Team Black Miner (ETHW ETC Vertcoin Ravencoin Zilliqa +dual +tripple mining ) - page 70.

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: snoby67 on October 29, 2021, 05:14:40 AM

I try to setup TBM on HiveOs with different GPU.
I try all --cl-devices [ i ] with i=0 to 15 and all the time the miner fail and restart.

You can try the short option instead -Y [ 0 ]

snoby67

newbie

Activity: 13

Merit: 0

Hello All,

I try to setup TBM on HiveOs with different GPU.
GPU 0 : 3070ti
GPU 1 : 6700XT
GPU2 : 3060ti
etc

I try different setting to take only 3060ti and 6700XT but seems doesn't work.
AMD driver : 20.40 (5.11.0701)
Nvidia driver : 495.44

1st test : -b --> take only amd card is working
2nd test : --auto-detect --> all card is working
3rd test : --cuda-devices [2] --> OK only 3060ti is working
4rd test : --cuda-devices [2] --cl-devices [1] --> 1 is the id of the 6700XT card // FAIL : source/sp_workers/cl_gpu.cpp:cl_setup_devices():576: A OpenCL runtime error occured: CL_INVALID_VALUE

I try all --cl-devices [ i ] with i=0 to 15 and all the time the miner fail and restart.

I just would like to take all the AMD card and a list of NVIDIA card (no-LHR card), do you have another way to do that or something I do wrong ?

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: kiss34000 on October 29, 2021, 02:04:30 AM

if other (gnminer and many other miner) can do core and mem from command line i'm sure you can do it too

Yes. It can be done, but most of the miners are using hivos, raveos, mpos and in these os'es you can set the clocks. Msi afterburner is good, you can set clocks, templimit, fans and let the program remember the settings on the next boot.

There are other more important issues to fix. Like faster AMD and LHR unlock..

kiss34000

newbie

Activity: 6

Merit: 0

thanks for the help so now it's work for the core frequency

don't ask me why but only add the program (.exe) admin right never work for me when i use a .bat file i always need to do a shortcut of the .bat and add the admin right on the shortcut and on the .exe and modify the real path of the .exe on the bat

not hard but i dont understand why and i read many people with the same problem

the memory is still not working properly with tbm and i dont want to use other program like afterburner or precision (from evga) for a manual oc i realy like the all in one way for a mining software

if other (gnminer and many other miner) can do core and mem from command line i'm sure you can do it too

tbm + afterburner fix config is the most efficient miner atm

but i will stay under gnminer until we can do a proper tuning of the card from the .bat file

thanks for your work

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: kiss34000 on October 29, 2021, 12:55:40 AM

hi very happy to see you add
-j --lock-cclock [[,],[,]] (Set an array of min and max core clock values for cuda devices)
-m --lock-mclock [[,],[,]] (Set an array of min and max memory clock values for cuda devices)

so i try under windows so
i create a shortcut with admin right for launch the .bat in admin mode
I modify the absolute path for a correct lauch of the the bat command

You have to give the exe file admin rights. Rightclick on the exe and change in the compability tab

Quote from: kiss34000 on October 29, 2021, 12:55:40 AM

i try
E:\_DL\TBM\TBMiner.exe --algo ethash --hostname eu-eth.hiveon.net --port 4444 --wallet %ETH_WALLET% --worker-name STRIX90 -Z --api --api-ip 127.0.0.1 --api-port 4068 --lock-cclock [1250] --lock-mclock [1700]

The correct syntax is: (min, max)

--lock-cclock [1250, 1250]

for 6 cards in the rig:

--lock-cclock [[1250, 1250], [1250, 1250], [1250, 1250], [1250, 1250], [1250, 1250], [1250, 1250]]

--lock-mclock only works for low memclocks supported by nvidia-smi. And these are too low for mining.
You should use msi-afterburner or similar program to set the memclock.

