why Jetson hash rate has only about 100~200kh/s? so far from therory number about 20Mh/s?

bobby20180331 commented 3 years ago

Operating System: Ubuntu 18.04.04 Linux 4.9.140-tegra Hardware Nvidia Jetson AGX Ethminer options used: -U -P stratum1+tcp://0xe5ad5e80f3ef1ed8b6978255461cd8713dcddd8f.xavier@eth.firepool.com:8008
ADD /libethash-cuda/CMakeLists.txt :
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_72,code=sm_72")
CUDA doesn't go wrong, it's normal GPU utilization reached 97.7% But the force is zero??

Like THIS:
 m 22:12:07 ethminer 0:00 A0 0.00 h - cu0 0.00
 i 22:12:08 ethminer Job: 2b59550f… eth.f2pool.com [47.108.142.254:6688]
 i 22:12:10 ethminer Job: e37014c7… eth.f2pool.com [47.108.142.254:6688]
 m 22:12:12 ethminer 0:00 A0 0.00 h - cu0 0.00
 i 22:12:14 ethminer Job: 8036a15f… eth.f2pool.com [47.108.142.254:6688]
 i 22:12:14 ethminer Job: 6b5ce8e5… eth.f2pool.com [47.108.142.254:6688]
If I use the command like this : ethminer -U -M 0 I can see the calculation speed LIKE THIS
 m 22:17:33 ethminer 0:01 A0 217.91 Kh - cu0 217.91
 m 22:17:38 ethminer 0:01 A0 217.91 Kh - cu0 217.91
 m 22:17:43 ethminer 0:01 A0 217.90 Kh - cu0 217.90
 m 22:17:48 ethminer 0:01 A0 217.90 Kh - cu0 217.90
 m 22:17:53 ethminer 0:01 A0 217.88 Kh - cu0 217.88
Too LOW (T-T)

I let it sit for a while and you can see the force

Originally posted by @ztfot in https://github.com/ethereum-mining/ethminer/issues/2047#issuecomment-706712829

bobby20180331 commented 3 years ago

m 19:37:57 ethminer 0:12 A0 0.00 h - cu0 0.00 i 19:37:59 ethminer Job: 4e084d2d… eth.f2pool.com [47.108.142.254:6688] i 19:38:01 ethminer Job: 03468b71… eth.f2pool.com [47.108.142.254:6688] i 19:38:02 ethminer Job: 53c85a31… eth.f2pool.com [47.108.142.254:6688] m 19:38:02 ethminer 0:12 A0 172.44 Kh - cu0 172.44 i 19:38:02 ethminer Job: 44366888… eth.f2pool.com [47.108.142.254:6688] m 19:38:07 ethminer 0:12 A0 172.44 Kh - cu0 172.44 i 19:38:08 ethminer Job: 250bd2c3… eth.f2pool.com [47.108.142.254:6688] i 19:38:11 ethminer Job: d595ec5a… eth.f2pool.com [47.108.142.254:6688] m 19:38:12 ethminer 0:12 A0 170.33 Kh - cu0 170.33 i 19:38:14 ethminer Job: 54a6ba92… eth.f2pool.com [47.108.142.254:6688] i 19:38:15 ethminer Job: 2f4187d9… eth.f2pool.com [47.108.142.254:6688] i 19:38:15 ethminer Job: 34e3c93a… eth.f2pool.com [47.108.142.254:6688] m 19:38:17 ethminer 0:13 A0 170.33 Kh - cu0 170.33 i 19:38:18 ethminer Job: f40d059f… eth.f2pool.com [47.108.142.254:6688] i 19:38:21 ethminer Job: 530f7b3f… eth.f2pool.com [47.108.142.254:6688] m 19:38:22 ethminer 0:13 A0 0.00 h - cu0 0.00

