Topic: SRBMiner-MULTI GPU & CPU Miner 0.9.4 - page 16. (Read 44750 times)

V2.2.2
+ Moved CPU part of the POUW work for algorithm 'dynex' on NVIDIA GPU's to the GPU*
+ Better utilization of CPU's that have more threads than running GPU's for algorithm 'dynex'
+ Parameter '--dynex-cpu-pow' now expects GPU id's of devices on which you want to use the old CPU POW method
+ Increased value of auto intensity a little bit under Linux OS for algorithm 'dynex'*
+ Minor bug fixes

***
Good news for NVIDIA GPU's :

A part of POUW work that was done on the CPU's is now done on GPU side.
This unfortunately comes with a penalty : less hashrate
More memory and additional work is needed on the GPU, which unfortunately decreases the total hashrate.

If you have a good CPU which can run all your cards without issues, then use the --dynex-cpu-pow parameter,
list all the GPU id's and the additional work will again be calculated on the CPU side.


***
Info:
Auto intensity (when you don't set --gpu-intensity parameter manually) can set a too high value now, mainly because
on some systems video driver reports more free memory for the GPU than there really is.
Also devices that have a display (monitor) attached will probably need to manually lower intensity.

In this case a few things can happen :

- gpu works but hashrate is lower than expected (previous version)
- gpu errors and won't start working
- gpu crashes

You should manually set the gpu intensity in these cases with --gpu-intensity parameter.
The initial (auto set) value can be seen on miner start.
Decrease that value and do some tests.

Hi, @doktor83, how about cn_gpu/cn_xvh/cn_upx for nvidia?

--gpu-intensity 480 - does not work, "no jobs for mallob/s: error" and the miner closes ..... how to spell it correctly? tried and 360, and other - Does not help

@doktor83, on Radiant I have very high CPU usage on older motherboards and that way the hashrate drops a lot, is there any way to improve it? On new modern CPUs there is no problem of course and your miner is giving the highest Radiant hashrate.

Mallob errors are not miner related, but dynex mallob related.
There are numerous reasons best is to seek help on Dynex discord.

I can't mine with versions and the newest V2.2.1 still the same mallob error at zergpool.
And I get invalid share #414.
And please provide parameters for core and memory clock control.

Same problem, can't start the miner.

"mallob registration error" , no mining DNX ! what to do?

V2.2.1
+ Moved CPU part of the POUW work for algorithm 'dynex' on AMD and INTEL GPU's to the GPU*
+ Added parameter '--dynex-cpu-pow' which uses the old way of 'dynex' hashing
+ Added parameter '--gpu-reset-oc' which resets OC settings to their default values on miner start [NVIDIA]
+ Reduced default value for intensity (number of chips) on algorithm 'dynex'
+ Removed auto-tune for algorithm 'dynex'
+ Bug fixes for algorithm 'dynex'


***
Good news for AMD & INTEL GPU's (and in an upcoming release probably for NVIDIA's) :
A part of POUW work that was done on the CPU's is now done on GPU side.
This unfortunately comes with a penalty : less hashrate [ up to ~10% ]
More memory and additional work is needed on the GPU, which unfortunately decreases the total hashrate.

If you have a good CPU which can run all your cards without issues, then add the --dynex-cpu-pow parameter
and the additional work will again be calculated on the CPU side.

Info:
Auto intensity (when you don't set --gpu-intensity parameter manually) got reduced a little bit, mainly because
on some systems video driver reports more free memory for the GPU than there really is, so mining won't start or
it crashes.
You can manually set the gpu intensity (number of chips) with --gpu-intensity parameter which can lead to higher hashrates.

Does anyone know what to set as msr presets/tweaks on linux for randomx? Trying to dual mine randomepic/rx0 but getting hashrate all over the place, and I've tried 50 different combinations of msr mods, my xmrig hashrate is a stable 14kh/s but with srb it sometimes goes to 14kh then settles back at 8-9kh, kind of annoying as it works great out of the box on windows.

On a ryzen 3900x btw.

Please add an option to lock the core and the memory clocks.

Greetings, dear ones, for some reason the new version of the miner gives a low hashrate on vega video cards - it is the same as on 480 cards. But the electricity consumption is also low. Is there any way to increase the mining speed for AMD VEGA graphics cards? coin dynex

https://prnt.sc/LsKAHgKvcP5g

How much RAM needed for dynex mining?
Dual mining Epic+Dynex not work (RTX3060 12GB):

Hi Doktor,

I have a question about the DNX integration.

On two different 4090 rigs (Win 11), I get this error:

Mallob error: [Couldn't resolve host name]

All other rigs on Win 10 are working. Maybe its the Win 11?

I'am using on all rigs dynex.neuropool.net:19331 and --mallob-endpoint mallob.neuropool.net:8081,https://dnx.eu.ekapool.com


Thanks

"mallob registration error" , no mining!

Nice, i waited for this update 
This will be a long night of testing 

V2.2.0
+ Added algorithm 'dynex' (DYNEX coin) for GPU mining, fee 3.0%* [AMD/NVIDIA/INTEL]
+ Added INTEL ARC support for algorithms : autolykos2, etchash, ethash, ubqhash, blake3_ironfish
+ Added INTEL ARC support for dual mining : autolykos2/kaspa, autolykos2/heavyhash, autolykos2/sha512_256d_radiant, autolykos2/sha256dt, autolykos2/blake3_alephium, autolykos2/blake3_ironfish
+ Added dynex specific parameter '--mallob-endpoint'
+ Fixed 'blake3_ironfish' algorithm [also for DUAL modes]
+ Fixed 'sha256dt' algorithm for Nvidia ADA and HOPPER architectures
+ Added parameter '--disable-ptx-check' [nvidia]
+ Fixed bugs with NVIDIA OC parameters

Dynex:

Multiple pools & mallobs can be used.
There is no 'adj' parameter, to manually set the number of chips use '--gpu-intensity' parameter.

Dynex can be 'dual' mined with ZIL :

You can define multiple pools with the same/different wallets, miner will automatically failover when needed:


Windows users [ AMD GPUS ]:
Dynex POUW computations are long, so you should apply TDR fix [can be found in the Help folder], or else driver crashes might occur.

It is good to have pufferfish2bmb algo for Pandanite mining. Thank you for your work..

I've ran into this same problem with the latest SRBMiner-MULTI using the 20.4 drivers which is newer than 19.x with FiroPow. It seems to think my max buffer size is still 4GB and I'm not sure what to make of the 3004.80 driver version:

[2023-02-18 07:55:33] Starting init of worker threads
[2023-02-18 07:55:33] GPU DeviceID 0 [BUS:03] driver version: 3004.80
[2023-02-18 07:55:33] GPU DeviceID 0 [BUS:03] reported total memory: 8192 MB
[2023-02-18 07:55:33] GPU DeviceID 0 [BUS:03] reported free memory: 8074 MB
[2023-02-18 07:55:33] GPU DeviceID 0 [BUS:03] max buffer size: 4040 MB

If newer drivers are now required, which ones are good? The latest ones really hate bifurcation and everything freezes up for a minute when trying to enumerate them so hopefully I can use older ones that don't freeze up.

Also, I have this problem when trying to use --password with firopow. It says more passwords than pools given, even if I only just set a password and not try to pass in any parameters.

Thanks for all your help and writing this miner for all these years!

Hal