Author

Topic: [ANN] TeamRedMiner v0.10.10 - Ironfish/Kaspa/ZIL/Kawpow/Etchash and More - page 121. (Read 211877 times)

jr. member
Activity: 194
Merit: 4
Watchdog could not restart the miner.

[2018-12-03 08:41:38] GPU 7: detected DEAD (09:00.0), will execute restart script watchdog.bat
[2018-12-03 08:41:38] Watchdog triggering miner shutdown after restart script execution.
[2018-12-03 08:41:38] Shutting down...
Watchdog restarting miner in 30 secs.

Bpeмя oжидaния 30 ceк., нaжмитe любyю клaвишy для пpoдoлжeния ...
[2018-12-03 08:41:38] GPU 3 CN thread 1 exiting.
[2018-12-03 08:41:38] GPU 0 CN thread 1 exiting.
[2018-12-03 08:41:38] GPU 4 CN thread 1 exiting.
[2018-12-03 08:41:38] GPU 9 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 6 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 8 CN thread 0 exiting.
               29[2018-12-03 08:41:39] GPU10 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 1 CN thread 1 exiting.
[2018-12-03 08:41:39] GPU 5 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 2 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 0 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 9 CN thread 1 exiting.
[2018-12-03 08:41:39] GPU 8 CN thread 1 exiting.
[2018-12-03 08:41:39] GPU10 CN thread 1 exiting.
[2018-12-03 08:41:39] GPU 4 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 6 CN thread 1 exiting.
[2018-12-03 08:41:39] GPU 3 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 2 CN thread 1 exiting.
[2018-12-03 08:41:39] GPU 1 CN thread 0 exiting.
               28[2018-12-03 08:41:40] GPU 5 CN thread 1 exiting.
               20[2018-12-03 08:41:48] GPU 7 thread 0 shutdown timed out.
[2018-12-03 08:41:48] GPU 7 thread 1 shutdown timed out.
[2018-12-03 08:41:48] Unclean shutdown.
                0
          Team Red Miner version 0.3.8
[2018-12-03 08:42:40] Auto-detected AMD OpenCL platform 0

The problem is from the way the miner is being started and stopped.

So far, stopping only happens with "CTRL+C", and since its started via .BAT script, it would throw a user prompt for
Code:
Terminate batch script? Y/N
newbie
Activity: 15
Merit: 0
Watchdog could not restart the miner.

[2018-12-03 08:41:38] GPU 7: detected DEAD (09:00.0), will execute restart script watchdog.bat
[2018-12-03 08:41:38] Watchdog triggering miner shutdown after restart script execution.
[2018-12-03 08:41:38] Shutting down...
Watchdog restarting miner in 30 secs.

Bpeмя oжидaния 30 ceк., нaжмитe любyю клaвишy для пpoдoлжeния ...
[2018-12-03 08:41:38] GPU 3 CN thread 1 exiting.
[2018-12-03 08:41:38] GPU 0 CN thread 1 exiting.
[2018-12-03 08:41:38] GPU 4 CN thread 1 exiting.
[2018-12-03 08:41:38] GPU 9 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 6 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 8 CN thread 0 exiting.
               29[2018-12-03 08:41:39] GPU10 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 1 CN thread 1 exiting.
[2018-12-03 08:41:39] GPU 5 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 2 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 0 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 9 CN thread 1 exiting.
[2018-12-03 08:41:39] GPU 8 CN thread 1 exiting.
[2018-12-03 08:41:39] GPU10 CN thread 1 exiting.
[2018-12-03 08:41:39] GPU 4 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 6 CN thread 1 exiting.
[2018-12-03 08:41:39] GPU 3 CN thread 0 exiting.
[2018-12-03 08:41:39] GPU 2 CN thread 1 exiting.
[2018-12-03 08:41:39] GPU 1 CN thread 0 exiting.
               28[2018-12-03 08:41:40] GPU 5 CN thread 1 exiting.
               20[2018-12-03 08:41:48] GPU 7 thread 0 shutdown timed out.
[2018-12-03 08:41:48] GPU 7 thread 1 shutdown timed out.
[2018-12-03 08:41:48] Unclean shutdown.
                0
          Team Red Miner version 0.3.8
[2018-12-03 08:42:40] Auto-detected AMD OpenCL platform 0
newbie
Activity: 15
Merit: 0
Faced a problem. The miner just stops working. It does not hang and can be closed.
member
Activity: 277
Merit: 23
It's a good result anyway, I have couple rigs 6x NITRO+ with with hynix can't pass 6.06 for entire rig with c1280 m2110 

6.06Kh/s ? Thats still an OK result. I would assume the Hynix memory is MJR. They are notreally good.

Worst possible combo

edw4032babg Micron/Eplida
h5gc4h24ajr Hynix

The Hynix AJR and the Elpida BABG are really good. They may not clock up high, but they do like tight timings.
With both of them, i was getting around 1Kh/s on XMR-Stak at ~1300Mhz core.

Just got 6.11H/s C1280 m2120

This is 580 card on my pulse 570 with pure hynix (same memory) i get easily same result.

Maybe I got spoiled with samsung Cheesy
jr. member
Activity: 194
Merit: 4
It's a good result anyway, I have couple rigs 6x NITRO+ with with hynix can't pass 6.06 for entire rig with c1280 m2110 

6.06Kh/s ? Thats still an OK result. I would assume the Hynix memory is MJR. They are notreally good.

Worst possible combo

edw4032babg Micron/Eplida
h5gc4h24ajr Hynix

The Hynix AJR and the Elpida BABG are really good. They may not clock up high, but they do like tight timings.
With both of them, i was getting around 1Kh/s on XMR-Stak at ~1300Mhz core.
member
Activity: 277
Merit: 23
It's a good result anyway, I have couple rigs 6x NITRO+ with with hynix can't pass 6.06 for entire rig with c1280 m2110 

6.06Kh/s ? Thats still an OK result. I would assume the Hynix memory is MJR. They are notreally good.

Worst possible combo

edw4032babg Micron/Eplida
h5gc4h24ajr Hynix
jr. member
Activity: 194
Merit: 4
It's a good result anyway, I have couple rigs 6x NITRO+ with with hynix can't pass 6.06 for entire rig with c1280 m2110 

6.06Kh/s ? Thats still an OK result. I would assume the Hynix memory is MJR. They are notreally good.
member
Activity: 277
Merit: 23
It's a good result anyway, I have couple rigs 6x NITRO+ with with hynix can't pass 6.06 for entire rig with c1280 m2110 
jr. member
Activity: 194
Merit: 4

Yes, the memclock changes do take effect. And yes, on Windows 10.

Which drivers?

I have Nitro+ 4gb samsung they work slightly better onBeta then on 18.6.1

I'm getting identical results as you

18.2.1.
member
Activity: 277
Merit: 23

Yes, the memclock changes do take effect. And yes, on Windows 10.

Which drivers?

I have Nitro+ 4gb samsung they work slightly better onBeta then on 18.6.1

I'm getting identical results as you
jr. member
Activity: 194
Merit: 4
I know what you mean, but i tested the scaling and....


Code:

Elpida 1280Mhz Core GPU 3
1850 - 1047H/s
1875 - 1050-1052H/s
1900 - 1052-1054H/s
1925 - 1052-1054H/s
1950 - 1043H/s


------

Elpida 1875Mhz Mem GPU 3

1280 - 1050-1052H/s
1250 - 1028H/s
1225 - 1010H/s
1200 - 991H/s

------
Elpida 1280Mhz Core GPU 0
1900 - 1049-1050H/s
1925 - 1052H/s
1950 - 1054H/s
1975 - 1057H/s

------

Elpida 1900Mhz Mem GPU 0

1280 - 1049-1050H/s
1250 - 1028-1029H/s
1225 - 1010-1011H/s
1200 - 991-992H/s

------

Samsung 2100Mhz Mem GPU 1

1280 - 1057H/s
1250 - 1035H/s
1225 - 1014H/s
1200 - 995H/s

------

Samsung 1280Mhz Core GPU 1

2075 - 1055H/s
2100 - 1057H/s
2125 - 1059-1061H/s
2150 - 1060-1061H/s


First is Memory clock scaling, then Core clock scaling.

GPU 0/3 -> 570/580 Elpida.
GPU 1 -> 580 Samsung.
Seems weird, for 75Mhz Memclock increase, to get only about 10Hs. Usually, it should be around 30H/s for ~75Mhz increase (at least on XMR-Stak).


EDIT: For clarification, are HWs actually considered Bad shares, that failed CPU check (meaning, GPU worked on them, and before submitting, it failed) ?

