Pages:
Author

Topic: tdxminer lyra2z/XZC Miner for AMD GPUs on Linux - page 8. (Read 44574 times)

newbie
Activity: 11
Merit: 0

Under water Vega hashes at 7.345Mh/s. But now on 33*C.
I made some testing under Windows. Normal settings for the cards in balance mode are 220W. And cards is throttling hits this power level. But you can set Power up to +50%. Than under full load I have 1630Mhz with 330W in some cases. Worth mentionned that getting 350W is no problem for the LC edition and VRM are able to delver more than 400W.

So the only way to get more H/s is to reflash bios from LC edition or maybe some improved ROCM release. Than, it shoud be around 8Mh/s.
Cheers:D

The gain of productivity is as a percentage equal to the price between cards on water and air? It seems to me that isn't present.
full member
Activity: 602
Merit: 114
did you try to install drivers again?seems driver problem.
or you can try another miner like claymore if it has driver problem or not..
jr. member
Activity: 44
Merit: 5
After a crash my miner is no longer running.
tdxminer => Failed to list OpenCL platforms
clinfo => Number of platforms 0
rom-smi shows both GPU

Does somebody has any idea? Thanks.
newbie
Activity: 34
Merit: 0
I tried installing Ubuntu on windows 10 and installed most drivers, except amd blockchain drivers which gave me an error while installing, and my conclusion is I had zero luck running the miner on windows 10 via Ubuntu platform.

The issue is I get error that it failed to list opencl platforms, is there anyway to get the miner to recognize my gpus?

https://i.imgur.com/LS1hgn9.png
newbie
Activity: 31
Merit: 0
Quote

yours is better. 106W / card. is it easy to adjust the core clock and mem clocks on linux ? i never installed linux for mining.

ohgodatool has no problem overclocking AMD
jr. member
Activity: 308
Merit: 4
Any idea what is the power consumption of AMD RX580 mining with tdxminer on lyra2z algo? More than Nvidia GTX 1080Ti or better?
I think they wont release the windows version?

8x rx580 cc1210 mc1000 cv900 gives 27kh/s and takes 950W, you can probably go lower with power consumption.

yours is better. 106W / card. is it easy to adjust the core clock and mem clocks on linux ? i never installed linux for mining.

Its around 100W per RX 580. If you use HiveOS its easy to control all the OC and voltage settings. With Ubuntu not so much.
full member
Activity: 602
Merit: 114
Any idea what is the power consumption of AMD RX580 mining with tdxminer on lyra2z algo? More than Nvidia GTX 1080Ti or better?
I think they wont release the windows version?

8x rx580 cc1210 mc1000 cv900 gives 27kh/s and takes 950W, you can probably go lower with power consumption.

yours is better. 106W / card. is it easy to adjust the core clock and mem clocks on linux ? i never installed linux for mining.
sr. member
Activity: 428
Merit: 250
Inactivity: 8963
Any idea what is the power consumption of AMD RX580 mining with tdxminer on lyra2z algo? More than Nvidia GTX 1080Ti or better?
I think they wont release the windows version?

8x rx580 cc1210 mc1000 cv900 gives 27kh/s and takes 950W, you can probably go lower with power consumption.
full member
Activity: 602
Merit: 114
My electric cost is so much so i really care that one.
250W/ cards is really high for lyra2z algo.
My GTX 1080 Ti cards power consumption is 125W per card.
full member
Activity: 602
Merit: 114
Any idea what is the power consumption of AMD RX580 mining with tdxminer on lyra2z algo? More than Nvidia GTX 1080Ti or better?
I think they wont release the windows version?
member
Activity: 212
Merit: 12
Our only hope at the moment is the new version of ROCm drivers, which should support newer kernels and offer more features (fingers crossed for undervolting).

Until then, I run my RX Vegas in level 4 and 2  Undecided

newbie
Activity: 11
Merit: 0
Quote

Did you attentively read the reference? And that you will burn the card and all.
Try it

--setoverdrive %            Set GPU OverDrive level (manual|high)

--setoverdrive #: This sets the percentage above maximum for the max Performance Level. For example, --setoverdrive 20 will increase the top sclk level by 20%. If the maximum sclk level is 1000MHz, then --setoverdrive 20 will increase the maximum sclk to 1200MHz

But personally I didn't try it.

May be --setoverdrive High

As a matter of fact I did read the reference thoroughly.

1. Setting --setoverdrive % is valid ONLY for the highest performance level which in case of Vega64 the 7'th. The 7'th performance level is @ 1711Mhz for the core. So increasing --setoverdrive % of 10% set (theoretically) the highest clock to 1882,1Mhz which is impossible to handle by the core at the stock voltage.

2. I don't know why Rocm set the 7'th clock level to 1711MHZ since the card have 7'th set to 1630Mhz under Windows.

3. Rocm can force Vega to work in 7'th clock level without the load( all diodes on the cards are on). Than it is 1630Mhz even though there is not such supported frequency.

First frequencies under load. Second idling.

