[LOCKED] cpuminer-opt v3.12.3, open source optimized multi-algo CPU miner

[ZenFr]

sorry for the dumb question, but if i download the files there is no *.exe to start. Where can i find it?

What do you mean by "files"? there is only one zip file to download for Windows.

You have to unzip the downloaded zip file.

[1714940359]

1714940359

[1714940359]

1714940359

[buckya]

Any chance for large pages support for Windows ?

[joblo]

Any chance for large pages support for Windows ?

This was asked before. I won't make any application changes to support huge pages, it's an OS issue.

[spronkin1]

What is the default number of threads?

[buckya]

What is the default number of threads?

As many as your CPU have them.

[xs.over]

cpuminer-opt on four socket system don't see more than one phusical CPU.
OS win server 2012 r2
How to run miner on multi socket systems?

[joblo]

cpuminer-opt on four socket system don't see more than one phusical CPU.
OS win server 2012 r2
How to run miner on multi socket systems?

It's the responsibility of the OS to present the available logical CPUs to the applications.
The miner can't use what it can't see. It's an OS issue.

[antaja]

i dont know if its an issue or is it my cpu, but ver 3.6.1 always stopped mining after a while, like it stuck after about 30m-1hour mining

[xs.over]

The miner can't use what it can't see. It's an OS issue.

OS - win server 2012 r2 see all four NUMA nodes ie 4 phusical CPU, but miner works only with one CPU. What I'm doing wrong?
This miner dodn't have even "cpu_threads_configuration" like XMR STAK CPU miner have  

[joblo]

i dont know if its an issue or is it my cpu, but ver 3.6.1 always stopped mining after a while, like it stuck after about 30m-1hour mining

I assume there was no error message or you would have mentioned it.
There was an issue reported on one algo where the miner would exit after 50 share submits
if stats were enabled. Stats are disabled by default so this is unlikely.

Most other SW issues would cause an exit much sooner.

Could be CPU or RAM.

[joblo]

The miner can't use what it can't see. It's an OS issue.

OS - win server 2012 r2 see all four NUMA nodes ie 4 phusical CPU, but miner works only with one CPU. What I'm doing wrong?
This miner dodn't have even "cpu_threads_configuration" like XMR STAK CPU miner have  

How would I know what you're doing wrong if you don't tell me what you're doing?

And if you want to start talking trash, just go away and use STAK.

[spronkin1]

If you have one hyperthreaded Intel CPU the default affinity should be correct for any thread count.
Just find the number of threads that gives the best performance.

Affinity should only be a concern with AMD CPUs that have a different way of implementing multicore HW.
As the previous discussions have shown it's still in the learning stage.

Otherwise your calculations are correct.

I tried a different number of threads without cpu-affinity. This gives a worse result than with cpu-affinity. Example:

Code:

:: c:\Miners\cpuminer-opt-3.6.1-windows\cpuminer-aes-avx2.exe -a zoin -o stratum+tcp://hexx.suprnova.cc:2876 -u spronkin1.Rig1 -p x --cpu-priority 2 -t 1
:: -t 1=478H/s
:: -t 2=660H/s
:: -t 4=1230H/s
:: -t 5=1530H/s
:: -t 6=1470H/s
:: -t 7=1600H/s
:: -t 8=1472H/s
:: -t 9=1423H/s
:: -t 10=1312H/s
:: -t 11=1220H/s
:: -t 12=1210H/s
:: -t 16=1070H/s
:: -t 20=1000H/s
:: -t 23=900H/s
:: -t 7 + --cpu-affinity 0xffffff= 1712H/s - why?:
:: c:\Miners\cpuminer-opt-3.6.1-windows\cpuminer-aes-avx2.exe -a zoin -o stratum+tcp://hexx.suprnova.cc:2876 -u spronkin1.Rig1 -p x --cpu-priority 2 -t 7 --cpu-affinity 0xffffff

Don't use affinity with default thread count, it just messes things up.

Try different affinities with 12 threads. The default should be the same as 0xffffff.

