w but then you can't remote manage?
 

Klee

Well-Known Member
Jun 2, 2016
1,289
396
83
@Marsh at what point do you think smaller systems are better/ worse than big ones? More power strip, switch ports and everything.

I'm also considering that it may be better to use no BMC boards since they'll save 5w but then you can't remote manage?
Well I will find out, just ordered one of these last night before it backordered again.

BIOSTAR TB85 LGA 1150 Intel B85 SATA 6Gb/s USB 3.0 ATX Motherboards - Intel - Newegg.com


I have a i5 4590 that will be put to use, really only need graphics cards .
 
Last edited:
  • Like
Reactions: Patrick

Marsh

Moderator
May 12, 2013
2,645
1,496
113
@Klee
Shuttle SH87
4.4.0-62-generic , xmr-stak-cpu
i5-4590 ( 3 threads ) idle at 15w ( 1gb ram ) , 221H/s , 49w = 4.51H/s per watt

I5-4590 has video on chip
 

zer0sum

Well-Known Member
Mar 8, 2013
849
474
63
@Klee
Shuttle SH87
4.4.0-62-generic , xmr-stak-cpu
i5-4590 ( 3 threads ) idle at 15w ( 1gb ram ) , 221H/s , 49w = 4.51H/s per watt

I5-4590 has video on chip
Interesting that your 4590 gets 220+ and I can only get about 110 out of an overclocked 3570k running at 4.4Ghz.
Any chance you could share your xmr stak config?
 

Marsh

Moderator
May 12, 2013
2,645
1,496
113
"cpu_thread_num" : 3,
"cpu_threads_conf" : [
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 0 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 1 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 2 },
],

"use_slow_memory" : "never",
"verbose_level" : 1,


# /etc/security/limits.conf
# * soft memlock 626688
# * hard memlock 626688
sudo sysctl -w vm.nr_hugepages=30

Running on Ubuntu 16.0.4 , Kernel 4.4.0-62-generic , xmr-stak-cpu.
I tested with the new Kernel 4.10.xxx , but it got slower with a handful of hashrate, so I just keep using Kernel 4.4.0-62-generic.

Here is another datapoints
I5-3470 ( 3 threads ) idle at 11w ( 8gb ram ) , 208H/s , 39w = 5.33H/s per watt


I5-4570 ( 3 threads ) idle at 17w ( 8gb ram ) 215H/s , 44w = 4.88H/s per watt
i5-4460 ( 3 threads ) idle at 16w ( 1gb ram ) , 202H/s , 43w = 4.69H/s per watt
 

Marsh

Moderator
May 12, 2013
2,645
1,496
113
I upgraded xmr-stak-cpu last night, turned on xmr-stak-cpu HTTP web server
Here is a screen shot of I5-3470 , may be later today, I'll bring up a I5-3570.

I think the previous version of xmr-stak-cpu was few hashrates faster, also http web server was not enabled in my previous measurement.
 

Attachments

Marsh

Moderator
May 12, 2013
2,645
1,496
113
This puzzles me.
OS : Ubuntu 16.0.4 , mining software: xmr-stak-cpu.

Created "startxmrcpu.sh" file
#!/bin/sh
sudo sysctl -w vm.nr_hugepages=30
cd /home/fun/bin/
sudo /home/fun/bin/xmr-stak-cpu

When I login , run "startxmrcpu.sh" script,
Hashrate is 276.7H/s ( see E3-1220v3B.jpg )

If starting the same "startxmrcpu.sh" script from /etc/rc.local
Hashrate is 249.5H/s ( see E3-1220v3A.jpg )

Power usage is also few watts higher.

I saw the same situation in many of my systems. Last night, I used systemd to start "startxmrcpu.sh", no difference .
Same issue ( slow hashrate , more power usage ) as starting script using /etc/rc.local

Ideas ????
 

Attachments

compuwizz

Member
Feb 25, 2017
46
45
18
40
Which website is this calculator on? I've searched and haven't found something as nice.

Looks like XMR has been over $20.00 all day.

Finally broke down and bought a kill-a-watt meter.

With two E5-2667 V3 ES cpu's running XMR-stack-Cpu on all 32 threads, two RX 480's running Wolf's Gpu miner, two ssd's, 64 gig's of ram ( 8- 8gig sticks of DDR4 3200) two 200mm fans, one 140mm fan, two 120mm cpu fans I am using 515-524 watts.

Stopping gpu mining its using 318-323 watts.

Stopping the cpu mining its 182-189 watts.

So the cpu's are using ~68 watts each mining

The gpu's are using ~101 watts each.

I have a 750 watt power supply.

So mining XMR it looks like I can easily add one more RX480 and if I add two I'm about 25 watts below the rated capacity of my power supply. Or add two RX 470's with plenty to spare.

Of course mining other coins I would probably need a higher wattage power supply depending on the coins.

I was guessing 550-575 watts and using 550 in my calculations.

So if I add two more identical RX480's :
 

Spartus

Active Member
Mar 28, 2012
323
121
43
Toronto, Canada
Update on the build below:
at 3.8 GHz it draws an extra 42 watts, and generates an extra 95 H/s. With prices where they currently are this is net profitable.

610 + 1640 = 2250 H/s @ ~484 W from wall.

Upgraded main pc to ryzen 1700, thought I might share some numbers

Ryzen 1700 Stock clocks: 515 H/S

2x R9 290 @1025/1400: 1640 H/S

System total: ~2155 H/S

power at wall: 442 W with platinum power supply. (120 V)

4.87 H/W
 

Marsh

Moderator
May 12, 2013
2,645
1,496
113
at what point do you think smaller systems are better/ worse than big ones? More power strip, switch ports and everything.
My goal is coming up with a "Summer Monero Mining Strategies".

My datacenter is my family room and spare bedroom serve as home office.
During winter time, heat generated from mining rigs are just perfect to heat the house. But summer time, not so good.

I lived in San Francisco Eastbay, ( hot day , cool night ). During hot summer day, I would adjusted numbers of low power mining rigs being used according to daytime temperature . The GPU mining rig consumes 240watt, it would be ok during mild summer days.

At night, I would open all the windows , power up the two ASUS 4 nodes servers, each ASUS 4 nodes server consumes 640watt each.
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,513
5,801
113
At night, I would open all the windows , power up the two ASUS 4 nodes servers, each ASUS 4 nodes server consumes 640watt each.
Here is the next question: at what point would it make sense to stick in a rack somewhere?
 

Marsh

Moderator
May 12, 2013
2,645
1,496
113
I avoid committing to fixed monthly cost.

I am treating this a FUN project,
because my other "mining" is done in the stock and real estate market and I am very grateful for the return.
 

DrPeter

New Member
Mar 10, 2017
12
3
3
69
E5-2648L V3 seems like a nice CPU for monero mining with 30MB cache.

Did anyone try it? (On another site two CPU are said to do 980 H/s would be 490 per CPU if that's true, not bad given the power consumption of those)