Monero Mining Performance

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

Kal G

Active Member
Oct 29, 2014
160
44
28
44
Bare metal, the E3-1240 v1 runs at 256 H/s with 4 threads.

Also bare metal, a dual E5-2630 v1 runs at 547 H/s with 14 threads.

Other thread counts gave lower hash rates.
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
Looks like a crash day in the crypto currency market today with the top 5 or so down >11%.
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
Quad E7-8870 V3

nproc / 2 = 2600H/s
nproc = 1586H/s
nproc -1 = 1584H/s

I am seeing a pattern.
 
Last edited:
  • Like
Reactions: gigatexal

fractal

Active Member
Jun 7, 2016
309
69
28
33
Well now, this IS interesting.

A Dual Xeon X5675 with 12288 KB cache per processor reported the following:

T21 (default): 334
T18 : 365
T13: 450
T12: 495
T11: 478

So 12 threads which coincidentally is T/2 and 1*C and cache/2 -- is the optimal value and produces significantly better results than letting wolf's cpuminer choose for you.
 
  • Like
Reactions: Patrick

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
Well now, this IS interesting.

A Dual Xeon X5675 with 12288 KB cache per processor reported the following:

T21 (default): 334
T18 : 365
T13: 450
T12: 495
T11: 478

So 12 threads which coincidentally is T/2 and 1*C and cache/2 -- is the optimal value and produces significantly better results than letting wolf's cpuminer choose for you.
Yea what I have been doing is making Docker images with scripts that run different thread counts. That seems about right.

On the quad E7 - 800-810W on 208V.
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
I am around 22KH/s now.

BTW on the quad E7 system, the hashrate falls off by around 160 blocks it hits 1944. I wonder if that is thermal related. All CPUs are reporting 40-43C which should not be an issue.
 

Kal G

Active Member
Oct 29, 2014
160
44
28
44
BTW on the quad E7 system, the hashrate falls off by around 160 blocks it hits 1944. I wonder if that is thermal related. All CPUs are reporting 40-43C which should not be an issue.
Could it be backing off from Turbo Boost as the processor heats up?
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
Added 3 more dual processor speeds to the list.

And @gigatexal prices keep falling but now up to 25KH/s so generating over 3 XMR/ day.
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
If anyone is building dedicated miners, I just dropped a E5-2650L V3 node from 128GB RAM using 16x8GB (RAM still was running at 2133MHz) to 32GB in 4x 8GB (2 DIMMs per socket) and here are the takeaways:
  • Mining performance: Still 809H/s (nproc/ 2 = 24 threads), 874H/s (t=30)
  • Power consumption: Down 17W (12 DIMMs removed)
That 17W savings is roughly 8% of total power consumption.

Breakeven on the 1U node is around $0.26kWh electricity costs
 
Last edited:

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
Here is an ugly chart based on what I have seen thus far:
upload_2017-1-15_16-4-46.png

That does not include power consumption (yet) but is at least interesting to look at.

@gigatexal But it requires over $100k of gear and $30K of power / rack space to generate that.
 
  • Like
Reactions: Marsh and gigatexal

Marsh

Moderator
May 12, 2013
2,644
1,496
113
Let me understand the return.

Invest $130K first year to generate $11K ( 8.5% ).
or
Invest $130K in Vanguard Total Stock Market Index Fund Investor Shares VTSMX
Year 2016 return for the fund is 12.53%.
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
Let me understand the return.

Invest $130K first year to generate $11K ( 8.5% ).
or
Invest $130K in Vanguard Total Stock Market Index Fund Investor Shares VTSMX
Year 2016 return for the fund is 12.53%.
Well, this is just running on idle capacity in nodes in the lab now. Things like having 40GbE in each node, multiple enterprise SSDs and etc. are not needed for Monero, but are useful in the lab.

Even spending $2200/ C6220 with 2x E5-2670's and 8GB RAM/ node and $400 on a cheap 1GbE switch you would have about the same Monero compute capacity for around $18K in servers/ networking. The struggle then would be to find ~10kW to make power profitable. You would not need a 1GbE+ WAN link and could get away with a 50mbps link instead keeping costs lower. There are certainly other options available if you wanted to keep costs low.

I think the point is that the dedicated Monero mining is going to be much more cost effective than what I am doing. On the other hand, if you have machines and power you are paying for, but not using, this is a potential way to offset operations costs. At the rate I am doing this, the opportunity is $700 to $1000/ month to help offset lab costs using hardware that is already configured and powered on.

Here is an example @Marsh, if I had access to unlimited numbers of those ASUS 4N2U at the price we paid, plus optimized power/performance CPUs, and minimal RAM, with $5/ 16GB m.2 drives, I would return 2% monthly even using V4 hardware. The downside is that doing so the infrastructure would be worthless for anything else. Alternatively, mining using cards like the RX 480 would be significantly more profitable as some of those have payback (minus power) measured in 8 months or less.

Nov 1 2016 was around $4.75/ XMR and Jan 5 2017 was $18.30 XMR so the fluctuations doing this are too risky for me to devote more concentrated efforts on.
 
  • Like
Reactions: gigatexal