[JCE]Fast & stable CN/v8/Heavy/Tube/XHV miner, CPU+GPU, Vega56 1800+ RX580 1200+

[NecronomicoN]

Hello,
I need some help,
I have 4 RX 580 nitro + (8go) , 1 hynix , 3 micron . I was mining with SRB at 850h/s each( V8 xmr)

I did this to increase my hashrate.
- Custom build with SRB Polaris ( I just pressed PIMP my strap then save BIOS and flash)
- then I DDU my old driver and install new one 18.6.1.
- then used atidmkpacher
- I tried the latest version of the miner

But now I'm mining at 350h/s each...

compute mode, enabled?

Thanks, I forgot that. Now it's better but I'm still at 850h/s with 1250 clock 2200 memory..

Experiment with the values "multi_hash" 832, 864, 896, 944.

[Lermite]

I currently use SRB to mine XMR.

The timings of both my RX 570 are optimized with Polaris Bios Editor 1.6.7 which gives more performance than SRB Polaris.

RX 570 4G Samsung
GPU 1280 MHz 990 mV
Memory 2110 990 mV
989 h/s

RX 570 8G Micron
GPU 1280 Hz 980 mV
Memory 2220 Mhz 980 mV
977 h/s

Their config for SRB Miner is pretty simple:
  { "id": 0, "intensity": 0, "threads": 2},
  { "id": 1, "intensity": 0, "threads": 2},

[lebuawu2]

I read here opposite results, so:
* Ensure you're talking about CN-Heavy and related, CN-v8 is not concerned by the 0.33b6 release
* Is it unstable on the 8Gb RX and Vega only? Sure that's the most common cards, but if this is the status, it's an important detail to track the bug down. I did my long-run tests on 2G and 4G cards only so far. If I need a 8G card to reproduce the bug, i've some, so i need to focus on them.

Sorry you are right - I haven't specified the algo - CN-heavy

580/8GB, 1250/2000

     { "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 0, "multi_hash":992 },
     { "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 0, "multi_hash":992 },

032q - stable 1165h/s

033b6 - unstable 1000-1220h/s (varies) and the total hash for the rig - lower that stable 1165 with older version for all cards.

Same situation is on vega56.

Lowering milti_hash, only decrease the hasrate, the fluctuations remain. 4GB no problem.

Hi,

how you can set multi_hash 992 for RX 580 8GB? what is your OS version and driver version?

I only got unstable hashrate for RX 580 8GB on version 0.33b5 and 0.33b6 (even after lower multi_hash to 864 or 832). Vega 56 got stable hashrate on both version 0.33b5 and 0.33b6.
so currently all 580 8GB still using 0.33b4 and Vega 56 using 0.33b6.

edit : all related to CN-Heavy variant.

Thanks.

[lstrike]

I read here opposite results, so:
* Ensure you're talking about CN-Heavy and related, CN-v8 is not concerned by the 0.33b6 release
* Is it unstable on the 8Gb RX and Vega only? Sure that's the most common cards, but if this is the status, it's an important detail to track the bug down. I did my long-run tests on 2G and 4G cards only so far. If I need a 8G card to reproduce the bug, i've some, so i need to focus on them.

Sorry you are right - I haven't specified the algo - CN-heavy

580/8GB, 1250/2000

     { "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 0, "multi_hash":992 },
     { "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 0, "multi_hash":992 },

032q - stable 1165h/s

033b6 - unstable 1000-1220h/s (varies) and the total hash for the rig - lower that stable 1165 with older version for all cards.

Same situation is on vega56.

Lowering milti_hash, only decrease the hasrate, the fluctuations remain. 4GB no problem.

Hi,

how you can set multi_hash 992 for RX 580 8GB? what is your OS version and driver version?

I only got unstable hashrate for RX 580 8GB on version 0.33b5 and 0.33b6 (even after lower multi_hash to 864 or 832). Vega 56 got stable hashrate on both version 0.33b5 and 0.33b6.
so currently all 580 8GB still using 0.33b4 and Vega 56 using 0.33b6.

edit : all related to CN-Heavy variant.

Thanks.

Windows 10, Adrenalin 18.6.1

https://i.imgur.com/SY1QjmV.jpg

Same config on vegas, hashing 1630h/s.

[NecronomicoN]

I read here opposite results, so:
* Ensure you're talking about CN-Heavy and related, CN-v8 is not concerned by the 0.33b6 release
* Is it unstable on the 8Gb RX and Vega only? Sure that's the most common cards, but if this is the status, it's an important detail to track the bug down. I did my long-run tests on 2G and 4G cards only so far. If I need a 8G card to reproduce the bug, i've some, so i need to focus on them.

