Pages:
Author

Topic: [ANN] dstm's ZCash / Equihash Nvidia Miner v0.6.2 (Linux / Windows) - page 17. (Read 224961 times)

newbie
Activity: 22
Merit: 2
What are "ALL" you're settings for you're GPUz? what % TPD? I run 200/660/90% in AB and constant voltage, looks like u need to turn TPD % down to 80-90%, or lower you're Overclawk Settings q=]  

How much power do your cards use at that settings?
Going for 4 sol/watt here at 70W per card (1060 gb oc)...
Hope this helps, Zardozer, sorry forgot to post these links, u can get 4sols/watt but you'll need to turn down TPD to about 78%, inrun kernel mode not user and force constant voltage in Afterburner, you can see clock settings, this is my new rig I made with some misc GPUz, I'll be buying more 1060z you get slot of bang for the buck with them, I've tried every miner with several rigs, zm miner (dstm) is by far the best/most efficient/most stable+, hope this helps q=]

https://ibb.co/jTRuj7
https://ibb.co/j33krn
legendary
Activity: 1260
Merit: 1046
Hi DSTM,

+5 merit for thank you to your job on this miner software :-).
newbie
Activity: 15
Merit: 0
What are "ALL" you're settings for you're GPUz? what % TPD? I run 200/660/90% in AB and constant voltage, looks like u need to turn TPD % down to 80-90%, or lower you're Overclawk Settings q=]  

How much power do your cards use at that settings?
Going for 4 sol/watt here at 70W per card (1060 gb oc)...
newbie
Activity: 22
Merit: 2
What are "ALL" you're settings for you're GPUz? what % TPD? I run 200/660/90% in AB and constant voltage, looks like u need to turn TPD % down to 80-90%, or lower you're Overclawk Settings q=] 
newbie
Activity: 1
Merit: 0
5x ASUS Dual GeForce GTX 1060 OC, DUAL-GTX1060-O6G, 6GB GDDR5

Core+145
Mem+500

Code:
2018-04-18 01:34:46|gpu_id 2 68 1 an illegal memory access was encountered
2018-04-18 01:34:46|gpu 2 unresponsive - check overclocking
2018-04-18 01:34:46|cudaMemcpy 1 failed

2018-04-19 05:12:25|gpu_id 2 87 0 an illegal memory access was encountered
2018-04-19 05:12:25|gpu 2 unresponsive - check overclocking
2018-04-19 05:12:25|cudaMemcpy 1 failed

What do the numbers after the gpu_id mean?
Please let me know what is causing the errors.

Cheers,
newbie
Activity: 10
Merit: 0
 --color          colorize the output ??

This beautiful piece of software needs more colorization options.

2018-04-19 12:01:54 PM|   ========== Sol/s: 5395.8 Sol/W: 3.11  Avg: 5419.2 I/s: 2897.5 Sh: 4.60   1.00 33

This stats line need to be controllable with user choice of color. I would prefer yellow or amber or orange to make it stand out.

The average column would be nice if it was also color controllable.

This is the only thing this software is lacking. Other than that, it's fantastic!

Cheers,

 
newbie
Activity: 11
Merit: 0
I run 8 x 1070 in one rig, 6 x 1080Ti in one rig and 5 x 1080Ti in one rig. they all run together in their own rig build. I mean, I run zm batch to work 8x1070 together and others respectably in their own.

while running, the target difficulty is set by the server (any chance to achieve static difficulty through miner?). the target difficulty is somehow too high for the 1080Tis which in return I get only 7,5 - 10 shares per minute in average; that being said I get zero returns on payment on some blocks randomly.

but with the 1070s, even though they are %60 solutions capable against 1080Tis, I have around 60 shares per minute in average.

you can say that the size of the share 1080Tis get is larger than 1070, I agree, but it is like more than 6 times the share I get with 1080Tis. so considering the solution power of two, 6 times is much bigger than %60 solution power in between these two.

I've tried this while mining BTCG in suprnova and MPH. I think this is because the target difficulty setting.

Honestly, I think you'll need to wait for a payout to see if you can see the earnings differences - looking at hash or share rate alone (particularly at the pool level!) to compare completely different rigs isn't going to be accurate. Earnings per rig is pretty much the only way to determine anything.

