ccminer 1.2.0/sgminer 0.1.3 for MTP: Fastest MTP miner for nvidia cards

[djm34]

I am still seeing Trex being faster for my 1660, 2060 and 1070 GPU's under linux.  Especially pool side.

yes, for some reason ccminer is a slower on the xx60 models (doesn't make lots of sense to me... I will probably get a sample to check this out and correct that)
For the 1070, I can't tell much on linux version, but the windows version is between faster and on par with t-rex.
What kind of speed do you get on the 1070 ? (please mention PL, core and mem... linux tends to run on a much slower mem clock this could have an impact somehow)

[1715616851]

1715616851

[1715616851]

1715616851

[1715616851]

1715616851

[dgeo]

yes would be good. How big was the memory leak ?

RAM and swap usage increases until all memory is full

[djm34]

yes would be good. How big was the memory leak ?

RAM and swap usage increases until all memory is full

was there a problem when it happens like a lost of connection ? (this really doesn't happen in normal condition)

[dgeo]

was there a problem when it happens like a lost of connection ? (this really doesn't happen in normal condition)

I haven't logs and screenshots now, but very similar like here:

...Sometimes
- miner stops completely / script loop not working because the process is still alive
- miner is flooded with stratum errors / no accepts
- miner is flooded with increasing nonce / no accepts ...

[djm34]

was there a problem when it happens like a lost of connection ? (this really doesn't happen in normal condition)

I haven't logs and screenshots now, but very similar like here:

...Sometimes
- miner stops completely / script loop not working because the process is still alive
- miner is flooded with stratum errors / no accepts
- miner is flooded with increasing nonce / no accepts ...

I think I have corrected the bulk of these problems, however for the remaining, I will force the miner to crash, so it can be restarted. I will publish an update tonight

[dgeo]

I think I have corrected the bulk of these problems, however for the remaining, I will force the miner to crash, so it can be restarted. I will publish an update tonight

Thanks for the development.

[du44]

Can't run miner on 1066, see error message. Can't run more than 4 gpu to work....
Tested new t-rex - it works fine with 12 gpu rig and does more hashes.

[petunder]

Looking forward to your AMD MTP miner...

he already has one: https://github.com/zcoinofficial/sgminer/releases
I get 1.7 Mh/s with it on rx580 8GB, but also waiting for an update...

I get only 1.1 Mh/s with Vega 56 and this miner

[djm34]

Looking forward to your AMD MTP miner...

he already has one: https://github.com/zcoinofficial/sgminer/releases
I get 1.7 Mh/s with it on rx580 8GB, but also waiting for an update...

I get only 1.1 Mh/s with Vega 56 and this miner

can you show me your command line or config ?
For vega, you need to use the kernel mtp_vega and use a worksize of 64

[djm34]

Can't run miner on 1066, see error message. Can't run more than 4 gpu to work....
Tested new t-rex - it works fine with 12 gpu rig and does more hashes.

what kind of error message are you getting ?
will get a 1066 next week... in anycase for now the 1066 is faster on t-rex
4 gpu's is a little strange, you should get 6 without any problem, after you just need to run 2 instances...

[petunder]

Looking forward to your AMD MTP miner...

he already has one: https://github.com/zcoinofficial/sgminer/releases
I get 1.7 Mh/s with it on rx580 8GB, but also waiting for an update...

I get only 1.1 Mh/s with Vega 56 and this miner

can you show me your command line or config ?
For vega, you need to use the kernel mtp_vega and use a worksize of 64

screenshot:
http://oi66.tinypic.com/24qqn9i.jpg

cmd:
--device 0,1,2 --kernel mtp_vega -o stratum+tcp://zcoin.mintpond.com:3000  -u  aDn7MMYjVQqenT11VFDYHfFdwXmSTRUTak.worker -p 0,strict,verbose,d=700 --intensity 18 --worksize 64