Sorry you are right - I haven't specified the algo - CN-heavy

580/8GB, 1250/2000

     { "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 0, "multi_hash":992 },
     { "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 0, "multi_hash":992 },

032q - stable 1165h/s

033b6 - unstable 1000-1220h/s (varies) and the total hash for the rig - lower that stable 1165 with older version for all cards.

Same situation is on vega56.

Lowering milti_hash, only decrease the hasrate, the fluctuations remain. 4GB no problem.

Hi,

how you can set multi_hash 992 for RX 580 8GB? what is your OS version and driver version?

I only got unstable hashrate for RX 580 8GB on version 0.33b5 and 0.33b6 (even after lower multi_hash to 864 or 832). Vega 56 got stable hashrate on both version 0.33b5 and 0.33b6.
so currently all 580 8GB still using 0.33b4 and Vega 56 using 0.33b6.

edit : all related to CN-Heavy variant.

Thanks.

580 8GB Micron "multi_hash": 992
https://fastpic.ru/view/98/2018/1110/b51b301024ef1a3f723873bfd4ecff3f.png.html

[Ultrasonik]

Unfortunately, I could not get a result more than SRB 1003 h/s (rx470  1165/2100 timing mod bios,  880mV)  
SRB
JCE mode --auto max 970 h/s
JCE manual config (checked all variants options above) give less than 970
as sample,  multi_hash ": 992  = give me  740 h/s
I wanted at least 1100  h/s  

[JCE-Miner]

online is

0.33f CPU Windows

* Cryptolight-Dark algo, --variation 17
* Parameterless config

The parameterless config, to run JCE like a service, will look, when no param is given, to a file named serviceconfig.txt
If it exists and contains parameters (in a JSON style) the miner will apply those params and start.
Example:

Code:

/* This is an example of implicit parameters 
   They are applied if and only if:
   * The miner is started with no parameter
   * And a file named exactly serviceconfig.txt is in the miner current directory
   * And it contains JSON-like parameters as comma-separated quoted strings
*/

// Usual parameters
"--auto", "--any", "--forever", "--keepalive", "--low",

// Using automatic fork selection, note that numbers don't need to be quoted
"--variation", 0,

// Example pool for Monero
"-o", "pool.monero.hashvault.pro:3333",

// Example wallet (that's the donation XMR wallet)
"-u", "43En6FLUc9EQPFjP21B73gLHBMPvUBV9LQxNA8ARsm7fLsMbybWeZgY6Dy1k77un6Z9BF8r7yxS6yTGCGdQhofYL6AvxwxZ",

// Password, no matter
"-p", "x"

[TigTex]

Three config arrays for three algo class (Light, normal, Heavy)

multiple algos on config.txt

That's exactly how it works. Example from the doc:

Code:

/* This is the configuration when Cryptolight class algo is mined */
"cryptolight_cpu_threads_conf" : 
[ 
     { "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true, "multi_hash":1 },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 1, "use_cache" : true, "multi_hash":1 },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 2, "use_cache" : true, "multi_hash":1 },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 3, "use_cache" : true, "multi_hash":1 },
],

/* This is the configuration when Cryptonight class algo is mined */
"cryptonight_cpu_threads_conf" : 
[ 
     { "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true, "multi_hash":1 },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 2, "use_cache" : true, "multi_hash":1 },
],

/* This is the configuration when CN-Heavy class algo is mined */
"cn_heavy_cpu_threads_conf" : 
[ 
     { "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true, "multi_hash":1 },
],

Well, I've had some spare time today and I'm testing this again.
Seems like the information you gave me is only valid for CPU mining. I'm using GPU mining and there is no information on the wiki
The wiki tells me to use --variation 16 for moneroocean, but it's an invalid value for the GPU miner so I just omit the value, and <algo>_cpu_threads_conf is not meant to be used on a gpu mining rig so I'm using --auto
Couldn't you just merge both versions in a gpu+cpu mining with the same features and maintain only one miner? Because the gpu miner also does cpu mining, but the wiki is missing information. Also you have so many .bat files and .txt, just keep it simple and put examples and all documentation inside the readme.txt for a clean and easy way to upgrade and access the documentation.
I've also experienced some bugs: For example, on the gpu miner, there's no information on how to use the gpu temperature monitoring. It always reports 0ºC (driver 18.6.1, windows 10 ltsb, no display connected to amd gpus). For some reason, --auto only runs on 4 of my 5 gpus, one is idle unless I configure it on config.txt (I have 5 RX580 4GB and 1 IGP, i've set -g 1,2,3,4,5 and it mines on 1,2,3,4)
Don't get me wrong, I've tested many miners and I would like to start using yours, but it needs some improvements before I can join. I'll keep on checking the changelogs

[JCE-Miner]

In short, the MoneroOcean mode is not supported yet in the GPU version. I'll add it, but not ready yet.