Our tests have shown that memory clock should have a noticeable effect.  Are you testing on windows?  Did you make sure that your memclock changes are taking effect by checking actual memclock speeds in say hwinfo?

As for the HWs, that is the number of shares that the GPU found that failed CPU verification.  This indicates that something went wrong on the GPU and it performed the share calculations incorrectly.  Typically, this is due to errors from aggressive core/mem clocks and/or voltages.
The miner does not submit these shares to the pool as it already knows that they are invalid.

Yes, the memclock changes do take effect. And yes, on Windows 10.
member
Activity: 176
Merit: 76
Another API woe:

GPU3 crashed, and was declared Dead, but the API status was still showing Alive.

http://i67.tinypic.com/2s7gxi1.png

The program seen there checks the JSON data and the Status Field specifically is like this:

Code:
GpuDetails.Substring(GpuDetails.IndexOf("Status=") + 7, GpuDetails.IndexOf(",Temperature") - (GpuDetails.IndexOf("Status=") + 7));


That is indeed a bug.  Sorry about that.  We'll fix it in the next release.
member
Activity: 176
Merit: 76
I know what you mean, but i tested the scaling and....


Code:

Elpida 1280Mhz Core GPU 3
1850 - 1047H/s
1875 - 1050-1052H/s
1900 - 1052-1054H/s
1925 - 1052-1054H/s
1950 - 1043H/s


------

Elpida 1875Mhz Mem GPU 3

1280 - 1050-1052H/s
1250 - 1028H/s
1225 - 1010H/s
1200 - 991H/s

------
Elpida 1280Mhz Core GPU 0
1900 - 1049-1050H/s
1925 - 1052H/s
1950 - 1054H/s
1975 - 1057H/s

------

Elpida 1900Mhz Mem GPU 0

1280 - 1049-1050H/s
1250 - 1028-1029H/s
1225 - 1010-1011H/s
1200 - 991-992H/s

------

Samsung 2100Mhz Mem GPU 1

1280 - 1057H/s
1250 - 1035H/s
1225 - 1014H/s
1200 - 995H/s

------

Samsung 1280Mhz Core GPU 1

2075 - 1055H/s
2100 - 1057H/s
2125 - 1059-1061H/s
2150 - 1060-1061H/s


First is Memory clock scaling, then Core clock scaling.

GPU 0/3 -> 570/580 Elpida.
GPU 1 -> 580 Samsung.
Seems weird, for 75Mhz Memclock increase, to get only about 10Hs. Usually, it should be around 30H/s for ~75Mhz increase (at least on XMR-Stak).


EDIT: For clarification, are HWs actually considered Bad shares, that failed CPU check (meaning, GPU worked on them, and before submitting, it failed) ?

Our tests have shown that memory clock should have a noticeable effect.  Are you testing on windows?  Did you make sure that your memclock changes are taking effect by checking actual memclock speeds in say hwinfo?

As for the HWs, that is the number of shares that the GPU found that failed CPU verification.  This indicates that something went wrong on the GPU and it performed the share calculations incorrectly.  Typically, this is due to errors from aggressive core/mem clocks and/or voltages.
The miner does not submit these shares to the pool as it already knows that they are invalid.
newbie
Activity: 8
Merit: 0
Is there a watchdog.sh available on Linux?

i'm trying to get a watchdog for Linux, I googled around, I could not find anything...
sr. member
Activity: 1484
Merit: 253
On nicehash pool many rejects "Error code: 2 - job not found." Can it be fixed?
member
Activity: 176
Merit: 10
Another API woe:

GPU3 crashed, and was declared Dead, but the API status was still showing Alive.

http://i67.tinypic.com/2s7gxi1.png

The program seen there checks the JSON data and the Status Field specifically is like this:

Code:
GpuDetails.Substring(GpuDetails.IndexOf("Status=") + 7, GpuDetails.IndexOf(",Temperature") - (GpuDetails.IndexOf("Status=") + 7));


hi  same here on 1 rig.

whatchdog says gpu X is dead but no:/
and stop working at the reload...."unclean quit" or something like that
full member
Activity: 1120
Merit: 131
Is there a reference setup for RX574 ?

maybe something like (using 8+7):
1244   937   1900   900  (965 h/s, 113 W) (really stable power draw compared to the other settings which have +5 +10 W peaks)
or
1100 912    1900 900 (865 h/s, 101 W)

still figuring stuff out, but higher mem clocks dont show a gain, or cause a drop, and voltage is important, if the hash stops, increase it by 25 or 12 and try again.