That said, if you're looking to have the same target difficulty across all your rigs, you might try running a mining proxy. Makes things slightly more efficient networking wise as well. I have recently created one that works with equihash (and other algos/coins) called aiostratum-proxy (see my signature as well).

thanks for your interest.

I speak of share numbers alone just to indicate the problem easier. other than that, I do check the earnings. for example in transaction history, 1080Tis get way low earnings despite their hash power. many shares in chunks surpasses low share numbers in larger -so to speak- amounts. that's why I ask about static target difficulty.

I will check your post, thank you.

just a suggestion to check,is there some stability problem you might miss?there should be comparable difference hash-power to payout,unless that miner is really unlucky(low number of shares because of really bad luck)...but give it few days and it should stabilise (also pplns vs other payout methods,don't switch pools)...so,check log if there are no restarts/reconnects or some gpu slowing down
newbie
Activity: 5
Merit: 0
Just for a benchmark I'm running two rigs each with a EVGA 1070 and a MSI GTX 1050ti and getting 650ish Sol's (450 on the 1070 and 185-190 on the 1050ti). 1070 is set to 65% power limit, 75 core clock, 100 mem and the 1050ti is at 80% power, 150 core and 200 mem.
full member
Activity: 350
Merit: 126
I run 8 x 1070 in one rig, 6 x 1080Ti in one rig and 5 x 1080Ti in one rig. they all run together in their own rig build. I mean, I run zm batch to work 8x1070 together and others respectably in their own.

while running, the target difficulty is set by the server (any chance to achieve static difficulty through miner?). the target difficulty is somehow too high for the 1080Tis which in return I get only 7,5 - 10 shares per minute in average; that being said I get zero returns on payment on some blocks randomly.

but with the 1070s, even though they are %60 solutions capable against 1080Tis, I have around 60 shares per minute in average.

you can say that the size of the share 1080Tis get is larger than 1070, I agree, but it is like more than 6 times the share I get with 1080Tis. so considering the solution power of two, 6 times is much bigger than %60 solution power in between these two.

I've tried this while mining BTCG in suprnova and MPH. I think this is because the target difficulty setting.

Honestly, I think you'll need to wait for a payout to see if you can see the earnings differences - looking at hash or share rate alone (particularly at the pool level!) to compare completely different rigs isn't going to be accurate. Earnings per rig is pretty much the only way to determine anything.

That said, if you're looking to have the same target difficulty across all your rigs, you might try running a mining proxy. Makes things slightly more efficient networking wise as well. I have recently created one that works with equihash (and other algos/coins) called aiostratum-proxy (see my signature as well).

thanks for your interest.

I speak of share numbers alone just to indicate the problem easier. other than that, I do check the earnings. for example in transaction history, 1080Tis get way low earnings despite their hash power. many shares in chunks surpasses low share numbers in larger -so to speak- amounts. that's why I ask about static target difficulty.

I will check your post, thank you.

Difficulty is set by the server. Mining software has to submit shares which satisfy the difficulty requirements set by the server otherwise the server will reject the shares. However some servers allow you to request a particular difficulty, they often simply use/'misuse' the password field for it. In this case the server reads the password field (containing your difficulty request) and starts to submit jobs according to your requested difficulty.

If the math is done right on the pool side there should be really no disadvantages for your 1080Ti rig - however there is probability involved so you really have to compare the numbers (e.g. payouts) of multiple days.
full member
Activity: 350
Merit: 126
Any info about dstm v7 ?

I'm working on it - I hope to be able to release it in 2 weeks.

  this is good news!
more power is expected in version v 0.7?

On linux systems potentially.

Hello dstm,

is it possible to make your miner compatible with Equihash 192,7 parameters ?
I hope so because there is only 1 for these parameters and it completely sucks and doesnt get updated...
So if your Miner would be only a bit more stable and would have a cuda implementation you would get the majority of miners on your miner.
That would generate you like atleast 2k Zero (which equals to about 2k dollars) a month.

If you want a bounty that is also possible. I look foward to your response.