Here are the results I received:

Code:

:: 0d4095=0b111111111111=0xfff=1230H/s
:: 0d8191=0x1FFF=1240H/s
:: 0d16383=0x3FFF=1240H/s
:: 0d32767=0b111111111111111=0x7FFF=1231H/s
:: 0d33023=0b1000000011111111=0x80FF=1450H/s
:: 0d33151=0b1000000101111111=0x817F=1460H/s
:: 0d33167=0b1000000110001111=0x818F=1763H/s
:: 0d33168=0b1000000110010000=0x8190=1850H/s
:: 0d33169=0b1000000110010001=0x8191=2140H/s
:: 0d33171=0b1000000110010011=0x8193=2200H/s
:: 0d33175=0b1000000110010111=0x8197=2200H/s
:: 0d33199=0bx1000000110101111=0x81AF=1700H/s
:: 0d33279=0b1000000111111111=0x81FF=1430H/s
:: 0d16777215=0b111111111111111111111111=0xffffff=1234H/s (as default affinity)

While I chose this result:
c:\Miners\cpuminer-opt-3.6.1-windows\cpuminer-aes-avx2.exe -a zoin -o stratum+tcp://hexx.suprnova.cc:2876 -u spronkin1.Rig1 -p x --cpu-priority 2 --cpu-affinity 0x8197
I did not check the hashrate with --cpu-affinity outside this range between 33279 and 65535 or more. I do not know, maybe there are better results. I just can not understand the principle of the -cpu-affinity parameter, how does it work?

[antaja]

i dont know if its an issue or is it my cpu, but ver 3.6.1 always stopped mining after a while, like it stuck after about 30m-1hour mining

I assume there was no error message or you would have mentioned it.
There was an issue reported on one algo where the miner would exit after 50 share submits
if stats were enabled. Stats are disabled by default so this is unlikely.

Most other SW issues would cause an exit much sooner.

Could be CPU or RAM.

yes there are no error message, now im pretty sure its a hardware issue, ive got an error message about ram today when shuting down

[joblo]

:: -t 7 + --cpu-affinity 0xffffff= 1712H/s - why?:
I just can not understand the principle of the -cpu-affinity parameter, how does it work?

0xffffffff should be the same as the default, some variation in hashrate is expected.

Google affinity mask.

It doesn't seem to be necessary on Intel CPUs, still trying to understand AMD.

[spronkin1]

I tried "-t 12 --cpu-affinity 0x555" (0b010101010101) and got a more high and stable result:
max 2300H/s, avg 2100 H/s on unlocked 2678v3

[joblo]

I tried "-t 12 --cpu-affinity 0x555" (0b010101010101) and got a more high and stable result:
max 2300H/s, avg 2100 H/s on unlocked 2678v3

That's strange you only had 6 bits in the mask so you're running 2 threads on each CPU.
Try 0x555555. On my 6700K I get about the same with 4 of 8 threads with default mask or 0x55.

[spronkin1]

I tried "-t 12 --cpu-affinity 0x555" (0b010101010101) and got a more high and stable result:
max 2300H/s, avg 2100 H/s on unlocked 2678v3

That's strange you only had 6 bits in the mask so you're running 2 threads on each CPU.
Try 0x555555. On my 6700K I get about the same with 4 of 8 threads with default mask or 0x55.

On each core or on each processor? I only have one Intel Xeon processor.
With the value 0x555555 I got 1220 H / s

[joblo]

I tried "-t 12 --cpu-affinity 0x555" (0b010101010101) and got a more high and stable result:
max 2300H/s, avg 2100 H/s on unlocked 2678v3

That's strange you only had 6 bits in the mask so you're running 2 threads on each CPU.
Try 0x555555. On my 6700K I get about the same with 4 of 8 threads with default mask or 0x55.

On each core or on each processor? I only have one Intel Xeon processor.
With the value 0x555555 I got 1220 H / s

I have no explanation for your results with a Xeon. My results differ.

The code will use the mask as a set of available logical CPUs to choose from. If there are more threads
than avaialable CPUs it is supposed to assign multiple threads per CPU.

Your results contradict that logic.

If you have the ability to monitor CPU usage on a per thread basis it may help to understand what's going on.
I'm not familiar with what tools are available on Windows but on Linux I can monotor CPU usage and core
temprerature. CPU usage monitors the logical CPUs while core temperature indicates how hard the physical
cores (hyperthreaded) are working.

[xs.over]

How to configure miner on shitty Win server 2012 r2?
System: 4 socket x Xeon E7 15 cores.
Task manager shows that miner see only one CPU

[th3.r00t]

shitty Win server 2012 r2
System: 4 socket x Xeon E7 15 cores.

There is an easy solution - switch to Ubuntu Server.