As mentioned above a little higher voltage needed than for say SRB, which I was running at:
1175   812   1750   800 (833 h/s, 92 W)

Its a 'bad' card based on previous xmr-stak and SRB results, compared to the other cards I have.

All mining on CNv8, graft on grft.codpool.com

Thanks for the info  Smiley
jr. member
Activity: 194
Merit: 4
Another API woe:

GPU3 crashed, and was declared Dead, but the API status was still showing Alive.

http://i67.tinypic.com/2s7gxi1.png

The program seen there checks the JSON data and the Status Field specifically is like this:

Code:
GpuDetails.Substring(GpuDetails.IndexOf("Status=") + 7, GpuDetails.IndexOf(",Temperature") - (GpuDetails.IndexOf("Status=") + 7));
jr. member
Activity: 194
Merit: 4
I have noticed something weird.

I have 4 cards.

1xRX570 4GB Elpida.
3xRX580 4GB, 2 of which Elpida.

The RX570 hashes absolutely the same as the RX580, despite the differences in the CUs.
Both cards are running SAME straps, SAME core clocks, SAME memory clocks.

You’re describing the very definition of a bandwidth limited process. Since they have the same memory controller and same straps and clocks, assuming you’re running 8+8, it’s not very surprising they end up with the same perf. That said, there is another aspect involved though that we will address in the future, might see a little per bump on the 580 then.

I have noticed something weird.

I have 4 cards.

1xRX570 4GB Elpida.
3xRX580 4GB, 2 of which Elpida.

The RX570 hashes absolutely the same as the RX580, despite the differences in the CUs.
Both cards are running SAME straps, SAME core clocks, SAME memory clocks.

It think it's because CryptoNight seems depending very much at memory, not so much at core performance. Like my RX 550 8 CU gets better hashrates than my RX 560 with 14/16 CUs. All about memory performance I think. I run all of my RX 550/560/570/580s at a very low core clock (1050-1150 MHz) and gets a good hashrate even then..

I know what you mean, but i tested the scaling and....


Code:

Elpida 1280Mhz Core GPU 3
1850 - 1047H/s
1875 - 1050-1052H/s
1900 - 1052-1054H/s
1925 - 1052-1054H/s
1950 - 1043H/s


------

Elpida 1875Mhz Mem GPU 3

1280 - 1050-1052H/s
1250 - 1028H/s
1225 - 1010H/s
1200 - 991H/s

------
Elpida 1280Mhz Core GPU 0
1900 - 1049-1050H/s
1925 - 1052H/s
1950 - 1054H/s
1975 - 1057H/s

------

Elpida 1900Mhz Mem GPU 0

1280 - 1049-1050H/s
1250 - 1028-1029H/s
1225 - 1010-1011H/s
1200 - 991-992H/s

------

Samsung 2100Mhz Mem GPU 1

1280 - 1057H/s
1250 - 1035H/s
1225 - 1014H/s
1200 - 995H/s

------

Samsung 1280Mhz Core GPU 1

2075 - 1055H/s
2100 - 1057H/s
2125 - 1059-1061H/s
2150 - 1060-1061H/s


First is Memory clock scaling, then Core clock scaling.

GPU 0/3 -> 570/580 Elpida.
GPU 1 -> 580 Samsung.
Seems weird, for 75Mhz Memclock increase, to get only about 10Hs. Usually, it should be around 30H/s for ~75Mhz increase (at least on XMR-Stak).


EDIT: For clarification, are HWs actually considered Bad shares, that failed CPU check (meaning, GPU worked on them, and before submitting, it failed) ?
jr. member
Activity: 98
Merit: 6
I have noticed something weird.

I have 4 cards.

1xRX570 4GB Elpida.
3xRX580 4GB, 2 of which Elpida.

The RX570 hashes absolutely the same as the RX580, despite the differences in the CUs.
Both cards are running SAME straps, SAME core clocks, SAME memory clocks.

It think it's because CryptoNight seems depending very much at memory, not so much at core performance. Like my RX 550 8 CU gets better hashrates than my RX 560 with 14/16 CUs. All about memory performance I think. I run all of my RX 550/560/570/580s at a very low core clock (1050-1150 MHz) and gets a good hashrate even then..
Jump to: