SRBMiner-MULTI AMD & NVIDIA & INTEL & CPU miner

[CallMeHog]

V3.6.2
+ Improved ...

Today up from 3.5.4 to 3.6.2
Miner shows backend error 1013.
I try downgrade to 3.6.1 and see same behavior
https://ibb.co/1JQZ7K7x
cards in order 5060 4080s 5060ti x3

[discoslap]

Hello,

with AMD drivers 6.4.4R(6.19.2606) Core Voltage dont work with last release 3.6.2. Use Stock Voltage per Frequency.

[doktor83]

V3.6.7
+ Improved algorithm 'quantus' on NVIDIA 2000, 3000, 4000 and 5000 GPUs
+ Improved algorithm 'noid' (Parano1d) for NVIDIA 3000, 4000, 5000 GPUs
+ Lowered devfee on algorithms 'noid' and 'quantus' to 2%
+ Auto tune is now re-run if miner version changes
+ Bug fixes

V3.6.6
+ Improved algorithm 'quantus' on NVIDIA SM70, SM90, SM100, SM103, SM120 GPUs
+ Improved algorithm 'quantus' on AMD GPUs

V3.6.5
+ Improved algorithm 'quantus' for all supported NVIDIA GPUs
+ Removed algorithm 'neuromorph'

V3.6.4
+ Added algorithm 'quantus' (Quantus Network) for NVIDIA and AMD RDNA GPUS, fee 2.5%*
+ Added parameter '--tls-cert-sha256' for Quantus QUIC protocol

***
'quantus' supports 2 types of pool connections : default QUIC and LuckyPool stratum.
Both use --pool and --wallet parameters, but for QUIC compatible pools you must also add --tls-cert-sha256 parameter

Can be mined on quantus.luckypool.io [ eu.lproute.com:5660 ], quantus.suprnova.cc [ quantus.suprnova.cc:7072 ], or any other pool that supports the QUIC protocol

Tip 1 : use --gpu-mclock 810
Tip 2 : On 5000 series NVIDIA GPU use driver 580+ for better performance


V3.6.3
+ Added algorithm 'noid' (Parano1d) for NVIDIA 3000, 4000, 5000 + H100, B200, B300 GPUs and AMD RDNA 2,3,4. fee 3.0%*
+ Added parameter '--noid-no-pause' which keeps the GPU loaded and avoids stop/resume transitions
+ Minor performance improvements to all 'randomx' based algorithms

***
'noid' on NVIDIA requires a CUDA13 capable driver (driver 580+)
It can be mined on noid.suprnova.cc [ noid.suprnova.cc:3337 ] (innovlab pool also works)
Tip : use --gpu-mclock 810

[dong8768]

Please increase the hash rate for AMD gpus  on 'quantus'

[pattanan.jumbo]

can anyone point me on how to unlocked cmp cards? I am on hive/linux, thx

[sssobp]

**SRBMiner 3.6.8 and 3.6.9 exit immediately with NAUGHTY on Vast.ai**

I'm running Ubuntu 24.04.4 LTS inside a Vast.ai container with 8 × NVIDIA RTX 5090 GPUs.

Both SRBMiner 3.6.8 and 3.6.9 exit immediately with `NAUGHTY`, even when running `--list-devices`.

Tests completed:

* Verified the v3.6.9 official MD5 checksum.
* Confirmed all eight GPUs are visible through nvidia-smi.
* No missing libraries reported by ldd.
* Tested as root and ubuntu.
* Removed LD_PRELOAD, LD_PRELOAD_ENV and LD_LIBRARY_PATH.
* Paused XMRig.
* Seccomp mode is 2, with one filter active.

PeakMiner works correctly on all eight GPUs in the same container.

Is SRBMiner detecting the virtualized/container environment, and is there an officially supported way to run it on Vast.ai?

[MrMens]

Good day!

I was on 3.6.4, mining PRL:

https://i.postimg.cc/s1mXwtHh/3-6-4.png

When the update for PRL 3.6.8 was released, I decided to update it, but an error occurred:

https://i.postimg.cc/Cd71mWrG/3-6-8.png

3.6.9 - nothing has changed:

https://i.postimg.cc/0rnQcLXD/3-6-9.png

3.7.0 - nothing has changed:

https://i.postimg.cc/t7kT2LvF/3-7-0.png

I see that the error is widespread, not just with the 3080 Ti; I’d really like to see a solution to the problem…

Thank you in advance.

[energy_xxx]

1660 ti 1660 süper noid ?

[lynxchat]

3.6.8+ fails on pre-AVX host CPU with F03/C04. Bisected to 3.6.8, looks like AVX baseline — confirm?


