Topic: Building Cheap Miners : My "Secret" - page 3. (Read 60219 times)

Is the 1500 H/s maintained when you are also mining on the GPU's or does it take a hit?

I have just now tested my Dell T5500 with dual X5670's. This is my main workstation and it is powered on 24/7.

At idle the power used is 110 watts.

When mining RandomX it goes to a constant 308 watts. Hash rate is 3124 H/s or 1562 H/s for each X5670 Xeon. So the power used mining RandomX is 198 watts for two X5670 Xeon's or 99 watts for one X5670 Xeon. That makes your original 95 watt number for the X5660 right on the mark.

So my Dell T5500 with dual X5670's rig mining RandomX produces 3124 H/s for an additional power usage of 198 watts which makes 15.78 H/s per watt figure.

THANKS for the information.   How much memory did you have installed for those E5-2640 and E5-2670 tests and how fast is the memory? Also which OS did you use?

Here we go:

IDLE SYSTEM : 86-watts
CPU MINING ONLY (RandomX @ 1500 H/S) : 186-watts
GPU MINING ONLY (1x1060 + 5x1050Ti on X16Rv2) : 465-Watts
CPU + GPU MINING : 550-Watts

But yes, your point is totally accurate if I am only CPU mining -- H/S per watt is not super great with all that idle usage.

Seems to be some synergy here, though... "CPU Only" adds 100-watts of draw, if I'm already GPU mining it only goes up 85-watts more (this is the number I was recalling).

So at my power cost it would run me $0.3122 per day in power to ONLY mine on the CPU -- that only jumps to $0.924 per day to do both ($0.61 more)... and worst case I can make $1 a day on NiceHash with this system GPUs. This is one of my lesser system in terms of GPU earnings; it's just the one I had a monitor hooked up to this morning.

So that is the way I'll go is to stack the RandomX CPU mining on top of the GPU and my *additional* earnings rings in at the ~16 H/S per watt figure.

That's true -- I am thinking only of the CPU draw -- in the past when I measured it I was just measuring what the CPU ADDED to the draw since I was also GPU mining (I will still be doing GPU on everything except the R815s as well).

Thermal Design Power for the Xeon X5660 is 95 watts. The actual power used can be upwards of 50% above this value just for the processor alone.

http://www.cpu-world.com/CPUs/Xeon/Intel-Xeon%20X5660%20-%20AT80614005127AA%20(BX80614X5660).html

Also you will need to consider the power used by the memory, motherboard, GPU and the efficiency of the power supply.

I have tested Dell T5600 workstations with dual Sandy Bridge E5-2640 and dual E5-2670 Xeon's. These are the results:

Dual E5-2640, 12 threads, 4554 H/s, 310 watts: 14.7 H/s/W

Dual E5-2670, 16 threads, 6406 H/s, 357 watts: 17.9 H/s/W

The E5-2640 also has a Thermal Design Power of 95 watts

http://www.cpu-world.com/CPUs/Xeon/Intel-Xeon%20E5-2640.html

Assuming that the two of then used 190 watts that still leaves 120 watts used for the memory, motherboard, Quadro 600 GPU and the 80% efficiency of the power supply.

Sandy Bridge processors are more efficient vs the Westmere processors which the X5660 is.

For actual power used you really need to measure at the wall.

Your guess of "16 H/S per watt" is not realistic.

I am hoping this Monero fork to RandomX will make my decision to own around 100 pcs Xeon X5660 worthwhile, haha!

I am getting close to 1500 H/S each on them... I forgot my power data (need to re-measure) but that is *under* the 90 watts max of the CPU... so 16 H/S per watt is the minimum.

I believe I have 8 Gigs per CPU on nearly all of them (except one that has a bad DIMM slot). I am in the process of moving buildings so I hope I don't miss out on the initial Monero rush =)

I am using XMRig for testing RandomX now and will use it for production when the fork happens November 30th. I am so done with XMR-STK with the abusive nature of the developer.