Code:
====================    ROCm System Management Interface    ====================
================================================================================
 GPU  Temp    AvgPwr   SCLK     MCLK     Fan      Perf    SCLK OD
  3   55.0c   218.0W   1536Mhz  945Mhz   82.75%   high      0%      
  1   58.0c   219.0W   1536Mhz  945Mhz   80.78%   high      0%      
  4   57.0c   219.0W   1536Mhz  945Mhz   82.75%   high      0%      
  2   59.0c   219.0W   1536Mhz  945Mhz   82.75%   high      0%      
  0   N/A     N/A      N/A      N/A      0%       N/A       N/A      
================================================================================
====================           End of ROCm SMI Log          ====================

Code:
====================    ROCm System Management Interface    ====================
================================================================================
 GPU  Temp    AvgPwr   SCLK     MCLK     Fan      Perf    SCLK OD
  3   49.0c   32.0W    1630Mhz  945Mhz   82.75%   high      0%      
  1   54.0c   39.0W    1630Mhz  945Mhz   80.78%   high      0%      
  4   53.0c   32.0W    1630Mhz  945Mhz   82.75%   high      0%      
  2   53.0c   32.0W    1630Mhz  945Mhz   82.75%   high      0%      
  0   N/A     N/A      N/A      N/A      0%       N/A       N/A      
================================================================================
====================           End of ROCm SMI Log          ====================

4. There is no information about what kind of power is displayed in rocm-smi. If this is a hole card power we have a lot of room to growth. But if it is a Core power only. hmm.. I have never seen numbers above 211W. TDP VEGA 64 Air is 295W and LC 345W. Under Windows we can monitor at lest 4 different temperatur of the VEGA 64 (CPU thermal diode, HBM2 temp, Hot spot temp, vrm temp) and power for the core and for the GPU package. Rocm can provide only CPU thermal diode temp and some kind of udefined power consumption.

5. The most important temp in terms of Lyra2Z algo is the Hot spot temperature which we don't have under rocm. There is many speculations about what is a Hot spot temp in fact. I like to belive that it is actually a SoC temp. It has no direct contact with the heatsink so gets very hot. Especially during gaming or, as i presume, during Lyra2z mining. Using cryptonightv7 algo we can observe that the hotspot is up to 10*C higher than the core thermal diode since it is a memory intensive algo. But during gaming the difference can be up to 20*C and easy hit 100*C.

Conclusion:
Vega 64 on air can work up to 6'th freq level when mining (only 6 diodes on the cars in on) The limit is the power. At least the Vega 64 air bios not allowing to pass 220W.
Since the PowerPlay table can not overclock the 6'th or 5'th freq level to achieve higher hashing speed we have to undervolt the card to meet 220W limit. That should boost the core frequency by itself.
Since I don't know any way how to undervolt the the GPU we're back to square one.

At the end of the week I'am planning to install a water cooling on my 4 Vegas. I wonder how the frequency is scaling with the temperature. Than I will re-flash those cards using Vega LC bios.

I'll keep you posted.
Cheers:D



I generally try to receive the maximum productivity, at the smallest consumption of energy. What from us is gradually taken away by producers since it is necessary for them that the video card has failed through certain time what we would buy the new video card.

Indirectly it is confirmed by three monthly guarantee for the video cards mining edition.

Therefore with interest I will watch your experiments.

I wish you good luck.
newbie
Activity: 11
Merit: 0
Quote

https://github.com/RadeonOpenCompute/ROC-smi

--setsclk LEVEL [LEVEL ...] Set GPU Clock Frequency Level Mask (manual)
  --setmclk LEVEL [LEVEL ...] Set GPU Memory Clock Frequency Mask (manual)
  --setfan LEVEL              Set GPU Fan Speed Level
  --resetfans                 Reset GPU fans to automatic control
  --setperflevel LEVEL        Set PowerPlay Performance Level

--setfan 150 (180+ for ref)
--setperflevel MANUAL
--setsclk 5
--setmclk 3

I don't wont setsclk to 5. It is 1401Mhz GPU. I wont to set it to 7 @ 1711Mhz.
This was easy done in Windows. Cards worked great. Here I can't force power play to pass 220W. Setting --setoverdrive to 10 and more crashes the system and does not change anything.

And --setmclk 0-2 does not change the mem freq. It is on level 3 constantly.

Even changing --setprofile # # # # # does nothing.
Code:
GPU[1] : Minimum SCLK: 1600MHz
GPU[1] : Minimum MCLK: 945MHz
GPU[1] : Activity threshold: 0%
GPU[1] : Hysteresis Up: 0ms
GPU[1] : Hysteresis Down: 200ms







Did you attentively read the reference? And that you will burn the card and all.
Try it

--setoverdrive %            Set GPU OverDrive level (manual|high)

--setoverdrive #: This sets the percentage above maximum for the max Performance Level. For example, --setoverdrive 20 will increase the top sclk level by 20%. If the maximum sclk level is 1000MHz, then --setoverdrive 20 will increase the maximum sclk to 1200MHz

But personally I didn't try it.