Symptom

3.6.8 and upper exits immediately, before GPU init — same for --help / --version:

Code:

Runtime diagnostic code: F03
Runtime diagnostic code: C04
Unsupported runtime environment.

Setup

Code:

RaveOS 1.0.26 (glibc 2.31, kernel 5.10.27-mining-mod)
CPU: Intel Celeron 847 — SSE4.2 max, no AVX/AVX2/AES (full flags checked)
GPUs: 8x RTX 3070 Laptop, driver 525.116.04, CUDA 12.0
Mode: --disable-cpu --algorithm pearlhash

Bisection (same --version probe, 8s + kill safeguard):

Code:

3.6.7 -> runs (no diagnostic codes, drops into guided setup)
3.6.8 -> F03 + C04
3.6.9 -> F03 + C04
3.7.0 -> F03 + C04
3.7.1 -> F03 + C04

3.6.8 is the "pearlhash improvement for NVIDIA 3000 series" release, so our guess is the Linux binary got rebuilt there with a new baseline (probably AVX2) plus an explicit runtime check. Supporting details: both binaries declare the same GLIBC max (2.30, system has 2.31), ldd is clean, and it fails before touching libcuda/nvidia-ml — so probably not a driver/CUDA issue, despite the popular "update your video driver" version.

Question

doktor83, can you confirm what F03/C04 check and what the minimum host CPU level is for 3.6.8+? And is there any path for pre-AVX host boards besides replacing hardware — e.g. a legacy SSE4 Linux build, a separate GPU-only binary, or a flag to bypass the check in --disable-cpu mode? There is still a lot of this host iron around (Celeron/Atom boards doing nothing but feeding GPUs), so a compatible branch would be useful. If that's off the table, just say so and we'll stay on 3.6.7.

[P00P135]

3.7.2 Is working very well on Quantus.  Still some invalids but pool hashrate is so much better than other miners I've tried.

[alphaTigris]

Is Epic mining broken? For several days, the miner has been trying to connect to the address 209.58.168.166 (sg.epicmine.io), but this address is currently unavailable (and it’s not listed among the pools on epicmine.io). Therefore, mining stops due to the unavailability of the dev‑fee.

[random8191]

pearl.

1-2 days ago there was a message from single-message-user about switching to another miner due to non-starting of miner due to unavailability of devfee server but later he have deleted it.

Same was for me.

At another ISP there is no problem (with relatively old verision of miner and new), but at my main ISP last days it works and not works several times per day.

3.7.5 now is working at bad ISP. (Mining is fully legal at my country.)

Accusing/punishing of user for unavailability of devfee server is illogical; no harm to anybody/anything if user mines in this case. Remove this stop-trigger.

Suggestion: make standalone app (to not bloat miner) to trace rout to devfee servers.

[Salens]

Great job on the performance improvements on Quantus—after the 3.7.3 update, the 3080 is hitting 420 MH at 135W. Is it possible to implement this for the CMP 90HX? Performance there is still the same as it was before the 3.7.3 update.

[P00P135]

Version 3.7.5 has a lot less stale/reject shares on Quantus and my farm has seen a huge jump in total Hashrate since 3.7.1. 

[egocentr]

On a 5070 Ti (Windows), `--gpu-mclock 810` doesn't work; it freezes and throws an error. On AMD cards in Windows quantus , power consumption fluctuates; there are drops and frequency dips.

[angelbbs]

Code:

SRBMiner-MULTI v3.7.7
Removed algorithms 'sha3t', 'blake3_an', 'argon2id_exfer'

sha3t is now very profitable on zpool.ca

[egocentr]

On Windows with an AMD gpu , ever since version 3.7.3, there have been constant performance drops. I’ve determined that this is due to cyclical CPU load spikes.

[egocentr]

On Windows with an AMD gpu , ever since version 3.7.3, there have been constant performance drops. I’ve determined that this is due to cyclical CPU load spikes.

and noticed that auto-tune isn't working

[P00P135]

On Windows with an AMD gpu , ever since version 3.7.3, there have been constant performance drops. I’ve determined that this is due to cyclical CPU load spikes.

and noticed that auto-tune isn't working

I havent tried the tuner, but i'm pretty sure its for Nvidia gpu's only.  Also a new version dropped v3.8.2.

[batsonxl]

Can someone help with the problem. I dont know what else to do.
I have a rig with mixed rtx 3000 series on windows10.
Mining pearl starts with error open_cl invalid id. My other rig same windows same drivers dont give this error. Same motherboard. I totally ddu drivers windows new installed still error. Rig mines like couple of hours then stops crash blue screen. My other rig runs days no problem.
Can someone please tell me what is the problem?