XMRig is top notch. You can also compile from the source and set the donation level to 0% if desired.

xmrig 4.5.0-beta (unified 3 in 1 miner, CPU+OpenCL+CUDA)
https://github.com/xmrig/xmrig/releases/tag/v4.5.0-beta

Here is what I do for testing right now:

RandomX testing


One thing to note that I have seen on the quad 16-core 6378 Opteron's is that disabling one core in each of the eight NUMA nodes increases the hash rate slightly while reducing power and heat slightly. The same reduction of one core per NUMA node may also be beneficial on your quad 16-core 6380 Opteron's.

Also note that you will need at least 8GB per Opteron or 32GB minimum for each R815 with quad Opterons. RandomX needs 2GB+256MB per NUMA node. Each Opteron has two NUMA Nodes so that equals 4.5GB per Opteron. That means that 4x 1GB sticks per Opteron will not work. You will either need 8x 1GB sticks or 4x 2GB sticks per Opteron at the minimum.

I am gearing up with twelve Dell R815 Servers. They are a mix of quad 6348 and 6378 Opteron's.

Two Dell R815's with 4x 6348's and two R815 with 4x 6378's produce:

43,892 H/s on RandomX while drawing 13.65 Amps @ 240 VAC or 3276 watts at the service panel.
That is an average of 819 watts per server.

I measured after mining constantly for 30 minutes so the Hash Rate and Power Usage numbers have stabilized along with the server fan speeds.

The quad 6348's produce: 10684 H/s while the quad 6378's produce: 11262 H/s.

The 43,892 H/s @ 3276 watts is 13.4 H/s/Watt.

Looking forward to the November 30th fork.

Great info here. I've had my R815s turned totally off for a bit... all Quad 6380 machines.

Looks like may have to turn them on again when Monero forks. Doesn't seem profitable for LOKI.

On the defective CPU be sure the pads on the bottom are clean and that no thermal residue is on any of those pads. Some of the ones I bought had thermal residue on the pads.

I examine the pads using a high power bright light and a magnifying glass.

Use Isopropyl alcohol to clean the bottom pads on the processor and canned air to dry the pads.

Gave up trying to get the benchmark to work. Went straight to xmrig 4.2.1.

edit/update: Thanks for the detailed guides in this thread. Using turion and the custom bios for overclocking, was able to get 12.8khs. Very happy.

I am building out another identical setup but after flashing the bios, CPU2 does not working. Meaning, if I take the CPU out of CPU2 slot, it boots. If I put it back in, it doesn't boot. I tried flashing the original stock bios but that didn't help either. No bios beeps, no screen, nothing with CPU2 in. I checked for bent pins but don't see anything. Have you experienced that with any of your work?

For now, I will have to just use 3 CPUs which seems a little like a waste but better than nothing I guess...

I fixed the invalid node by swapping ram into all blue slots instead of all white. Now, still having the issue with having to enable --softAes. I checked the bios again but nothing jumped out. Doesn't the 6378 support AES?

Update: Just kidding. Still having 5 and 6 invalid now.

lscpu
Architecture:        x86_64
CPU op-mode(s):      32-bit, 64-bit
Byte Order:          Little Endian
CPU(s):              64
On-line CPU(s) list: 0-63
Thread(s) per core:  2
Core(s) per socket:  8
Socket(s):           4
NUMA node(s):        8
Vendor ID:           AuthenticAMD
CPU family:          21
Model:               2
Model name:          AMD Opteron(tm) Processor 6378
Stepping:            0
CPU MHz:             1444.930
CPU max MHz:         2400.0000
CPU min MHz:         1400.0000
BogoMIPS:            4800.29
Virtualization:      AMD-V
L1d cache:           16K
L1i cache:           64K
L2 cache:            2048K
L3 cache:            6144K
NUMA node0 CPU(s):   0-7
NUMA node1 CPU(s):   8-15
NUMA node2 CPU(s):   16-23
NUMA node3 CPU(s):   24-31
NUMA node4 CPU(s):   32-39
NUMA node5 CPU(s):   40-47
NUMA node6 CPU(s):   48-55
NUMA node7 CPU(s):   56-63


numactl --show
policy: default
preferred node: current
physcpubind: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63
cpubind: 0 1 2 3 4 7
nodebind: 0 1 2 3 4 7
membind: 0 1 2 3 4 7

That cleared up the turion errors and appears to be working.

