Monero Mining Performance

[fractal]

I took a chance on eBay for a "for parts" Z220 system that included a CPU E3-1240 V2 @ 3.40GHz. I bid for the chassis and power supply and got it. Amazingly the processor worked so I got a great deal on it.

Anyway, the motherboard was dead but I put the processor in a SM X9SCL-M and gave it a go using the latest version of Lucas cpuminer-multi.

3 threads, 132 H/s
4 threads, 143 H/s, 37 watts (power_gadget), 68c (i7z)
5 threads, 120 H/s
7 threads, 125 H/s
8 threads, 120 H/s, 40 watts, 66c

It looks like 1*cores, threads/2, cache/2m wins again. I did notice that the OS would occasionally assign the worker threads to the same core and the processor would switch from a multiplier of 36 for 4 cores active to 37 for 3 cores active. I wonder if that extra turbo is kicking in on other tests and whether that makes any difference.

 

[Patrick]

@fractal nice grab!

Interesting that it is throttling a whole core. You could always try changing BIOS power settings.

On the other tests bit I have been using logs for hours of mining so I would likely see that happen if it were occurring. Very few E3 machines in the lab.

 

[fractal]

Ack, typo there. I fixed it. It went FASTER with only three cores active using standard Intel turbo rules. That processor is

" Max TURBO Multiplier (if Enabled) with 1/2/3/4 Cores is 38x/38x/37x/36x"

so it went from 36x when 4 threads were running on 4 cores to 37x when 4 threads were running on 3 cores.

 

[Patrick]

Glad to see the slide has stopped and we are back up to 12USD today!

I got off the mountain today and saw 12.90. Big move!

Over the next few weeks I think I am going to start watching XMR/ USD more closely. By mid-Feb I should be around 100XMR making the swings more noticeable.

I am also going to post more XMR denominated hardware in my next FS thread.

 

[cafcwest]

Patrick,

Question for someone high core counts who is mining - I've pulled the trigger on my 64 core Opteron build as a dedicated Monero miner. 4 sockets, I need to populate with 4 DIMMs. Any reason to believe I couldn't get away with 4 x 4GB?

 

[Patrick]

@cafcwest I have not tried on the Opteron side. On the Xeon D and E5's 1-2 DIMMs per CPU has very little impact on performance. Sub 0.2% which is basically a margin of error. The benefit, of course, is lower power consumption. I am interested to see how well those Opterons do.

 

[Patrick]

Monero over $13 today!

For those that may have seen, there were a number of updates to the docker images. Moved them to the servethehome docker hub account. There are also now images for both minergate and moneropool. CPU images I have working on Docker swarm clusters. NVIDIA GPU images are online but need to test them a bit more before I say they are OK (I am not declaring any of these OK to use until they have processed at least 10,000 shares).

Also, expect a video later this week on getting up and running.

 

[Marsh]

After 7 days ( 12 hours per day ), Blocks:1

Total mined:1.906324324545 Blocks:1
Good shares:354,617 Bad shares:48
What is a block ? Is it equal to XMR?

 

[KioskAdmin]

Doesn't their block mean you've had a client that found a block for the pool?

I think you've got 1.906 XMR

 

[Kybber]

I understand that Monero mining can in some cases be profitable, depending on cost of electricity and heating requirements. However, has wear and tear on equipment been considered? I would be surprised if a CPU running at 100% capacity 24/7 doesn't wear out quicker than one idling most of the time. The added cost of replacing worn out CPUs and GPUs could possibly negate any profit margin. This of course doesn't matter if you're using hired hardware...

Any thoughts?

 

[Patrick]

I understand that Monero mining can in some cases be profitable, depending on cost of electricity and heating requirements. However, has wear and tear on equipment been considered? I would be surprised if a CPU running at 100% capacity 24/7 doesn't wear out quicker than one idling most of the time. The added cost of replacing worn out CPUs and GPUs could possibly negate any profit margin. This of course doesn't matter if you're using hired hardware...

Any thoughts?

It is a bigger issue on consumer GPUs. Server CPUs will last 5-7 years under 100% load with proper cooling. No issues there.

On the consumer GPU side cooling can be an issue and there is a higher failure rate.

 

[Kybber]

So in other words: It should be safe to use Xeon CPUs and Nvidia Quadro/AMD Firepro cards for mining, but don't use Nvidia GTX or AMD Radeon unless the profit margin is high enough to cover the cost of relatively frequent equipment failure.

I burned out a Radeon HD 7950 a few years ago after just a few weeks of mining, and am therefore very hesitant to start over. May have been just bad luck, but still...

 

[Marsh]

I have a GTX750Ti that was replaced by newer model, the card is still works well in GPU mining rig.

The whole mining concept is not profitable due to high CA power cost $$.
I am retire , and have lots of free time to learn fun thing.

The real "mining" is best done in the stock market.

 

[Patrick]

@Marsh did you get the RX 470's working?

BTW some stats:
Shares: Good - 3,421,958, Bad - 298
Total Mined: 26.849565464175 XMR
Blocks: 3

That seems about right as an over time average. Each block is worth under 9 XMR.

 

[Marsh]

I chicken out on buying the RX 470 card, couldn't decided on RX480 vs RX470.
I'll wait for next time when the RX470 card goes onsale.

My 8days x 12hr per day
Unconfirmed:0.002745276390
Total mined:2.110772941472
Good shares:394,679 Bad shares:50
Blocks:1
$27.92

 

[SavageWS6]

So got bored and tested/running it on my E3-1225 v3 @ 3.2

Using this release since this is my Windows storage server: Releases · fireice-uk/xmr-stak-cpu · GitHub

2012 R2 - 32GB Total
EDIT: I upped it to all threads. http://i.imgur.com/tIcKEVI.png

 

[Patrick]

So got bored and tested/running it on my E3-1225 v3 @ 3.2

Using this release since this is my Windows storage server: Releases · fireice-uk/xmr-stak-cpu · GitHub

2012 R2 - 32GB Total
EDIT: I upped it to all threads. http://i.imgur.com/tIcKEVI.png

Any config tips? If I can get stak running easily in a Docker container I will update it. Right now too much config editing.

 

[SavageWS6]

Any config tips? If I can get stak running easily in a Docker container I will update it. Right now too much config editing.

Per the base config, I've only enabled LARGE PAGE SUPPORT since I'm on Windows which gave it a big boost, also per these lines.

"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 },
],

The only change was setting no_prefetch to true from false

but the weird part... On my single E5-2670 v1 system (Pulled 2nd CPU) I'm only getting 325H/s atm on Server 2016.

UPDATE: So, I had to knock it back to 10 threads, I was too tired to mess with it last night. I also unparked cores. So I'm chilling at 389H/s with system mentioned above

 

[SavageWS6]

Quick update, threw in my second 2670v1 cause I was bored. I didn't pull one of the hard drives that is idle 24/7, but I don't think its a big issue regarding getting power numbers.

According to PMBUS: Average: 234 watts with both CPUs. Only 25 threads, anymore and performance goes to crap.

140 watts if I only run 1 CPU.

Hashing Speed: 667-671H/s

 

[Marsh]

Here is my E5-2650 v1 x2 hash rate
Intel S26K chassis, dual e5-2650 x 2 , 257w-277w , 1 node
STH docker CPU miner image: 740-760H/s