config:
{
  "pools": [
    {  
      "url": "stratum+tcp://zcoin.mintpond.com:3000",
      "user": "aDn7MMYjVQqenT11VFDYHfFdwXmSTRUTak.worker",
      "pass": "0",
      "no-extranonce": true
    }
  ],
  "profiles": [],
  "failover-only": true,
  "algorithm": "mtp_vega",
  "device": "0,1,2",
  "intensity": "22",
  "worksize": "64",
  "temp-cutoff": "95,95",
  "temp-overheat": "90,90",
  "temp-target": "80,80",
  "gpu-memdiff": "0,0",
  "shares": "0",
  "api-mcast-port": "4028",
  "api-port": "4028",
  "expiry": "28",
  "failover-switch-delay": "60",
  "gpu-dyninterval": "7",
  "gpu-platform": "0",
  "hamsi-expand-big": "4",
  "keccak-unroll": "0",
  "log": "5",
  "no-pool-disable": true,
  "no-client-reconnect": true,
  "queue": "1",
  "scan-time": "7",
  "tcp-keepalive": "30",
  "temp-hysteresis": "3",
  "watchpool-refresh": "30"
}

[djm34]

looks ok to me, you probably need to let it stabilizes a bit

[petunder]

looks ok to me, you probably need to let it stabilizes a bit

what MTP hashrate is normal for vega 56? vega 64?

[djm34]

looks ok to me, you probably need to let it stabilizes a bit

what MTP hashrate is normal for vega 56? vega 64?

1.7/1.8  (the best amd card for the moment for mtp is the radeon Vii which is 3.something)

[djm34]

new release: https://github.com/zcoinofficial/ccminer/releases/tag/1.1.23

attempt to solve problem with 6+ gpu's on 2 core/thread cpu's
force the miner to crash in case of a loss of correction (this was removed in the previous version as the problem had been solved, however, it appears some issues may remain).
therefore in case it happens again, the miner will crash so it can be restarted in a loop

[KaydenC]

yes would be good. How big was the memory leak ?

RAM and swap usage increases until all memory is full

was there a problem when it happens like a lost of connection ? (this really doesn't happen in normal condition)

I have the same problem. Solved it by running 7 instances of 1 gpu each.

If I run 7 gpu in a single instance, the memory leak fills up my 4gb ram to 99% within 15 seconds, then hangs the rig.

[notsure85]

thanks for development.
Got 2 more crashes on 22, first one even caused 2 gpus to be "dropped" from system.
Power cycle brings em back.
https://imgur.com/Y3poYFQ
https://imgur.com/7cusMOZ

Second one I have seen for the first time.
Started running .23 just now.

My running line is

:RUN

ccminer -a mtp -o stratum+tcp://xzc.2miners.com:8080 -u XXX.XXX -p x -i 21

timeout /t 5
goto RUN

Rig is corei5/8gb

[notsure85]

thanks for development.
Got 2 more crashes on 22, first one even caused 2 gpus to be "dropped" from system.
Power cycle brings em back.
https://imgur.com/Y3poYFQ
https://imgur.com/7cusMOZ

Second one I have seen for the first time.
Started running .23 just now.

My running line is

:RUN

ccminer -a mtp -o stratum+tcp://xzc.2miners.com:8080 -u XXX.XXX -p x -i 21

timeout /t 5
goto RUN

Rig is corei5/8gb

how much hashrate u got before in this config

~23 mh/s

80% tdp

[du44]

Can't run miner on 1066, see error message. Can't run more than 4 gpu to work....
Tested new t-rex - it works fine with 12 gpu rig and does more hashes.

what kind of error message are you getting ?
will get a 1066 next week... in anycase for now the 1066 is faster on t-rex
4 gpu's is a little strange, you should get 6 without any problem, after you just need to run 2 instances...

cudaErrorInvalidValue - can't run even 1 gpu - looks like problem with indexation ( used command -d 4 ) in 7 gpu rig. 7 gpu's runs ok but slower than t-rex

[djm34]

Can't run miner on 1066, see error message. Can't run more than 4 gpu to work....
Tested new t-rex - it works fine with 12 gpu rig and does more hashes.

what kind of error message are you getting ?
will get a 1066 next week... in anycase for now the 1066 is faster on t-rex
4 gpu's is a little strange, you should get 6 without any problem, after you just need to run 2 instances...

cudaErrorInvalidValue - can't run even 1 gpu - looks like problem with indexation ( used command -d 4 ) in 7 gpu rig. 7 gpu's runs ok but slower than t-rex

which version ? actually the indexation has been re-arranged in the latest version (however this is transparent...). The main problem is with cpu with 2 threads