bobby20180331 commented 3 years ago

./ethminer -P stratum1+tcp://****.ethminer_xavier@eth.f2pool.com:6688

ethminer 0.19.0-17+commit.ce52c740.dirty Build: linux/release/gnu

No OpenCL platforms found i 19:25:17 ethminer Configured pool eth.f2pool.com:6688 i 19:25:17 ethminer Selected pool eth.f2pool.com:6688 i 19:25:19 ethminer Stratum mode : Eth-Proxy compatible i 19:25:19 ethminer Established connection to eth.f2pool.com [47.108.142.254:6688] i 19:25:19 ethminer Spinning up miners... cu 19:25:19 cuda-0 Using Pci Id : 00:00.0 Xavier (Compute 7.2) Memory : 12.85 GB i 19:25:19 ethminer Epoch : 397 Difficulty : 8.59 Gh i 19:25:19 ethminer Job: 2d92818a… eth.f2pool.com [47.108.142.254:6688] i 19:25:22 ethminer Job: 9d3ccfb5… eth.f2pool.com [47.108.142.254:6688] cu 19:25:22 cuda-0 Generating DAG + Light(on GPU) : 4.17 GB m 19:25:22 ethminer 0:00 A0 0.00 h - cu0 0.00

bobby20180331 commented 3 years ago

DDU? anyone had a try on jetson Xavier ? Chewey28 commented on Sep 3, 2020 • Hello, I'm not part of the team, but I'll try to help... Check on the docs if there's Reference to what Nvidia divers are optimal to use... Then Uninstaller completely the current Nvidia drivers using DDU! Use it on safe mode! Then install the recommended drivers... Try again then post results here!

PS - next time tell us more about your setup, windows or Linux, I had to check the print to find out...

Hi Josevora, Thank you for help! I uninstall drivers using DDU, and install the latest version of Nvidia driver 452.06 + ethminer-0.19.0-alpha.0-cuda10.0-windows-amd64. And now it looks like that: image I guess it's OK right now?)

Maybe you have any suggestions how can I improve my Mh? Like virtual memory, or smth like that. I'm still on default 4864. Maybe any ideas how to make it more cost effective in terms of energy consumption etc?

bobby20180331 commented 3 years ago

I wonder it caused by "No OpenCL platforms found". And Nvida forum offcial said openCL is not support on jetson series.

https://forums.developer.nvidia.com/t/can-the-xavier-run-opencl-applications/70262 https://forums.developer.nvidia.com/t/does-jetson-tx1-or-tx2-support-opencl/50229

And may anyone have a try "OpenCL and CUDA translation? http://snucl.snu.ac.kr/snucl-tr.html

codegastudio commented 3 years ago

Hi @bobby20180331 I try to run ethminer on Jeston Xavier NX. I have update cmake to version 3.20 and cuda to version 11. Compilation is fine but when i want run ethminer i get the following error on start

ethminer 0.19.0-4+commit.dae359df
Build: linux/release/gnu

CUDA Error : unknown error
Error: No usable mining devices found

I have try the master branch, and the cuda11 branch. I have try different build option, like cmake .. -DETHASHCUDA=ON -DETHASHCL=OFF or just cmake ... Nothing change! I get always the same error.

You seem to be able to run ethminer program on your AGX. Maybe you have an idea of what i miss please ?

Regards

codegastudio commented 3 years ago

I have finally succeed to run ethminer compiled binary !

Job is to do it

ADD /libethash-cuda/CMakeLists.txt :
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_72,code=sm_72")

Now i have the following message

 i 13:28:09 ethminer Spinning up miners...
cu 13:28:09 cuda-0   Using Pci Id : 00:00.0 Xavier (Compute 7.2) Memory : 2.71 GB
 i 13:28:09 ethminer Authorized worker 0x42855FF3F65015C295e36D736E3aaC78B5251DB0
 i 13:28:09 ethminer Epoch : 399 Difficulty : 8.73 Gh
 i 13:28:09 ethminer Job: 56d0f1e2… eth.2miners.com [51.195.105.101:2020]
 i 13:28:13 ethminer Job: 767a5f60… eth.2miners.com [51.195.105.101:2020]
cu 13:28:13 cuda-0   Epoch 399 requires 4.12 GB memory.
cu 13:28:13 cuda-0   This device hasn't enough memory available. Mining suspended ...
 m 13:28:13 ethminer 0:00 A0 0.00 h - cu0 0.00

Jetson Xavier NX come with 8Go of memory.

tegrastats return the following informations :

RAM 1461/7766MB (lfb 338x4MB) SWAP 0/3883MB (cached 0MB) CPU [100%@1420,1%@1420,0%@1420,0%@1420,0%@1420,0%@1420] EMC_FREQ 0% GR3D_FREQ 0% AO@41C GPU@41.5C PMIC@100C AUX@41C CPU@43.5C thermal@41.75C

So some one have any idea of what are the limited 2.71 GB detected which block mining process ?

codegastudio commented 3 years ago

At least !!! I have succeed to mine. Hash rate is in the same range than @bobby20180331. Do you have find a way to increase this ?

I don't understand why OpenCL should be the way cause it is the competitor of CUDA if i have well understood.

bobby20180331 commented 3 years ago

@codegastudio Nice job. I have tried to close opencl option .But No progress yet.

bobby20180331 commented 3 years ago

https://blog.csdn.net/weixin_43480227/article/details/114323718

codegastudio commented 3 years ago

Some people could give me some explanation of the graph below plz ? This is a capture of ethminer work on 2miners.com pool. I get spike at 14MH/s and after 20min, 2miners.com alert me than the worker is offline while it is running like a charm and tegrastats tool show me GPU usage at 99%.

And this all the time during 12H as you can see on the subgraph.

kivvi3412 commented 3 years ago

@codegastudio @bobby20180331

This is my previous successful address. At that time, my Jetson Xavier 16G had only 180kH of computing power. How did you get 14mH /s? (· _ ·;

@ztfot in #2047 (comment)

bobby20180331 commented 3 years ago

well. I guess

the program maybe not so stable so maybe offline accur for a while during several hours. I got 8 alerts per day that the worker is offline above 10 minute ,and if i check the software, maybe it have software problem for 2 times.
And i guess maybe the development fee matters, the developer connect their wallet adrress instead of yours maybe last more than 10 minute. so cut a piece more than the portion than they announced.
the chart time windows meaning method works at lazy strategy. so the chart data is not so accurate.

bobby20180331 commented 3 years ago

well. I guess

1. the program maybe not so stable so maybe offline accur for a while during several hours. I got 8 alerts per day that the worker is offline above 10 minute ,and if i check the software, maybe it have software problem for 2 times.

2. And i guess maybe the develop fee matters, the developer  connect their wallet adrress instead of yours maybe last more than 10 minute. so cut a portion more than the portion than they announced.

3. the chart time windows meaning method works at lazy strategy. so the chart data is not so accurate.

this chart above is based on Nvidia 1660Ti notebook version ,not Xavier.

chamboin commented 3 years ago

Just got 216KH/s, wonder how to improve it cause it should be higher in theory

codegastudio commented 3 years ago

active (running) since Wed 2021-03-10 13:23:15 CET; 22h ago

This is a joke ...

Real-ZeminJiang commented 3 years ago

I have the same issue. I just get 2mh/s hashrate.. Not sure what to do.

codegastudio commented 3 years ago

@ztfot this is a mystery for the 14MH/s I have try many and many things before succeed to run it on Xavier NX. By default, the program not run on it cause it have not enough memory when it check pcie. I have apply a trick to increase the pcie bus speed and this allow program to run. I give you the result i get the first time. After many reset to install from scratch to try to understand and find the way to have this result, i am not able to recover them.

I had a discussion with nvidia guys and they don't understand why the pcie is used in the case of Xavier cause the GPU is not plug on it. So maybe we have a limitation about it. Some specs are getting from pcie for data transfert but there are wrong to exchange with the GPU cause pcie is not the connected to GPU. It's just my feeling about it but i am sure about it.

Someone could explain here why we should get 20MH/s on this hardware plz ? Maybe this will give more data to understood what we are missing.

Regards

bobby20180331 commented 3 years ago

@ztfot this is a mystery for the 14MH/s I have try many and many things before succeed to run it on Xavier NX. By default, the program not run on it cause it have not enough memory when it check pcie. I have apply a trick to increase the pcie bus speed and this allow program to run. I give you the result i get the first time. After many reset to install from scratch to try to understand and find the way to have this result, i am not able to recover them.

I had a discussion with nvidia guys and they don't understand why the pcie is used in the case of Xavier cause the GPU is not plug on it. So maybe we have a limitation about it. Some specs are getting from pcie for data transfert but there are wrong to exchange with the GPU cause pcie is not the connected to GPU. It's just my feeling about it but i am sure about it.

Someone could explain here why we should get 20MH/s on this hardware plz ? Maybe this will give more data to understood what we are missing.

Regards

@codegastudio The Dagger Hashimoto hash used by ETH is intentionally memory constrained such that the relationship between hash rate and memory bandwidth is virtually linear. So, a little math,

A tuned Vega 64 GPU achieves 50MH/S at 125watts, and has a memory bandwidth of 483.8 GB/s. The Xavier AGX memory is rated at a theoretical max. of 137GB/s, so it should deliver up to a quater of 50MH/s,that's about 10~20MH/s

https://github.com/ethereum-mining/ethminer/issues/2158

kivvi3412 commented 3 years ago

So what method do you use to improve Jetson's calculation speed? Use ethminer. In addition, do any of you use f2pool? How are your parameters adjusted?

kivvi3412 commented 3 years ago

codegastudio commented 3 years ago

New data

Below we can see 3 blocs :

Init process to connect 2miners pool and some informations about hardware.
jobs with 0h while DAG data are generated (in 162.3 seconds)
jobs at 144 Kh during more 5h before one job was accepted and processed

mars 16 10:37:35 :  miner Build: linux/release/gnu
mars 16 10:37:35 :  miner 3rd Party: GCC 7.5.0, CUDA 10.2, Boost 1.75.0
mars 16 10:37:35 :  miner 3rd Party: OpenSSL 1.1.1i, Ethash 0.5.0
mars 16 10:37:35 :  miner Configured pool eth.2miners.com:2020
mars 16 10:37:35 :  miner Selected pool eth.2miners.com:2020
mars 16 10:37:35 :  miner Stratum mode : Stratum
mars 16 10:37:35 :  miner Established connection to eth.2miners.com:2020
mars 16 10:37:35 :  miner Spinning up miners...
mars 16 10:37:35 :  cu-0  Using Pci 0000:00:00.0: Xavier (Compute 7.2) Memory : 7.58 GB
mars 16 10:37:35 :  miner Authorized worker 0x42855FF3F65015C295e36D736E3aaC78B5251DB0
mars 16 10:37:35 :  miner Epoch : 401 Difficulty : 8.73 Gh
mars 16 10:37:35 :  miner Job: dbf7256e eth.2miners.com:2020
mars 16 10:37:36 :  miner Job: ed9b4d53 eth.2miners.com:2020
mars 16 10:37:36 :  miner Job: 5520a1cb eth.2miners.com:2020
mars 16 10:37:38 :  cu-0  Required GPU mem: Total 4.20 GB, Cache 66.12 MB, DAG 4.13 GB, Miscellaneous 56.00 B
mars 16 10:37:40 :  miner 0:00 A0 0.00 h - cu0 0.00
mars 16 10:37:43 :  miner Job: ef2c635a eth.2miners.com:2020
mars 16 10:37:43 :  miner Job: be53851a eth.2miners.com:2020
mars 16 10:37:45 :  miner 0:00 A0 0.00 h - cu0 0.00
mars 16 10:37:47 :  miner Job: b54f5d7c eth.2miners.com:2020
mars 16 10:37:47 :  miner Job: f6a10784 eth.2miners.com:2020
mars 16 10:37:50 :  miner 0:00 A0 0.00 h - cu0 0.00
mars 16 10:37:51 :  miner Job: b541e24e eth.2miners.com:2020
...
mars 16 10:40:21 :  cu-0  4.13 GB of DAG data generated in 162.3 seconds
mars 16 10:40:25 :  miner 0:02 A0 1.53 Kh - cu0 1.53
mars 16 10:40:25 :  miner Job: 8019e89c eth.2miners.com:2020
mars 16 10:40:25 :  miner Job: 6af0fad1 eth.2miners.com:2020
mars 16 10:40:30 :  miner 0:02 A0 144.70 Kh - cu0 144.70
mars 16 10:40:32 :  miner Job: 45b7fb0b eth.2miners.com:2020
mars 16 10:40:35 :  miner 0:03 A0 143.71 Kh - cu0 143.71
mars 16 10:40:40 :  miner Job: 43097b92 eth.2miners.com:2020
mars 16 10:40:40 :  miner 0:03 A0 144.06 Kh - cu0 144.06
...
mars 16 16:09:08 :  miner Job: ebe32303 eth.2miners.com:2020
mars 16 16:09:11 :  miner 5:31 A0 143.79 Kh - cu0 143.79
mars 16 16:09:15 :  miner Job: 9f134d30 eth.2miners.com:2020
mars 16 16:09:16 :  miner 5:31 A0 144.05 Kh - cu0 144.05
mars 16 16:09:21 :  miner 5:31 A0 143.84 Kh - cu0 143.84
mars 16 16:09:22 :  miner Job: 1addc88d eth.2miners.com:2020
mars 16 16:09:26 :  miner 5:31 A0 144.37 Kh - cu0 144.37
mars 16 16:09:29 :  miner Job: 51a5ef6f eth.2miners.com:2020 <==========================>
mars 16 16:09:29 :  cu-0  Job: 51a5ef6f Solution: 0xf789c8118ffeaece <=  Spike of 14MH/s   =>
mars 16 16:09:29 :  miner **Accepted  23 ms. eth.2miners.com:2020 <============ =========>

Logs of tegrastats before and during spike (nothing special, GPU is used same all the time)

mars 16 16:09:26 tegrastats: RAM 5491/7766MB (lfb 241x4MB) SWAP 0/3883MB (cached 0MB) CPU [2%@1190,0%@1190,off,off,off,off] EMC_FREQ 9%@1600 GR3D_FREQ 99%@
mars 16 16:09:27 tegrastats: RAM 5491/7766MB (lfb 241x4MB) SWAP 0/3883MB (cached 0MB) CPU [0%@1190,0%@1190,off,off,off,off] EMC_FREQ 9%@1600 GR3D_FREQ 99%@
mars 16 16:09:28 tegrastats: RAM 5491/7766MB (lfb 241x4MB) SWAP 0/3883MB (cached 0MB) CPU [0%@1190,0%@1190,off,off,off,off] EMC_FREQ 9%@1600 GR3D_FREQ 99%@
mars 16 16:09:29 tegrastats: RAM 5491/7766MB (lfb 241x4MB) SWAP 0/3883MB (cached 0MB) CPU [2%@1190,0%@1190,off,off,off,off] EMC_FREQ 9%@1600 GR3D_FREQ 99%@
mars 16 16:09:30 tegrastats: RAM 5491/7766MB (lfb 241x4MB) SWAP 0/3883MB (cached 0MB) CPU [0%@1190,0%@1190,off,off,off,off] EMC_FREQ 9%@1600 GR3D_FREQ 99%@
mars 16 16:09:31 tegrastats: RAM 5491/7766MB (lfb 241x4MB) SWAP 0/3883MB (cached 0MB) CPU [0%@1190,0%@1190,off,off,off,off] EMC_FREQ 9%@1600 GR3D_FREQ 99%@

My miner is running since few days without be stopped. Since the start, i get 7 spikes between 9 MH/s and 14 MH/s. Someone could explain line like this one please miner 5:31 A0 143.79 Kh - cu0 143.79 ? I guess than miner is not mining all the time. It catches the jobs that need to be done but is not selected to do them. Some people could confirm that please ? And if possible explain what are criteria to get the job.

Xavier NX is describe to have a memory at 51.2GB/s. So theoretically, i should get an hashrate around 5 MH/s.

Regards

microwaven commented 3 years ago

The estimates of 5..20MH/s above assume that the performance is equally memory bandwidth bounded as on a dGPU. As you can see in the tegrastats log above, memory is only utilized at 9% while the GPU is 99%. As a result, the GPU can`t keep up with the memory bandwidth. On my AGX I made an interesting experiment and set the GPU to different frequencies:

GPU @ 1377M gives 216KH/s at 8.5GB/s at 4W for the GPU rail GPU @ 905M gives 190KH/s at 7.6GB/s at 1.8W for the GPU rail GPU @ 523M gives 139KH/s at 6.6GB/s at 1W for the GPU rail

The GPU in all 3 data points reports 99% utilization but interestingly the hashrate doesn`t get reduced linearly with the GPU frequency.

I`d say that it is hard to understand why reducing the GPU clock to 65% only reduces the hash rate to 88%. Might indicate that there is optimization potential for the implementation. If you have time, maybe you want to have a look with Nsight what is going on?

JamAndCheese commented 3 years ago

I would like to run some of these experiments. Can you set me up with what source and compilation arguments to use? I have AGX 16GB and a couple of TX2s

hermanw commented 3 years ago

The estimates of 5..20MH/s above assume that the performance is equally memory bandwidth bounded as on a dGPU. As you can see in the tegrastats log above, memory is only utilized at 9% while the GPU is 99%. As a result, the GPU can`t keep up with the memory bandwidth. On my AGX I made an interesting experiment and set the GPU to different frequencies:

GPU @ 1377M gives 216KH/s at 8.5GB/s at 4W for the GPU rail GPU @ 905M gives 190KH/s at 7.6GB/s at 1.8W for the GPU rail GPU @ 523M gives 139KH/s at 6.6GB/s at 1W for the GPU rail

The GPU in all 3 data points reports 99% utilization but interestingly the hashrate doesn`t get reduced linearly with the GPU frequency.

I`d say that it is hard to understand why reducing the GPU clock to 65% only reduces the hash rate to 88%. Might indicate that there is optimization potential for the implementation. If you have time, maybe you want to have a look with Nsight what is going on?

I think the GPU is powerful enough. The bottleneck is still on the memory bandwidth. I also noticed the same interesting point: the EMC is only around 10% all the time. I make a simple comparison test, to allocate 1 GB device buffer and copy bytes inside the buffer. I did two tests:

something like buffer[index] = buffer[index+1]
something like buffer[index] = buffer[random_index] The first one is significantly fast than the 2nd. EMC close to 100% in the 1st and 10% in the second. I guess the memory design is not friendly to random RAM access. Maybe related to cache. It is also true for CPU access to RAM (I also have code tested) because the CPU and GPU share the same memory architecture.

microwaven commented 3 years ago

@JamAndCheese

I would like to run some of these experiments. Can you set me up with what source and compilation arguments to use?

You clone ethminer, add arch=compute_72,code=sm_72 to libethash-cuda/CMakeLists.txt as @codegastudio has shown above and then build from source.

Then you can run ethminer -U -M 0. To see the frequencies and power consumption, run tegrastats. For setting the GPU clock, see https://elinux.org/Jetson/Performance

@hermanw

I think the GPU is powerful enough. The bottleneck is still on the memory bandwidth

I wonder what measurements support your believe.

agallers commented 2 years ago

FYI for those that might still look into this. I don't see how to get about 200khz-ish on the following HW.

Xavier NX 8GB

ethminer...
 m 15:08:02 ethminer 0:14 A0 110.55 Kh - cu0 110.55
 m 15:08:07 ethminer 0:14 A0 110.55 Kh - cu0 110.55
 m 15:08:12 ethminer 0:14 A0 0.00 h - cu0 0.00
 m 15:08:17 ethminer 0:14 A0 0.00 h - cu0 0.00
 m 15:08:22 ethminer 0:15 A0 110.55 Kh - cu0 110.55

tegrastats
RAM 5932/7774MB (lfb 243x4MB) SWAP 0/3887MB (cached 0MB) CPU [1%@1907,0%@1907,0%@1907,0%@1907,off,off] EMC_FREQ 0% GR3D_FREQ 99% AO@42.5C GPU@43C PMIC@100C AUX@42.5C CPU@44C thermal@43.1C
RAM 5932/7774MB (lfb 243x4MB) SWAP 0/3887MB (cached 0MB) CPU [0%@1907,1%@1907,0%@1907,0%@1907,off,off] EMC_FREQ 0% GR3D_FREQ 99% AO@42C GPU@42.5C PMIC@100C AUX@43C CPU@44C thermal@43.15C
RAM 5932/7774MB (lfb 243x4MB) SWAP 0/3887MB (cached 0MB) CPU [0%@1907,0%@1907,0%@1907,0%@1907,off,off] EMC_FREQ 0% GR3D_FREQ 99% AO@42.5C GPU@43C PMIC@100C AUX@42.5C CPU@44C thermal@43.15C
RAM 5932/7774MB (lfb 243x4MB) SWAP 0/3887MB (cached 0MB) CPU [0%@1907,0%@1907,0%@1907,0%@1907,off,off]

sudo /usr/sbin/nvpmodel --query --verbose
NVPM VERB: Config file: /etc/nvpmodel.conf
NVPM VERB: parsing done for /etc/nvpmodel.conf
NV Fan Mode:quiet
NVPM VERB: Current mode: NV Power Mode: MODE_10W_DESKTOP
5

cat /etc/nvpmodel.conf | grep -A 20 'POWER_MODEL' | egrep -i 'POWER_MODEL|EMC|GPU'
...
< POWER_MODEL ID=0 NAME=MODE_15W_2CORE >
GPU_POWER_CONTROL_ENABLE GPU_PWR_CNTL_EN on
GPU MIN_FREQ 0
GPU MAX_FREQ 1109250000
GPU_POWER_CONTROL_DISABLE GPU_PWR_CNTL_DIS auto
EMC MAX_FREQ 1600000000

< POWER_MODEL ID=1 NAME=MODE_15W_4CORE >
GPU_POWER_CONTROL_ENABLE GPU_PWR_CNTL_EN on
GPU MIN_FREQ 0
GPU MAX_FREQ 1109250000
GPU_POWER_CONTROL_DISABLE GPU_PWR_CNTL_DIS auto
EMC MAX_FREQ 1600000000

< POWER_MODEL ID=2 NAME=MODE_15W_6CORE >
GPU_POWER_CONTROL_ENABLE GPU_PWR_CNTL_EN on
GPU MIN_FREQ 0
GPU MAX_FREQ 1109250000
GPU_POWER_CONTROL_DISABLE GPU_PWR_CNTL_DIS auto
EMC MAX_FREQ 1600000000

< POWER_MODEL ID=3 NAME=MODE_10W_2CORE >
GPU_POWER_CONTROL_ENABLE GPU_PWR_CNTL_EN on
GPU MIN_FREQ 0
GPU MAX_FREQ 803250000
GPU_POWER_CONTROL_DISABLE GPU_PWR_CNTL_DIS auto
EMC MAX_FREQ 1600000000

< POWER_MODEL ID=4 NAME=MODE_10W_4CORE >
GPU_POWER_CONTROL_ENABLE GPU_PWR_CNTL_EN on
GPU MIN_FREQ 0
GPU MAX_FREQ 803250000
GPU_POWER_CONTROL_DISABLE GPU_PWR_CNTL_DIS auto
EMC MAX_FREQ 1600000000

< POWER_MODEL ID=5 NAME=MODE_10W_DESKTOP >
GPU_POWER_CONTROL_ENABLE GPU_PWR_CNTL_EN on
GPU MIN_FREQ 0
GPU MAX_FREQ 510000000
GPU_POWER_CONTROL_DISABLE GPU_PWR_CNTL_DIS auto
EMC MAX_FREQ 1600000000

ethereum-mining / ethminer

why Jetson hash rate has only about 100~200kh/s? so far from therory number about 20Mh/s? #2192