Quote from: KillrBee on October 28, 2021, 02:48:03 PM

1200 for me reduces my hashrate from 136 to 131. I found 1560 to be the best for my card / intensity

Increasing the coreclock to add a few MHASH'es is a nice find in The Team Black Miner.. No other miner out there can push it to 136MHASH on the rtx 3090.. You watt is high, your clocks are crazy. But 136MHASH verified poolside is very good.

[moderator's note: consecutive posts merged]

kiss34000

newbie

Activity: 6

Merit: 0

hi very happy to see you add
-j --lock-cclock [[,],[,]] (Set an array of min and max core clock values for cuda devices)
-m --lock-mclock [[,],[,]] (Set an array of min and max memory clock values for cuda devices)

so i try under windows so
i create a shortcut with admin right for launch the .bat in admin mode
I modify the absolute path for a correct lauch of the the bat command

i try
   E:\_DL\TBM\TBMiner.exe --algo ethash --hostname eu-eth.hiveon.net --port 4444 --wallet %ETH_WALLET% --worker-name STRIX90 -Z --api --api-ip 127.0.0.1 --api-port 4068 --lock-cclock [1250] --lock-mclock [1700]


or

   E:\_DL\TBM\TBMiner.exe --algo ethash --hostname eu-eth.hiveon.net --port 4444 --wallet %ETH_WALLET% --worker-name STRIX90 -Z --api --api-ip 127.0.0.1 --api-port 4068 --lock-cclock 1250 --lock-mclock 1700


for fix the core clock and the memory clock for my 3090strix and it's just dont work can you say me what i do wrong ?

https://i.imgur.com/mcQNOSv.png
https://i.imgur.com/h3L1KKu.png
thanks in advance

i use it the same way in gnminer and it's work on it but not on tbm 1.21 5

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: dle378 on October 28, 2021, 09:25:41 PM

480m [2021-10-28 21:20:37.407] ID BOARD HASHRATE/W HASHRATE AVERAGE SHARES

I see that you are not running the latest version 1.21. Why?

Quote from: dedizonesv2 on October 28, 2021, 06:51:22 PM

I put 2 rigs Gminer VS TBM

Rigs of what? TBM is not faster on all cards, but some cards are faster with the right mining pool and settings.

[moderator's note: consecutive posts merged]

dle378

copper member

Activity: 77

Merit: 0

Quote from: sp_ on October 28, 2021, 04:42:50 PM

Quote from: dle378 on October 28, 2021, 04:29:10 PM

It seems as the miner submits more accepted shares on xintensity 160 vs 256.

Depends on the miningpool and what it expect. For some pools perhaps yes, other pools no.

Got it.

Here is current status after 8 hours on --xintensity 512 & --dagintensity 0:

480m [2021-10-28 21:20:37.407] ID   BOARD   HASHRATE/W   HASHRATE     AVERAGE      SHARES
480m [2021-10-28 21:20:37.407] GPU0   3060Ti   422.45 kH/W   64.64 MH/s   64.64 MH/s   200/0/0 (100.00)
480m [2021-10-28 21:20:37.407] GPU1   3060Ti   347.45 kH/W   64.63 MH/s   64.63 MH/s   201/0/0 (100.00)
480m [2021-10-28 21:20:37.407] GPU2   3060Ti   416.87 kH/W   64.61 MH/s   64.62 MH/s   201/1/0 (99.50)
480m [2021-10-28 21:20:37.407] GPU3   3060Ti   347.45 kH/W   64.63 MH/s   64.62 MH/s   196/0/0 (100.00)
480m [2021-10-28 21:20:37.407] GPU4   3060Ti   416.92 kH/W   64.62 MH/s   64.62 MH/s   236/0/0 (100.00)
480m [2021-10-28 21:20:37.407] GPU5   3060Ti   417.03 kH/W   64.64 MH/s   64.63 MH/s   194/0/0 (100.00)
480m [2021-10-28 21:20:37.407]         394.69 kH/W   387.76 MH/s   387.76 MH/s   1228/1/0 (99.92)

dedizonesv2

jr. member

Activity: 152

Merit: 3

Quote from: sp_ on October 28, 2021, 07:51:40 AM

Quote from: KillrBee on October 28, 2021, 07:47:40 AM

I would say that's the understatement of the year! Looking back through my 2Miners API logs, it was running in excess of 140MH/s at points over night!
Now if we can get my AMD cards running like the 3090 I'll be back to pre-EIP1559 Ether / day!

Go post it on reddit and let them burn some comment karma.

.

The Team Black is a good profitable mining software that actually works pretty good. We don't have the funding and resources of the other teams, but bugs are being fixed and the miner is now stronger than ever. The Team Black coders are gpu/miner opensource developers since 2013.

https://www.reddit.com/r/EtherMining/comments/qhl6wy/team_black_miner_v121_has_been_released_faster/

There is no point in having 140Mhs if your shares are not ready, with an increase in consumption.

If we lower the consumption, we easily lose 50/70 shares per hour.

I put 2 rigs Gminer VS TBM

each time we were +25 more shares on Gminer ... and 80w less so where the real savings are apart from putting in the electric bill and losing ETH in the final

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: dle378 on October 28, 2021, 04:29:10 PM

It seems as the miner submits more accepted shares on xintensity 160 vs 256.

Depends on the miningpool and what it expect. For some pools perhaps yes, other pools no.

Quote from: ayman_abm on October 28, 2021, 04:26:50 PM

Daggerhashimoto needs more Clocks on Memory as I know. what do you suggest?

Depends on how high the memory can be clocked. You need to find the sweetspot's between memory and coreclock. We are close to maximum on the 3080 and the 3090 with tbminer, but the 3070 might have a few mhashes left.

[moderator's note: consecutive posts merged]

dle378

copper member

Activity: 77

Merit: 0

Quote from: sp_ on October 28, 2021, 12:13:00 AM

You can increase the speed with higher intensity. Why don't you try it out. What numbers are you getting?

--xintensity 224 (default)
--xintensity 256
--xintensity 288
--xintensity 320

I am getting 64-65 MH per card on xintensity 160/256/512.

It seems as the miner submits more accepted shares on xintensity 160 vs 256.

Thoughts?

ayman_abm

newbie

Activity: 24

Merit: 0

Quote from: sp_ on October 28, 2021, 03:09:35 PM

Quote from: ayman_abm on October 28, 2021, 03:01:56 PM

What happens if you use --lock-cclock [[1200,1200] ? for me its same hash rate but way less power used as 1500/1600, and temps less for sure since its less power consuming

Higher clocks need more power. When you tell the device that you don't want higher than 1200 clock it will undervolt automaticly. For the laptop gpu's this is giving a nice boost in the Hashrate. Less Power, no trottleing, run cooler, more hash..,

For Laptop it works well with 1200, what about Desktop's GPU! do you think 3070 and 3060Ti needs more than 1200 Core Clock ? Daggerhashimoto needs more Clocks on Memory as I know. what do you suggest?

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: ayman_abm on October 28, 2021, 03:01:56 PM

What happens if you use --lock-cclock [[1200,1200] ? for me its same hash rate but way less power used as 1500/1600, and temps less for sure since its less power consuming

Higher clocks need more power. When you tell the device that you don't want higher than 1200 clock it will undervolt automaticly. For the laptop gpu's this is giving a nice boost in the Hashrate. Less Power, no trottleing, run cooler, more hash..,

ayman_abm

newbie

Activity: 24

Merit: 0

Quote from: KillrBee on October 28, 2021, 09:42:31 AM

Quote from: ayman_abm on October 28, 2021, 09:25:56 AM

@KillrBee

I have a rig of 3090s (6 cards) running on HiveOS. I couldn't reach 11326 mem clock (+3000 in Hive) max could reach was +2900 and after several minutes miner crashes. Cards cooled up using AC direct cold air to them.

I think you are using Windows ?!

one more thing, I used +1200 on Core, didn't try +1600 since Ethhash a memory hungry algorithm

Yeah, Linux and Windows are different beasts when it comes to the GPU clocks. +1200 would probably be just fine -- essentially I'm running at .75 - .80 Volts. I was very lucky with how well my card overclocks, I can get that memory up to 11400 for a while, but eventually the card throws an error (even though temps are well controlled).
Try pushing the intensity up and keeping the memory overclock to +2600 (I'm running XInt = 3108 right now and I can keep the memory clock down to a level that the miner doesn't drop or gpu hang for 24hr+ {which is 11326ish for my card}). It's taken A LOT of experimenting and testing to find a stable and fast hashrate/share rate, and it continues to be refined every day and every release. I am probably spending way too much time tweaking vs. doing my day job, but I'm a performance freak and always have to find the useable limits of everything. :-)

What happens if you use --lock-cclock [[1200,1200] ? for me its same hash rate but way less power used as 1500/1600, and temps less for sure since its less power consuming

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

I think 1500+ is needed to push the 3080 to 110MHASH and the 390 to 135MHASH.

As for the 3070, you can try to increase the power use --dagintensity 0 and increase the clocks. Windows can oc more than linux. Perhaps you can reach 67MHASH. But the last MHASH'es would be expensive. Use to much power to be justified...

KillrBee

jr. member

Activity: 139

Merit: 3

Quote from: ayman_abm on October 28, 2021, 02:41:03 PM

--lock-cclock [[1500,1500],[1500,1500],[1500,1500],[1500,1500],[1500,1500],[1500,1500]]

Is This Good to use with RTX 3060Ti 3070 and 3090 ?? or better use 1200 ?

I can confirm that 1200 uses less power and same hash-rate as 1500! but in previous posts I saw many likes 1500 !!!

1200 for me reduces my hashrate from 136 to 131. I found 1560 to be the best for my card / intensity

ayman_abm

newbie

Activity: 24

Merit: 0

--lock-cclock [[1500,1500],[1500,1500],[1500,1500],[1500,1500],[1500,1500],[1500,1500]]

Is This Good to use with RTX 3060Ti 3070 and 3090 ?? or better use 1200 ?

I can confirm that 1200 uses less power and same hash-rate as 1500! but in previous posts I saw many likes 1500 !!!

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: PeaMine on October 28, 2021, 10:20:11 AM

Is it possible to get 1.21 with Cuda 11.5 for Linux?

Not yet.

Quote from: PeaMine on October 28, 2021, 10:20:11 AM

Also if I have a high CPU load should I use the no cpu usage switch?

Doesn't work for AMD

Quote from: PeaMine on October 28, 2021, 10:20:11 AM

Does dual mining eth zil on the same pool need any tags or slow things down like ssl?

AMD cards doesn't have a dag cache implemented yet, so the dag creation on switch will slow down since you need to generate new dag's every 90 minutes. epoch0 and epoch 450.
NVIDIA cards with more than 6GB memory will cache the dag and the switch is smooth.

PeaMine

hero member

Activity: 979

Merit: 510

Is it possible to get 1.21 with Cuda 11.5 for Linux?
I'm using Ubuntu 18.04 which has always done solid for me.
Should I upgrade to 20.04?

Also if I have a high CPU load should I use the no cpu usage switch?
I've tried increasing the niceness priority but for some reason can't get below 0 without renice which breaks my script
Does dual mining eth zil on the same pool need any tags or slow things down like ssl?

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

I gained 4 mhash on the 3070 Laptop gpu by using the tbminer parameter

--lock-cclock [1200,1200]

From ~49MHASH to ~53MHASH at 80 watt. (v1.21)

Lock-clock is actually lowering the voltage needed / power. So I guess the lost power went to the memory controller instead.

in v1.21 you should also use --dagintensity 0 and boost the memory clock until the dag is failing to validate. As long as the validation doesn't fail, the black search code in TBminer can run stable for weeks on high clocks.

Topic: Team Black Miner (ETHW ETC Vertcoin Ravencoin Zilliqa +dual +tripple mining ) - page 70. (Read 35100 times)