Still getting this error:
seq 0 7 | xargs -P 0 -I node numactl --localalloc -N node ./randomx-benchmark --mine --largePages --jit --nonces 100000 --init 8 --threads 8 --softAes
libnuma: Warning: node argument 0 is out of range

<0> is invalid
<3> is invalid
<6> is invalid

Also, still required to have --softAes on the command line. Anything ideas about that?

Instructions for installing TurionPowerControl from Source Code:

https://github.com/mh0rst/turionpowercontrol/releases

Install TPC:

https://hardforum.com/threads/multiple-processor-g34-checklist.1669698

Install TurionPowerControl


So that you don't have to do the modprobe's after you reboot do this:

sudo Edit /etc/modules file and add these lines:
 cpuid
 msr

------------------------------------------

To Run TurionPowerControl commands at Boot:

sudo Edit /etc/rc.local

add /home/miner/turionpowercontrol/TurionPowerControl -psmax 1
add /home/miner/turionpowercontrol/TurionPowerControl -htctemplimit 75

and any other TurionPowerControl commands you would like to run at boot

Change the "/home/miner/turionpowercontrol" in the above line(s) to whatever directory TurionPowerControl is installed in.

------------------------------------------

sudo ./TurionPowerControl -l

sudo ./TurionPowerControl -psmax 1

sudo ./TurionPowerControl -spec

sudo ./TurionPowerControl -temp
sudo ./TurionPowerControl -mtemp

sudo ./TurionPowerControl -dram

sudo ./TurionPowerControl -gettdp

sudo ./TurionPowerControl -CM

sudo ./TurionPowerControl -htc

// To Prevent Throttling when Garage Mining
sudo ./TurionPowerControl -htctemplimit 75

You need cpuid installed:
sudo apt install cpuid

And you need cpuid and msr loaded:
sudo modprobe cpuid
sudo modprobe msr

To have cpuid and msr loaded on boot:

Edit /etc/modules file and add these line to it:
 cpuid
 msr



First make sure you have enough LargePages:

sudo sysctl -w vm.nr_hugepages=9600
cat /proc/meminfo

also make sure you have numactl installed:
sudo apt install numactl

For a Quad Opteron 6378 system this is the command:

seq 0 7 | xargs -P 0 -I node numactl --localalloc -N node ./randomx-benchmark --mine --largePages --jit --nonces 100000 --init 8 --threads 8

I pulled down the turionPowercontrol but getting this error when I run it:

sudo ./TurionPowerControl -dram
TurionPowerControl 0.44-rc2 (tpc-0.44-rc2-r144)
Turion Power States Optimization and Control - by blackshard

cpuid:open: No such file or directory
cpuid:open: No such file or directory
cpuid:open: No such file or directory
cpuid:open: No such file or directory
cpuid:open: No such file or directory
No supported processor detected, sorry.

Quad opteron 6378 on ubuntu 18

Also, I can't run the benchmark without using --softAes. Why is that? I see your logs show hardware AES. Is that something in the bios I need to setup?

Also, Getting this error:

seq 0 1 | xargs -P 0 -I node numactl -N node ./randomx-benchmark --mine --largePages --jit --nonces 10000 --init 8 --threads 8
libnuma: Warning: node argument 0 is out of range

can opteron 6376 be profitable at .18/kwh?

You are a troll f*cking up this useful thread, shilling a scam, right?