Topic: [ANN] cpuminer-opt v3.14.2, open source optimized multi-algo CPU miner - page 10. (Read 10284 times)

Physical memory limits the number of threads but adding more memory may not increase hashrate. Try different thread/affinity configurations
to find the sweet spot, half the number of threads with alternating affinity (0x55555555) seems to work best in many cases.

Hi JayDDee and everybody else

Im trying to setup my cpuminer-opt in HiveOS, mining on a Ryzen 5950x, XBTX coin on ScryptN algo, 1048576.
The thing is that the cpu has 16cores and thus 32 threads, but, im not able to get the miner working with more than lika 18  threads, doing around 60h/s, but, with lower amount of threads i get same hashrate (higher / thread)...
Do you have nay idea why like only using 50% of available threads?

Best regards!
// Pontus

Hyperthreading (SMT) has only been available on AMD CPUs since Ryzen.
A SMT core is actually 1 1/2 CPU cores. It has 2 register files to manage 2 threads
but only one set of shared execution units. Simply put 2 threads can run simultaneously
if they don't need the same execution units. For example one thread could do an FP operation
while the other is doing integer.

The problem here is it's a multi socket system. They typically aren't used as a monolithic system,
often partitioned by the OS to avoid crossing CPU boundaries to access resources. In effect it
looks like a bunch of virtual machines and the miner is likely only using one.

Multi-CPU support has never been properly tested for mining and likely requires significant system
admin skills to understand how to fully use it. It starts with providing clear and concise technical
information about what the user, miner, and OS are doing.

This only happens with 'Threading Tech' applied ...

Opteron (as with any CPU that uses HyperThreading Type of technology) is classically using only the cores allotted, but with HyperThreading Tech enabled, it uses the SAME cores but different time slices. Which means you have 64 Physical Cores (if this machine is a Quad CPU Machine) and another 64 'SoftWare' Cores - making 128 Cores in total as far as the OS can 'see'. That is IF this is happening here with this machine.

So even though it is showing only 50% usage of all 'SoftWare' Cores, the physical cores are actually being used to close to full capacity. You may get some increase in hashing, but I would assume very little if you set your miner thread count at 128.

You can test it and see, it can't really hurt the Physical CPU's themselves, but the whole OS may slow down considerably so be prepared to wait for things to happen when you click Wink

These Opteron CPU's were incredible in their day. Now the Epyc CPU's show what REAL CPU Power is all about!

#crysx

I believe its 64 threads, its a 16 core AMD Opteron 6378 CPU x4 with 256GB of RAM

How many threads are you using? Some algorithms perform better with fewer threads, typically 50%.

Although this might be the same answer, what if I WANT to use 100% of the CPU.  I have a system with 64 cores and using this with awesome miner and its barely passing 50% usage and I'm getting 10.60 KH/s

There's no way to limit CPU usage except by by reducing the number of threads.

What is command to limit cpu usage if i dont wanna go 100% ?

JayDDee
I tried to send you a private message but I can't. Don't have enough "credibility"

my apology, the proxy does not close connection properly if fail-over pool is not specified. Error on my side

Ok, I do a bit more testing, the proxy is closing the connection, that's why I don't understand it. It doesn't happen with previous versions either.

The proxy needs to close the connection so the miner can detect the error and reconnect properly.

Some weird things happening on algo m7m. Miner compiled locally with proper settings. Miner is running via a local proxy set in another machine.
If there is a disconnection between pool and proxy, this happens:
If I shut miner down and restart it, all goes well again. First time i see something like this.
Any idea? I can troubleshoot if you need me to. I have several machines exactly the same to test if need be.
I set roundrobin very low on purpose on the proxy to test it and can replicate it.
It doesn't happen on 3.12 or 3.14 versions. I recompiled a fresh version of latest, tried 4 times, once it recovered as normal, 3 times it failed.
~~~ few friends told me I will get bored after retirement... I am enjoying every second of it ~~~

 2021-05-04 19:31:00] 819 A714 S0 Rejected 105 B0, 82.304 sec (144787ms)
                      Diff 0.00016911, Block 3040289, Job 3c88
[2021-05-04 19:31:00] Reject reason: Share is above target
                      Hash:   00001719235eb24adf00000000000000ffffffffffffffff
                      Target: 00002000000000000000000000000000ffffffffffffffff
[2021-05-04 19:31:00] CPU temp: curr 93 C max 94, Freq: 0.000/0.000 GHz
[2021-05-04 19:31:30] CPU temp: curr 93 C max 94, Freq: 0.000/0.000 GHz
[2021-05-04 19:31:30] New Block 3040290, Job 3d38
                      Diff: Net 1.7483, Stratum 8, Target 0.00012207
                      TTF @ 8836.83 h/s: Block 9d20h, Share 0m59s
[2021-05-04 19:32:30] CPU temp: curr 93 C max 94, Freq: 0.000/0.000 GHz
[2021-05-04 19:32:30] m7m: 192.168.1.100:3331
                      Periodic Report     5m20s        12h20m
                      Share rate        0.94/min     1.11/min
                      Hash rate          0.00h/s   8068.42h/s   (8836.83h/s)
                      Lost hash rate  8181.51h/s    1274.89h/s
                      Submitted             5          822
                      Accepted              0          714       86.9%
                      Rejected              5          105       12.8%
                      Hi/Lo Share Diff  0.26387 /  0.00012514
[2021-05-04 19:32:30] Share count mismatch: 3, stats may be incorrect
[2021-05-04 19:32:30] New Work: Block 3040290, Net diff 1.7483, Job 3d65
[2021-05-04 19:33:30] CPU temp: curr 93 C max 94, Freq: 0.000/0.000 GHz
[2021-05-04 19:33:30] New Work: Block 3040290, Net diff 1.7483, Job 3d93
[2021-05-04 19:33:36] 823 Submitted Diff 0.00015248, Block 3040290, Job 3d93
[2021-05-04 19:33:37] 820 A714 S0 Rejected 106 B0, 36.251 sec (265387ms)
                      Diff 0.0005164, Block 3040290, Job 3c88
[2021-05-04 19:33:37] Reject reason: Share is above target
                      Hash:   000007907f283395e500000000000000ffffffffffffffff
                      Target: 00002000000000000000000000000000ffffffffffffffff
[2021-05-04 19:33:37] CPU temp: curr 93 C max 94, Freq: 0.000/0.000 GHz
[2021-05-04 19:33:42] New Block 3040291, Job 3dbd
                      Diff: Net 1.7483, Stratum 8, Target 0.00012207
                      TTF @ 8692.41 h/s: Block 9d23h, Share 1m00s
[2021-05-04 19:34:15] 824 Submitted Diff 0.00013701, Block 3040291, Job 3dbd
[2021-05-04 19:34:15] 821 A714 S0 Rejected 107 B0, 24.664 sec (279048ms)
                      Diff 0.0011594, Block 3040291, Job 3cb2
[2021-05-04 19:34:15] Reject reason: Share is above target
                      Hash:   0000035e8856735b9ba0000000000000ffffffffffffffff
                      Target: 00002000000000000000000000000000ffffffffffffffff
[2021-05-04 19:34:15] CPU temp: curr 93 C max 94, Freq: 0.000/0.000 GHz
[2021-05-04 19:34:41] CPU temp: curr 93 C max 94, Freq: 0.000/0.000 GHz
[2021-05-04 19:34:41] New Work: Block 3040291, Net diff 1.7483, Job 3de9