May be --setoverdrive High
newbie
Activity: 11
Merit: 0
Because tdxminer doesn't show info about temperature and fan speed, you can use utility of rocm drivers. Smiley

If you  open new terminal beside your tdxminer and type:

Code:
/opt/rocm/bin/rocm-smi

you will get info about temperature, average power consumption, fan speed, and other info about your GPUs.
Problem with this command is that it only shows current state of your GPUs. If you want to monitor GPUs info for a certain interval, you can use watch command. Smiley Type in your terminal:

Code:
watch -n 0,5 /opt/rocm/bin/rocm-smi

and you will get your GPUs info every 0.5 second. Smiley

I hope that this post will be helpful for some of you. Cheers. Smiley

This was a great advice. Thanks Cheesy

My rig is working great.
4xVega46 @28.Mh/s on stock settings except the fan speed.

Unfortunetly stock setings becouse I cannot OC.
Tried to use rocm-smi. No luck.

Code:

GPU[2] : Supported GPU clock frequencies on GPU2
GPU[2] : 0: 852Mhz
GPU[2] : 1: 991Mhz
GPU[2] : 2: 1084Mhz
GPU[2] : 3: 1138Mhz
GPU[2] : 4: 1200Mhz
GPU[2] : 5: 1401Mhz
GPU[2] : 6: 1536Mhz *
GPU[2] : 7: 1711Mhz

Code:
/opt/rocm/bin/rocm-smi  --setsclk 7
But that does not change anything.
Since those cards will be water cooled I wondering how to OC?

The only way is to flash my Vegas with LC bios? Really?
Cheers. Smiley



https://github.com/RadeonOpenCompute/ROC-smi

--setsclk LEVEL [LEVEL ...] Set GPU Clock Frequency Level Mask (manual)
  --setmclk LEVEL [LEVEL ...] Set GPU Memory Clock Frequency Mask (manual)
  --setfan LEVEL              Set GPU Fan Speed Level
  --resetfans                 Reset GPU fans to automatic control
  --setperflevel LEVEL        Set PowerPlay Performance Level

--setfan 150 (180+ for ref)
--setperflevel MANUAL
--setsclk 5
--setmclk 3
member
Activity: 212
Merit: 12
I'm making some progress with this miner Smiley

I had issues for a few days, still not sure what went wrong. I wasn't able to connect more than 1 GPU (RX Vega) and I was getting PCIe errors during boot. Finally, the boot process fails with more than one GPU.

I've reinstalled Ubuntu on another motherboard, the ROCm drivers, and it seems everything works just fine now.

I've even written a simple Python program to monitor GPU temps and increase or decrease fan speed. So far so good.

Now, I would also like to be able to undervolt the GPUs but I couldn't find a way to do it using the current ROCm driver.
newbie
Activity: 11
Merit: 0
The more difficulty, the more has to be productivity in Mh to find the solution. If difficulty above the nominal productivity of your video cards, the longer they look for the decision. On the picture it is presented 6x584x24h from constant difficulty. Compare the schedule to the schedule on angrypool.

https://ibb.co/b89sCT

The schedule for constant difficulty for raiven. Too it is very indicative in comparison with other pools.

https://ibb.co/eyvXU8
newbie
Activity: 31
Merit: 0
Quote

Problem in a pool. If it is impossible to set difficulty, there will be a fall of profit.

Minecore.pro there is an opportunity to regulate difficulty. On ravein everything is good, but on gin for some reason there are no payments. If who knows still pools where difficulty can regulate, tell about it please.

On the picture two 8x1080ti from the set difficulty, and 6x584 (20,5Mh) difficulty appointed by a pool.

https://ibb.co/gUM82T

and why not put 6x584 manual difficulty? and I do not quite understand what gives manual complexity and how it affects the profit  Roll Eyes Roll Eyes
newbie
Activity: 11
Merit: 0
confirm, the problem is on lyra2z. On phi2 or any other algorithm 3-4% deviation with calculators but not 27%... And about the fact that the calculators specifically do it - it's not exactly, just the Lira was apparently not designed for such a power network, and now there is an error in the calculation of complexity.... Generally when podseti arrived at лиpe2з necessary from the final profit take 27%  Grin

Problem in a pool. If it is impossible to set difficulty, there will be a fall of profit.

Minecore.pro there is an opportunity to regulate difficulty. On ravein everything is good, but on gin for some reason there are no payments. If who knows still pools where difficulty can regulate, tell about it please.

On the picture two 8x1080ti from the set difficulty, and 6x584 (20,5Mh) difficulty appointed by a pool.

https://ibb.co/gUM82T
newbie
Activity: 9
Merit: 0
I am very impressed with the hashrate this miner is getting on amd.  That being said... It's glitchy, it overheats gpus like crazy and its still in beta mode.  I must say that 3% dev fee is uncalled for.
newbie
Activity: 9
Merit: 0
I get around 2.6 mh/s on my rx 470 cards, but only 1.1 mh/s on my gpu0. Any ideas?
Pages:
Jump to: