[ANN] cpuminer-opt v24.5, Optimized multi-algo CPU miner for x86_64 and AArch64

[JayDDee]

I don't think he said "at" work. He said "to" work.

Where else would you find a cluster? The point being it's a cluster issue, not an application issue.
If the application needs to be modified to run in a cluster, too bad.

[JayDDee]

cpuminer-opt-3.18.1

More speed for scrypt:
 - additional scryptn2 optimizations for all CPU architectures,
 - AVX2 is now used by default on CPUS with SHA but not AVX512,
 - scrypt:1024 performance lost in v3.18.0 is restored,
 - AVX512 & AVX2 improvements to scrypt:1024.

Big speedup for SwiFFTx AVX2 & SSE4.1: x22i +55%, x25x +22%.

Issue #337: fixed a problem that could display negative stats values in the
first summary report if the report was forced prematurely due to a stratum
diff change. The stats will still be invalid but should display zeros.


Although, in my testing on a i5-2400, scryptn2 AVX is faster than v3.18.0 it's still slower than 3.17.1.
I have no idea why. It uses essentially the same code for AVX512, AVX2, AVX & SSE4.2. When I do a
benchmark on a bigger newer CPU there is not a significant drop from AVX2 to AVX, in fact less than
the lost AVX performance from v3.17.1 to v3.18.0. I can't explain this.

If any users have AVX CPUs it would be appreciated if you could run a quick test to either confirm or
contradict by results. Either way would be a step forward to understanding this mystery.

[JayDDee]

Here's an arcticle about enabling AVX512 [& VAES] on Alderlake CPUs running Linux.
Unfortunately there were no details about compiling. "Official" Alderlake should
be compatible with, ironically, Zen1. With AVX512 [& VAES] it should be compatible with
Rocketlake [Icelake].

https://www.phoronix.com/scan.php?page=article&item=alder-lake-avx512&num=1

In a seperate arcticle it was mentioned that Intel has yet to supply Thread Director patches
for the Linux kernel.

Any sharing of user experience with Alderlake would be appreciated.

"Tom" also did a comparison test of the top end Alderlake with Ryzen 5900x & 5950X. Although Alderlake
won overall in Gaming performance using Win11 with Alderlake optimizations, Ryzen had better multithreaded
performance using less power.

Edit: corrected for VAES which is also available but not mentioned in the arcticle.

[JayDDee]

cpuminer-opt-3.19.0

Note: updated Windows binaries for CPUs with AVX or SSE* due to problems with CPU groups.
These builds have affinity limited to 64 CPUs but may still work with XP.

---

Windows binaries [for CPUs with AVX2 and above] are now built with support for CPU groups, requires Windows 7.

Changes to cpu-affinity:
  - PR#346: Fixed incorrect CPU affinity on Windows built for CPU groups,
  - added support for CPU affinity for up to 256 threads or CPUs, [AVX2 and above]
  - streamlined code for more efficient initialization of miner threads,
  - precise affining of each miner thread to a specific CPU,
  - added an option to disable CPU affinity with "--cpu-affinity 0"

Faster sha256t with AVX512 & AVX2.

Added stratum error count to stats log, reported only when non-zero.

[tomganleylee]

So i am having some issues trying to get one of my machines minning.

System:
OS: Windows server 2016
CPU: 5800x

I keep getting this error:

Prepared for Windows - NTver: 0x601
CPU: AMD Ryzen 7 5800X 8-Core Processor
SW built on Nov 10 2021 with GCC 10.0.0
CPU features:  AVX2    AES
SW features:   AVX2   VAES SHA
Algo features: AVX2   VAES
The SW build requires a CPU with VAES!

anyon have any idea whats wrong, i have 5800x and 5900x working fine on other machines

Thanks in advance

[JayDDee]

So i am having some issues trying to get one of my machines minning.

That's someone else's fork, can't help you.

[rambetiko]

what is EPYC performance? i have 2 pcs  7F52 and 2 pcs 7F32

[JayDDee]

what is EPYC performance? i have 2 pcs  7F52 and 2 pcs 7F32

https://openbenchmarking.org/test/pts/cpuminer-opt-1.4.0

[Cryptomining2003]

question for RTM setup:

whats the impact of ram for the rtm hashrate?

more mhz is as important as low latency or more important or is latency more important?


3600 better than 2666, sure
cl13 better than cl16, sure, but what about:

3600 cl16 vs 2666 cl13 e.g

so whats more important, mhz or latency? big impact on price

cpu is ryzen 9 5900

[JayDDee]

question for RTM setup:

Wrong place, no RTM support here.

[DRFTA]

Just Loaded up your miner on three of my pc's with the default tuning taking around 2 hours on a Ryzan 5 3600 and a I7 7700k.
My 3960x Threadripper took 8 hours to get through the default tune.... I'm assuming because it has 48 threads and not 12 like the AMD 3600 its taking 4x longer?
or am I missing something that's not quite right?

Thanks!   

[Cryptomining2003]

question for RTM setup:

Wrong place, no RTM support here.

where s the right place?

[JayDDee]

where s the right place?

From wherever you got the miner, it's not mine.

[JayDDee]

New in v3.19.1

Changes to Windows binaries package:
 - builds for CPUs with AVX or lower have CPU groups disabled,
 - zen3 build renamed to avx2-sha-vaes to support Alderlake as well as Zen3,
 - zen build renamed to avx2-sha, supports Zen1 & Zen2,
 - avx512-sha build removed, Rocketlake CPUs can use avx512-sha-vaes,
 - see README.txt for compatibility details.

Fixed a few compiler warnings that are new in GCC 11.
Other minor fixes.

[robminer80]

Which argon2 algo should I select to mine TRTL (argon2/chukwav2)?

[JayDDee]

Which argon2 algo should I select to mine TRTL (argon2/chukwav2)?

cpuminer-opt doesn't support that variation of argon2, you need a different miner.

[robminer80]

Which argon2 algo should I select to mine TRTL (argon2/chukwav2)?

cpuminer-opt doesn't support that variation of argon2, you need a different miner.

No plans for future support?

[msduzaman]

Will you please add cn/rwz algo :)Please

[SecureFreak]

I have a 5900x mining QOGE but can't seem to get it to use axv2-sha-vaes. It always starts with SSE2 SHA. What am I missing?

https://imgur.com/a/vwMWtiq

[JayDDee]

I have a 5900x mining QOGE but can't seem to get it to use axv2-sha-vaes. It always starts with SSE2 SHA. What am I missing?

The algorithm is limited to SSE2, note "algo features".