The CPU-only still exists because:
* It allows 32-bits mode
* It doesn't need OpenCL at startup (could be fixed in the GPU version)
* It has a linux version

-g is a restriction, so if a GPU is not enabled by the autoconfig, -g won't force it.
Is the missing GPU a Intel one? Intel GPU are detected but not enabled by default.

It's right I should clean some parts of the docs, but it's still better to keep two pages, because CPU and GPU config have each some exclusive features.

The GPU temperature/fan should work automatically, but if not, it's not blocking and it displays 0. I got reports about this bug but could never reproduce it

[TigTex]

Thank you for the reply.

The missing gpu is a RX580.
GPU0 is intel HD and if I set --auto without using the -g option, it will try to mine with gpu 0,1,2,3,4 (that is intel graphics + 4 RX580) and gives a known opencl error because you can't mine on an intel HD graphics. GPU5 is listed but it doesn't start any thread for it. Using -c config.txt and specifying the 5 gpu's works as intended.
This is on a Asus z270-p with intel G4560 cpu and windows 10 ltsb (version 1607), jce_cn_gpu_miner.033b6
I have the intel graphics enabled so that the windows dwm.exe process doesn't allocate video memory on any rx580. Without it, I have to reduce intensity on the main gpu or the system lags and BSOD's

[christiano88]

Please please please fix the warmup on RX470-480-570-580 8gb series... still having really bad speeds with b6. Only one out of 3-4 cards make it to full speed.

[UnclWish]

I think, in starting of mining heavy algo 5 minutes warm-up needs to replace 15-20. I playing with different multi_hash parameters, but still no numbers that give stable speed... But warm-up period on heavy algo is much more than 5 min.

[JCE-Miner]

that mix of Intel and autoconfig is clearly a bug, right i'll fix it. and it's pretty bad since it prevent the basic examples to start…

i'm still working on the warmup and stability problem on 8G+ cards.

[UnclWish]

i'm still working on the warmup and stability problem on 8G+ cards.

Good! Cause there is some bug... Differenet launches this the same parameters can result different speeds...

[dnkn]

Been using 33b6 since it came out and I can say there's nothing changed since b5. I can achieve stable hashrate but with much lower multihash. I'm talking about 570 4gb. I can get ~1000 out of those but only for a couple of minutes. Stable hashrate is around 800-900 (I used to get ~820 using 0.32q). So +20% boost is somewhat artificial for now. But on rx vega it does seem to be stable enough. I'm getting ~1840 per card for 24h+

[PIOUPIOU99]

jesus no v8 for cpu 33f ?

[vmozara]

alpha and beta are internal data packing modes. I cannot tell more without revealing code secrets. They must be power of two, and only three values are decent for them, respectively 32/64/128 and 4/8/16, except for very old cards where extreme values like alpha=256 or beta=32 can be good.

Other greeks are deprecated, note they no longer appear in the doc.

Thanks for the clarification! I was still using them (transfered from some of the older configs). So the any numbers to play with are alpha and beta,  and other letters we can delete from the config?

Anyway, in 3 hours it will be about 24 hours since I switched my rigs from CN8 TR miner, to JCE heavy miner. 6 out of 7 Vega rigs are working without one single issue, except one rig that restarts every couple of hours, which brings me to following suggestion:

JCE,
please can you implement a warning message into the miner, when one card/thread stops hashing, to immediately write this message into log file? Now, if the card hangs the rig will crash before periodic hashrate reports. If there is one immediate warning message about the card, we can see it later in logfile and tune that card.
Thanks!

[sergneo]

0.33f CPU Windows

* Parameterless config

thanks, that's it

[JCE-Miner]

My first tests on hybrid miner give interresting results on old cards. But not on new ones since the cost of sync tend to be higher than their native speed. But the main goal remains possible: put a big boost on older cards.

JCE,
please can you implement a warning message into the miner, when one card/thread stops hashing, to immediately write this message into log file? Now, if the card hangs the rig will crash before periodic hashrate reports. If there is one immediate warning message about the card, we can see it later in logfile and tune that card.
Thanks!

Isn't the Watchdog enough? Otherwise no problem to put just a message. GPU Stuck -> Watchdog enabled: Stop, disabled: Message.

jesus no v8 for cpu 33f ?

a VERY BIG thank for this report, it probably confused lots of people. I'll re-upload the package with the comments fixed.

I could reproduce the drop problem on my RX on Heavy, at full speed my miner is clearly above others, but the drop is like -15% and this cancels any gain, as some stated. I've some ideas how to fix, need more time and tests.

Online is Linux CPU 0.33f
exact same as Windows version

[fluxy12]

I could reproduce the drop problem on my RX on Heavy, at full speed my miner is clearly above others, but the drop is like -15% and this cancels any gain, as some stated. I've some ideas how to fix, need more time and tests.

Happy to heard that we are all behind you (at least the rx580 8gb owners)