⛏Team Black Miner (ETHB3 ETHW ETC VTC KAWPOW FIROPOW EVRPROGPOW MEOWPOW +dual)

[ekgd39]

I spent about 4 days on tests,

Which version did you test. Which mining pool did you test? Let it run for 5000 shares. TBM should be faster if you configure it correctly.. What speed are you getting in the miner window. Screenshots please.

I tried on 2miners , on a 12x3080 rig, Hive OS cuda 11.5 driver 495.46 cuda 8 xint 3108 , I have 1321mh in the miner window. Average per pool per day 1.24 ghs. It works well for the first hour, then it starts to fall behind. I can put it for you again for a test for a day, so that you can be convinced of this yourself. Maby in hive i must use 11.4 cuda , i dont know, but it not work ....

Now I will turn it on on cuda 11.4 And I will throw off the test result in 2-3 hours here. Cuda 11.5 doesn't work on Hive os, probably because autotuning always chooses core 1 for my cards on any settings.

[sp_]

I tried on 2miners , on a 12x3080 rig, Hive OS cuda 11.5 driver 495.46 cuda 8 xint 3108 , I have 1321mh in the miner window. Average per pool per day 1.24 ghs. It works well for the first hour, then it starts to fall behind. I can put it for you again for a test for a day, so that you can be convinced of this yourself. Maby in hive i must use 11.4 cuda , i dont know, but it not work ....

Try v1.42 cuda v11.4 build. 11.5 might have issues. As reported in the previous post. 3108 might not be the optimal config for your rig. This setting works good on 1 rtx 3090.

[ekgd39]

I tried on 2miners , on a 12x3080 rig, Hive OS cuda 11.5 driver 495.46 cuda 8 xint 3108 , I have 1321mh in the miner window. Average per pool per day 1.24 ghs. It works well for the first hour, then it starts to fall behind. I can put it for you again for a test for a day, so that you can be convinced of this yourself. Maby in hive i must use 11.4 cuda , i dont know, but it not work ....

Try v1.42 cuda v11.4 build. 11.5 might have issues. As reported in the previous post. 3108 might not be the optimal config for your rig. This setting works good on 1 rtx 3090.

Cuda 11.4 in the miner immediately increase in hashrate

Cuda 11.5 RTX 3080 102.5mh

Cuda 11.4 RTX 3080 110.5mh

Apparently the libraries and compatibility of the miner in HIVE OS with cuda 11.5 are at a low level and the miner does not work correctly.

[sp_]

Cuda 11.4 RTX 3080 110.5mh

Now you need to find the correct intensity for your rig so that the mining speed is reflected poolside. When you are done you will understand why alot of miners have switched.
AMD/NVIDIA. Try to combine high --xintenity with --tweak --kernel . The defaults will probably not work optimal.

[ekgd39]

Cuda 11.4 RTX 3080 110.5mh

Now you need to find the correct intensity for your rig so that the mining speed is reflected poolside. When you are done you will understand why alot of miners have switched.
AMD/NVIDIA. Try to combine high --xintenity with --tweak --kernel . The defaults will probably not work optimal.

https://i.ibb.co/DYSTntT/3080.png

After 2 minutes, the hashrate dropped again, but still more than cuda 11.5

[sp_]

After 2 minutes, the hashrate dropped again, but still more than cuda 11.5

with 13 cards you should get at least 1.404GHASH in the miner window on optimal settings. 108MHASH each. But perhaps you are running in powersaver mode.

[KillrBee]

I spent about 4 days on tests,

Which version did you test. Which mining pool did you test? Let it run for 5000 shares. TBM should be faster if you configure it correctly.. What speed are you getting in the miner window. Screenshots please.

I tried on 2miners , on a 12x3080 rig, Hive OS cuda 11.5 driver 495.46 cuda 8 xint 3108 , I have 1321mh in the miner window. Average per pool per day 1.24 ghs. It works well for the first hour, then it starts to fall behind. I can put it for you again for a test for a day, so that you can be convinced of this yourself. Maby in hive i must use 11.4 cuda , i dont know, but it not work ....

Now I will turn it on on cuda 11.4 And I will throw off the test result in 2-3 hours here. Cuda 11.5 doesn't work on Hive os, probably because autotuning always chooses core 1 for my cards on any settings.

I will say that I consistently get over 135 MH/s poolside on my 3090 FTW running for 24hr + at 2miners, with the recent update, I've seen a good boost on my AMD cards, especially in Windows (65-68 MH/s) over long stretches as well.  Setup is key and I have no experience with the 3080s, but with the 3090, intensity at 3801 and GPU running no less than 1630 MHz (which keeps the voltage at a set point that seems to work well for throughput), and memory speed to whatever you can get as a stable setting, and it's important to force p0 state and ensure "prefer maximum performance" is set in the nvidia-settings for each nvidia card.

Oh, and definitely set the kernel to 8 on the 3080s when using 3801 as the intensity on the miner.

[ekgd39]

After 2 minutes, the hashrate dropped again, but still more than cuda 11.5

https://i.ibb.co/3kMjPCD/Screenshot-2022-01-20-at-19-56-35-W5-ethash-1-324-GH-s-Ed-Konst.png

with 13 cards you should get at least 1404 in the miner window on optimal settings. 108MHASH each. But perhaps you are running in powersaver mode.

https://ibb.co/bQfY9dg

[sp_]

You are locking the coreclock at 1250, you probolby need to increase it to 1500 or something. The watt on the card is unstable which indicate trottleing. The memory temp could be too high or power too low.

[ekgd39]

https://i.ibb.co/17SBW2f/Screenshot-2022-01-20-at-20-02-13-W5-ethash-1-324-GH-s-Ed-Konst.png

You are locking the coreclock at 1250, you probolby need to increase it to 1500 or something. The watt on the card is unstable which indicate trottleing. The memory temp could be too high or power too low.

https://ibb.co/jws3Csr

dont work my friend.........

Maybe  need to enable P0 mode?

[KillrBee]

You are locking the coreclock at 1250, you probolby need to increase it to 1500 or something. The watt on the card is unstable which indicate trottleing. The memory temp could be too high or power too low.

dont work my friend.........

Maybe  need to enable P0 mode?

P0 is a must, GPU at 1650 and your memory OCs seem really low; I'm running +3375 in HiveOS (although normally with temperature being an issue, I would run +3000 (+1500 in windows)).

Try intensity of 2380 since the 3801 came from a 3090 card, which has GDDR6x and I think the 3080 non-lhr just have GDDR6, and the (x) ram basically doubles the message rate per cycle.

[dle378]

Added 2x 3080 LHRs to my setup and was hoping to get a stable hash rate on the cards.

Any tips or options to add as a starting point?

Current Options/Settings

--lhr-unlock 1
--lock-cclock [[1500,1500]]
--xitensity 2048
Memory Clock: 1000 (2000 if HiveOS)
P0 State enabled.

I am currently on TBM 1.42 (CUDA 11.4) with Nvidia GPU driver CUDA 11.5.

[sp_]

Added 2x 3080 LHRs to my setup and was hoping to get a stable hash rate on the cards.
Any tips or options to add as a starting point?

On windows I use 75% powerlimit and low coreclock as possible on the 3080ti. No need to lock the clock.

If the card fail to unlock try with -1, -2, -3, -4 ... etc until it's stable.

--lhr-unlock [-1,-1,0,0,0,0]

[KillrBee]

Added 2x 3080 LHRs to my setup and was hoping to get a stable hash rate on the cards.
Any tips or options to add as a starting point?

On windows I use 75% powerlimit and low coreclock as possible on the 3080ti. No need to lock the clock.

If the card fail to unlock try with -1, -2, -3, -4 ... etc until it's stable.

--lhr-unlock [-1,-1,0,0,0,0]

I'm not familiar with how the LHR unlock works, but I am curious about how the lhr-unlock and the tweak settings work together..  Also, I assume that the intensity at -2 on the 3080's means something different than it does on the non-LHR cards, and the kernel selection for the 3080s being low values is also an LHR thing??

[ekgd39]

Added 2x 3080 LHRs to my setup and was hoping to get a stable hash rate on the cards.
Any tips or options to add as a starting point?

On windows I use 75% powerlimit and low coreclock as possible on the 3080ti. No need to lock the clock.

If the card fail to unlock try with -1, -2, -3, -4 ... etc until it's stable.

--lhr-unlock [-1,-1,0,0,0,0]

https://ibb.co/hXvNmmk

In Hive OS, it always sets kernel 1 automatically, on any settings, this kernel shows the best result in terms of hashrate, I tried kernels from 5 to 8 - they work poorly, very few shares are found. My result in 1 hour 15 minutes, until I see no increase even to the declared hashrate is not close ...... I continue the test further ....

[sp_]

I'm not familiar with how the LHR unlock works, but I am curious about how the lhr-unlock and the tweak settings work together..  Also, I assume that the intensity at -2 on the 3080's means something different than it does on the non-LHR cards, and the kernel selection for the 3080s being low values is also an LHR thing??

In the LHR more tweak will increase the load sent to the gpu since xintensity is not supported. It has a different meaning than in standard mode. In standard mode tweak will split jobs in order try to remove stale shares. This is not needed on the 2miners pool.

[sp_]

In Hive OS, it always sets kernel 1 automatically

Make sure that -J 1 is not passed. The program should autotune, or you can set -J -1 to force autotune.

I tried kernels from 5 to 8 - they work poorly, very few shares are found.

On windows and cuda 11.5 kernel 5-8 gives the best results on RTX cards. The next version will have 16 kernels to try to improve the performance on low power.

You need to find out why your cards in the rig is using 168- 241 watt. Cards seems to trottle. (downclock and perform poorly) They need to run at the same wattage all the time. Did you manage to run them in p0 state?

v1.42 windows 11.5 build

Kernel 1 RTX 2060 31  MHASH
Kernel 8 RTX 2060 33.5MHASH (but needs more power)

v1.43 will be released in a few hours.

[ekgd39]

In Hive OS, it always sets kernel 1 automatically

Make sure that -J 1 is not passed. The program should autotune, or you can set -J -1 to force autotune.

I tried kernels from 5 to 8 - they work poorly, very few shares are found.

On windows and cuda 11.5 kernel 5-8 gives the best results on RTX cards. The next version will have 16 kernels to try to improve the performance on low power.

You need to find out why your cards in the rig is using 168- 241 watt. Cards seems to trottle. (downclock and perform poorly) They need to run at the same wattage all the time. Did you manage to run them in p0 state?

v1.42 windows 11.5 build

Kernel 1 RTX 2060 31  MHASH
Kernel 8 RTX 2060 33.5MHASH (but needs more power)

v1.43 will be released in a few hours.

I use P0 , tbm 1.42 cuda 11.4 , on cuda 11.5 I have worse results .  video cards do not throttle, core frequencies do not drop off, this is out of the question.  raising the power does not give an increase in hashrate, ethereum mining does not need a high core value, it only uses memory.
https://ibb.co/YyYTDPV

[sp_]

video cards do not throttle, core frequencies do not drop off, this is out of the question.  raising the power does not give an increase in hashrate, ethereum mining does not need a high core value, it only uses memory.

Some of your cards are not clocked correct, and some have fluctating power.

[Dojo76]

In Hive OS, it always sets kernel 1 automatically

Make sure that -J 1 is not passed. The program should autotune, or you can set -J -1 to force autotune.

I tried kernels from 5 to 8 - they work poorly, very few shares are found.

On windows and cuda 11.5 kernel 5-8 gives the best results on RTX cards. The next version will have 16 kernels to try to improve the performance on low power.

You need to find out why your cards in the rig is using 168- 241 watt. Cards seems to trottle. (downclock and perform poorly) They need to run at the same wattage all the time. Did you manage to run them in p0 state?

v1.42 windows 11.5 build

Kernel 1 RTX 2060 31  MHASH
Kernel 8 RTX 2060 33.5MHASH (but needs more power)

v1.43 will be released in a few hours.

Can you suggest a kernel for full AMD rigs? It seems that autotune always select kernel 0.

About Hiveon pool rejected shares are always a problem, 1/2% average (tried xintensity 71 and 24), while they are close to zero in crazypool or 2miners  for example