[2021-05-04 19:35:13] 825 Submitted Diff 0.00019973, Block 3040291, Job 3de9
[2021-05-04 19:35:13] 822 A714 S0 Rejected 108 B0, 83.489 sec (254211ms)
                      Diff 0.00048466, Block 3040291, Job 3d0c
[2021-05-04 19:35:13] Reject reason: Share is above target
                      Hash:   0000080f4e9bfde90880000000000000ffffffffffffffff
                      Target: 00002000000000000000000000000000ffffffffffffffff
[2021-05-04 19:35:13] CPU temp: curr 93 C max 94, Freq: 0.000/0.000 GHz
[2021-05-04 19:35:18] 826 Submitted Diff 0.00015702, Block 3040291, Job 3de9
[2021-05-04 19:35:19] 823 A714 S0 Rejected 109 B0, 156.891 sec (102429ms)
                      Diff 0.00015248, Block 3040291, Job 3d93
[2021-05-04 19:35:19] Reject reason: Share is above target
                      Hash:   0000199e6efe49f24200000000000000ffffffffffffffff
                      Target: 00002000000000000000000000000000ffffffffffffffff
[2021-05-04 19:35:29] CPU temp: curr 93 C max 94, Freq: 0.000/0.000 GHz
[2021-05-04 19:35:29] New Block 3040292, Job 3e13
                      Diff: Net 1.8436, Stratum 8, Target 0.00012207
                      TTF @ 8841.81 h/s: Block 10d08h, Share 0m59s
[2021-05-04 19:35:55] CPU temp: curr 93 C max 94, Freq: 0.000/0.000 GHz
[2021-05-04 19:35:55] New Block 3040293, Job 3e3e
                      Diff: Net 1.8436, Stratum 8, Target 0.00012207
                      TTF @ 8838.01 h/s: Block 10d08h, Share 0m59s
[2021-05-04 19:36:22] CPU temp: curr 93 C max 94, Freq: 0.000/0.000 GHz
[2021-05-04 19:36:22] New Block 3040294, Job 3e6b
                      Diff: Net 1.9918, Stratum 8, Target 0.00012207
                      TTF @ 8834.17 h/s: Block 11d04h, Share 0m59s
[2021-05-04 19:36:23] New Work: Block 3040294, Net diff 1.9918, Job 3e95
[2021-05-04 19:36:38] 827 Submitted Diff 0.00016472, Block 3040294, Job 3e95
[2021-05-04 19:36:38] 824 A714 S0 Rejected 110 B0, 38.333 sec (143984ms)
                      Diff 0.00013701, Block 3040294, Job 3dbd
[2021-05-04 19:36:38] Reject reason: Share is above target
                      Hash:   00001c827b55aa83a200000000000000ffffffffffffffff
                      Target: 00002000000000000000000000000000ffffffffffffffff

the tec cooler was a failure.  Great temps if CPU idle or running single core or gaming... complete bust for mining.  result were worse than the AIO 360.  thermal throttling with avx2 (i.e. 100C)... didn't even try avx512.

I'm going to go full custom water loop next attempt. 
Phanteks Enthoo Prime case with a 420 radiator (or a 480) push-pull configuration on top of case with 4 120mm fans on the bottom of case to suck cold air inand maybe and additional two 240 radiators, one on the side and one on the front.

Fair enough, the community is not too bad on it, https://miningpoolstats.stream/raptoreum
and the few miners available are based on your work! The problem is that is better to compile locally to optimize the build for your own hardware.
In my link from github there is the module for gr algo and thought would have been simple for you to check and add it.

I'm not aware of any working ethash CPU miner.

Sorry to disturb.
there was an old version of your miner working with etash ?
I am looking anywhere till cgminer.. to find one etash cpu miner, so please understand if I disturb.

I've stopped trying to add avery algorithm, most of them die off in a few months. I have to see
some potential for a future. Among other things it must be mineable in a major pool with autoexchange.

any chance you can add parameters for Raptoreum coin?
Unfortunately adding new algos is beyond my capabilities.

It's using Ghostrider algo, infos I found:
https://docs.raptoreum.com/#/howtomine

and this miner:
https://github.com/npq7721/cpuminer-gr