Other Equihash instances is really something I want to support in future versions since it was requested multiple times - however zm's core development consumes a lot time currently so I can't say anything about the time frame.
newbie
Activity: 3
Merit: 0
I run 8 x 1070 in one rig, 6 x 1080Ti in one rig and 5 x 1080Ti in one rig. they all run together in their own rig build. I mean, I run zm batch to work 8x1070 together and others respectably in their own.

while running, the target difficulty is set by the server (any chance to achieve static difficulty through miner?). the target difficulty is somehow too high for the 1080Tis which in return I get only 7,5 - 10 shares per minute in average; that being said I get zero returns on payment on some blocks randomly.

but with the 1070s, even though they are %60 solutions capable against 1080Tis, I have around 60 shares per minute in average.

you can say that the size of the share 1080Tis get is larger than 1070, I agree, but it is like more than 6 times the share I get with 1080Tis. so considering the solution power of two, 6 times is much bigger than %60 solution power in between these two.

I've tried this while mining BTCG in suprnova and MPH. I think this is because the target difficulty setting.

Honestly, I think you'll need to wait for a payout to see if you can see the earnings differences - looking at hash or share rate alone (particularly at the pool level!) to compare completely different rigs isn't going to be accurate. Earnings per rig is pretty much the only way to determine anything.

That said, if you're looking to have the same target difficulty across all your rigs, you might try running a mining proxy. Makes things slightly more efficient networking wise as well. I have recently created one that works with equihash (and other algos/coins) called aiostratum-proxy (see my signature as well).

thanks for your interest.

I speak of share numbers alone just to indicate the problem easier. other than that, I do check the earnings. for example in transaction history, 1080Tis get way low earnings despite their hash power. many shares in chunks surpasses low share numbers in larger -so to speak- amounts. that's why I ask about static target difficulty.

I will check your post, thank you.
jr. member
Activity: 95
Merit: 2
I run 8 x 1070 in one rig, 6 x 1080Ti in one rig and 5 x 1080Ti in one rig. they all run together in their own rig build. I mean, I run zm batch to work 8x1070 together and others respectably in their own.

while running, the target difficulty is set by the server (any chance to achieve static difficulty through miner?). the target difficulty is somehow too high for the 1080Tis which in return I get only 7,5 - 10 shares per minute in average; that being said I get zero returns on payment on some blocks randomly.

but with the 1070s, even though they are %60 solutions capable against 1080Tis, I have around 60 shares per minute in average.

you can say that the size of the share 1080Tis get is larger than 1070, I agree, but it is like more than 6 times the share I get with 1080Tis. so considering the solution power of two, 6 times is much bigger than %60 solution power in between these two.

I've tried this while mining BTCG in suprnova and MPH. I think this is because the target difficulty setting.

Honestly, I think you'll need to wait for a payout to see if you can see the earnings differences - looking at hash or share rate alone (particularly at the pool level!) to compare completely different rigs isn't going to be accurate. Earnings per rig is pretty much the only way to determine anything.

That said, if you're looking to have the same target difficulty across all your rigs, you might try running a mining proxy. Makes things slightly more efficient networking wise as well. I have recently created one that works with equihash (and other algos/coins) called aiostratum-proxy (see my signature as well).
jr. member
Activity: 95
Merit: 2
Is there anything for Ubuntu, how MSI AFTenberner for Windows?

Yes, there are lots of posts all over the internet about how to adjust OC settings in Linux, but there aren't any (I don't think?) graphical tools like Afterburner. You use the command line (see below), and once you figure things out, it's actually better than Afterburner. Everything based in percentages is really useless for sharing with others. For example, telling someone you set the power to 70% is useless without knowing what the TDP is for a particular card. However, telling someone the specific wattage setting + the card is better.

Here's what I use with my 6x Asus ROG Strix 1070s:

Code:
# These two lines allow me to run the rig 'headless' (without a monitor)
# Without them, the nvidia commands below won't run unless you log into the Ubuntu graphical UI.
export XAUTHORITY=/var/run/lightdm/root/:0
export DISPLAY=:0

# This enables power management
sudo nvidia-smi -pm 1
# This sets the desired power consumption to 115 watts (no percentages here; specificity!)
sudo nvidia-smi -pl 115
# this tweaks core and memory clock settings and sets the fan speed (ok, fan speed is a percentage...)
sudo nvidia-settings -a GPUPowerMizerMode=1 -a GPUGraphicsClockOffset[3]=200 -a GPUMemoryTransferRateOffset[3]=1050 -a GPUFanControlState=1 -a GPUTargetFanSpeed=25
jr. member
Activity: 95
Merit: 2
One of my rig is constantly crashing - "cuda 1 memory failed". It has 6 msi armor 1060 6gbs.

I have tried almost all clocks in order to stabilise it.
Now I am running with no clocks only 80% power limit and it keeps crashing. I noted that when temp limit is set to 70C rig is more stable.

Can you please someone assist me as I am desperate for help. Please help me what can be the problem here?!

As you've tried different clocks, I'd guess it's a hardware issue. Perhaps riser/cable related - it isn't uncommon from what I've read/seen.
newbie
Activity: 3
Merit: 0
hey there,

thanks for your work dstm. I'm gonna ask a question about not being able to achieve enough shares.

I run 8 x 1070 in one rig, 6 x 1080Ti in one rig and 5 x 1080Ti in one rig. they all run together in their own rig build. I mean, I run zm batch to work 8x1070 together and others respectably in their own.

while running, the target difficulty is set by the server (any chance to achieve static difficulty through miner?). the target difficulty is somehow too high for the 1080Tis which in return I get only 7,5 - 10 shares per minute in average; that being said I get zero returns on payment on some blocks randomly.

but with the 1070s, even though they are %60 solutions capable against 1080Tis, I have around 60 shares per minute in average.

you can say that the size of the share 1080Tis get is larger than 1070, I agree, but it is like more than 6 times the share I get with 1080Tis. so considering the solution power of two, 6 times is much bigger than %60 solution power in between these two.

I've tried this while mining BTCG in suprnova and MPH. I think this is because the target difficulty setting.

thanks,
newbie
Activity: 16
Merit: 0
Any info about dstm v7 ?

I'm working on it - I hope to be able to release it in 2 weeks.

  this is good news!
more power is expected in version v 0.7?

On linux systems potentially.

thanks for the answer ,
this is great!
Is there anything for Ubuntu, how MSI AFTenberner for Windows?
newbie
Activity: 3
Merit: 0
Any info about dstm v7 ?

I'm working on it - I hope to be able to release it in 2 weeks.

  this is good news!
more power is expected in version v 0.7?

On linux systems potentially.

Hello dstm,

is it possible to make your miner compatible with Equihash 192,7 parameters ?
I hope so because there is only 1 for these parameters and it completely sucks and doesnt get updated...
So if your Miner would be only a bit more stable and would have a cuda implementation you would get the majority of miners on your miner.
That would generate you like atleast 2k Zero (which equals to about 2k dollars) a month.

If you want a bounty that is also possible. I look foward to your response.
full member
Activity: 350
Merit: 126
I am still a bit confused about the following: Is temp target parameter really temperature limit parameter or will the miner try to reach and maintain this temperature as the word "target" suggests? Thank you.

ZM maximizes the amount of submitted work (i.e. the utilization of the GPU) such that the temperature of the GPU is less or equal to the temp-target.
full member
Activity: 350
Merit: 126
I am using DSTM to all my rigs. I am mining with 55 gpus NVIDIA only.

One of my rig is constantly crashing - "cuda 1 memory failed". It has 6 msi armor 1060 6gbs.

I have tried almost all clocks in order to stabilise it.
Now I am running with no clocks only 80% power limit and it keeps crashing. I noted that when temp limit is set to 70C rig is more stable.

Can you please someone assist me as I am desperate for help. Please help me what can be the problem here?!

Regards,
Hristo

The error message contains also the gpu-id of the GPU which failed to execute this particular call. Try to disable the affected GPU by using the '--dev' option - it could be faulty. In general: if you have stability issues on multi-gpu systems, try to to run zm on a single GPU only, in order to minimize the effects of thermal or power supply issues.
full member
Activity: 350
Merit: 126
Any info about dstm v7 ?

I'm working on it - I hope to be able to release it in 2 weeks.

  this is good news!
more power is expected in version v 0.7?

On linux systems potentially.
Pages:
Jump to: