Pages:
Author

Topic: Re: [ANN] Technobit HEX16A2 - 24 Ghs - 50 EUR 48 h shippment - page 26. (Read 104895 times)

legendary
Activity: 1610
Merit: 1000

http://users.atw.hu/ifa/duplazas.png

Here you are!

SOMEBODY HELP!!

I have 3 boards and I manage by RPI and when they are running sometimes they compile the same block. There are lots of duplicate blocks in cgminer.  
How can i solve it? Can somebody help me? Any ideas?

Thanks:)

At least provided us with screenshot & your setting.

Your setting to high.
Best results 23-24 Gh/s (16 chips) in terms of stability and clocking are observed at 1100mV core voltage + 1500 MHz Clock
+1 on Zich,
besides:
you can always calc your pool reject rate % as A/R*100 in your case for hexc 1: 440/17088*100 = 2.57% - this depends on variance also

If all is good including your pool conectivity it should be below 0.5%

But once again az Zich said try it 1500/1100 you should be getting stable 23Gh+ per board Smiley
And check your psu with your setings you should be able to provide about (guess) 9A/12 per board. move your molexes on different lines. If they are on a single line voltage drop is huge + the chance for them to melt also
From a point on OC matters for your power bill only + increasing the chance for board failure Wink And decreasing your hashrate also. Life sucks sometimes Grin
legendary
Activity: 1190
Merit: 1000

http://users.atw.hu/ifa/duplazas.png

Here you are!

SOMEBODY HELP!!

I have 3 boards and I manage by RPI and when they are running sometimes they compile the same block. There are lots of duplicate blocks in cgminer.  
How can i solve it? Can somebody help me? Any ideas?

Thanks:)

At least provided us with screenshot & your setting.

Your setting too high.
Best results 23-24 Gh/s (16 chips) in terms of stability and clocking are observed at 1100mV core voltage + 1500 MHz Clock
newbie
Activity: 32
Merit: 0

http://users.atw.hu/ifa/duplazas.png

Here you are!

SOMEBODY HELP!!

I have 3 boards and I manage by RPI and when they are running sometimes they compile the same block. There are lots of duplicate blocks in cgminer. 
How can i solve it? Can somebody help me? Any ideas?

Thanks:)

At least provided us with screenshot & your setting.
legendary
Activity: 1610
Merit: 1000
SOMEBODY HELP!!

I have 3 boards and I manage by RPI and when they are running sometimes they compile the same block. There are lots of duplicate blocks in cgminer.  
How can i solve it? Can somebody help me? Any ideas?

Thanks:)
Some logs of your screen session pls?
PS: you are fast zich Smiley
legendary
Activity: 1190
Merit: 1000
SOMEBODY HELP!!

I have 3 boards and I manage by RPI and when they are running sometimes they compile the same block. There are lots of duplicate blocks in cgminer. 
How can i solve it? Can somebody help me? Any ideas?

Thanks:)

At least provided us with screenshot & your setting.
newbie
Activity: 32
Merit: 0
SOMEBODY HELP!!

I have 3 boards and I manage by RPI and when they are running sometimes they compile the same block. There are lots of duplicate blocks in cgminer. 
How can i solve it? Can somebody help me? Any ideas?

Thanks:)
legendary
Activity: 1610
Merit: 1000
Looks like DHL said my miners are coming tomorrow.  I am excite.  Grin
Nice,

Here is the hub i am using for nonstop operation of HEX boards + TpLink host

http://www.amazon.de/dp/B00602C91U/ref=pe_386171_38075861_TE_item - it is MTT hub and working great no issues or whatsoever

I am recommending highly this one and nothing else.

Happy mining:)


It doesn't need a powered hub right?

I got one hub that I still have lying around from my block erupter days.  Grin

I'm just worried I might not have enough molex power plugs. lol (Ordered some of Amazon. Might take a while)
Are you waiting for HEXA2 (Avalon 2) boards?
If yes according to marto post they will feed the hub with 5V. However if you gona use them with tp-link and you have it in hand i suggest to run lsusb when hub is connected and see if tplink supports that hub. It happen to me with some d-links that they work with PC but tp-link is not seeing them

Once again if you live in Germany and want nonstop operation just by the one I suggested. I can guarantee you you will not regret Wink



Yep. I hope it does (Worked fine on a Ras Pi I owned)

In America man. lol.  Yep. I have that tp-link router. Haven't flashed it yet though.

OK..
Then flash it and check if it sees the hub with lsub Wink
full member
Activity: 147
Merit: 100
Looks like DHL said my miners are coming tomorrow.  I am excite.  Grin
Nice,

Here is the hub i am using for nonstop operation of HEX boards + TpLink host

http://www.amazon.de/dp/B00602C91U/ref=pe_386171_38075861_TE_item - it is MTT hub and working great no issues or whatsoever

I am recommending highly this one and nothing else.

Happy mining:)


It doesn't need a powered hub right?

I got one hub that I still have lying around from my block erupter days.  Grin

I'm just worried I might not have enough molex power plugs. lol (Ordered some of Amazon. Might take a while)
Are you waiting for HEXA2 (Avalon 2) boards?
If yes according to marto post they will feed the hub with 5V. However if you gona use them with tp-link and you have it in hand i suggest to run lsusb when hub is connected and see if tplink supports that hub. It happen to me with some d-links that they work with PC but tp-link is not seeing them

Once again if you live in Germany and want nonstop operation just by the one I suggested. I can guarantee you you will not regret Wink



Yep. I hope it does (Worked fine on a Ras Pi I owned)

In America man. lol.  Yep. I have that tp-link router. Haven't flashed it yet though.
legendary
Activity: 1610
Merit: 1000

By the way i am trying new timing for nanos, because i think we are loosing some speed with latest 3.10
driver-hexmineru.c line 176
   if (long_sleep) {
         //applog(LOG_ERR, "Long sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
         // 6+ Gh/s Wink
          //cgsleep_us_r (&ts_start, 600*1000);
            cgsleep_ms(200);
       } else {   
         if(!job_switch) {
         //   applog(LOG_ERR, "Short sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
            //cgsleep_us_r (&ts_start, 5*1000);
              cgsleep_ms(5);
         } else {
            //applog(LOG_ERR, "No sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
         }
      }
    }

Cpu goes bit high again but i think nanos work better

If you have time you can play with   cgsleep_ms(200); - increase reduces cpu load but may impact nanos performance - do not go above 800
No need to decrease it


Yesterday i try to change  (&ts_start, 600*1000) to (&ts_start, 800*1000). Got WU around 70/m for 2 nano  Grin
Honestly speaking WU is a mistery for me. No joks it was explained by kano..You can dig the info for it. but changing it to 800 will reduce CPU load for sure. However it can reduce nanos hashrate. In theory up to 1300 should be fine without impact on the hashrate but according to my observations it is not quite true. In general you have to pump same job to the chip (and getting response from it) and watch when it will ask for new one. I do not know exactly speaking of timing when chip is asking for new one. But what i know that as fast you pump it the chance to delay the time for the new job is lower. Unfortunately pumping each 5 ms as it was in first version generates a lot of IO and CPU load but insures that you will never leave chip workless Grin. We need to found good compromise and sweet spot how much IO/CPU load to suffer.
Got it?
PS: I am not bitfury chip expert and if someone can comment on it i am all years

Got it. Then i will focus on (&ts_start, 800*1000) but change cgsleep_ms(200)
I have tried 800 and after a day i though that there were 2-3 % hash loss. But it might be variance though...
if it is OK at 800 speaking of hashrate leave it there Smiley

And i found that even tp-lin is at 90% load performace of HEX16 is not suffering - 5 HEX16 + 2 Nanos
So i left it pumping Grin
In curent implementation when you feed the chip with new job you wait longer then 5 ms 800 in your case, then you pump it each 5 ms with same job not to miss the moment. First big delay reduces cpu load ....
legendary
Activity: 1190
Merit: 1000

By the way i am trying new timing for nanos, because i think we are loosing some speed with latest 3.10
driver-hexmineru.c line 176
   if (long_sleep) {
         //applog(LOG_ERR, "Long sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
         // 6+ Gh/s Wink
          //cgsleep_us_r (&ts_start, 600*1000);
            cgsleep_ms(200);
       } else {   
         if(!job_switch) {
         //   applog(LOG_ERR, "Short sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
            //cgsleep_us_r (&ts_start, 5*1000);
              cgsleep_ms(5);
         } else {
            //applog(LOG_ERR, "No sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
         }
      }
    }

Cpu goes bit high again but i think nanos work better

If you have time you can play with   cgsleep_ms(200); - increase reduces cpu load but may impact nanos performance - do not go above 800
No need to decrease it


Yesterday i try to change  (&ts_start, 600*1000) to (&ts_start, 800*1000). Got WU around 70/m for 2 nano  Grin
Honestly speaking WU is a mistery for me. No joks it was explained by kano..You can dig the info for it. but changing it to 800 will reduce CPU load for sure. However it can reduce nanos hashrate. In theory up to 1300 should be fine without impact on the hashrate but according to my observations it is not quite true. In general you have to pump same job to the chip (and getting response from it) and watch when it will ask for new one. I do not know exactly speaking of timing when chip is asking for new one. But what i know that as fast you pump it the chance to delay the time for the new job is lower. Unfortunately pumping each 5 ms as it was in first version generates a lot of IO and CPU load but insures that you will never leave chip workless Grin. We need to found good compromise and sweet spot how much IO/CPU load to suffer.
Got it?
PS: I am not bitfury chip expert and if someone can comment on it i am all years

Got it. Then i will focus on (&ts_start, 800*1000) but change cgsleep_ms(200)
legendary
Activity: 1610
Merit: 1000

By the way i am trying new timing for nanos, because i think we are loosing some speed with latest 3.10
driver-hexmineru.c line 176
   if (long_sleep) {
         //applog(LOG_ERR, "Long sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
         // 6+ Gh/s Wink
          //cgsleep_us_r (&ts_start, 600*1000);
            cgsleep_ms(200);
       } else {   
         if(!job_switch) {
         //   applog(LOG_ERR, "Short sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
            //cgsleep_us_r (&ts_start, 5*1000);
              cgsleep_ms(5);
         } else {
            //applog(LOG_ERR, "No sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
         }
      }
    }

Cpu goes bit high again but i think nanos work better

If you have time you can play with   cgsleep_ms(200); - increase reduces cpu load but may impact nanos performance - do not go above 800
No need to decrease it


Yesterday i try to change  (&ts_start, 600*1000) to (&ts_start, 800*1000). Got WU around 70/m for 2 nano  Grin
Honestly speaking WU is a mistery for me. No joks it was explained by kano..You can dig the info for it. but changing it to 800 will reduce CPU load for sure. However it can reduce nanos hashrate. In theory up to 1300 should be fine without impact on the hashrate but according to my observations it is not quite true. In general you have to pump same job to the chip (and getting response from it) and watch when it will ask for new one. I do not know exactly speaking of timing when chip is asking for new one. But what i know that as fast you pump it the chance to delay the time for the new job is lower. Unfortunately pumping each 5 ms as it was in first version generates a lot of IO and CPU load but insures that you will never leave chip workless Grin. We need to found good compromise and sweet spot how much IO/CPU load to suffer.
Got it?
PS: I am not bitfury chip expert and if someone can comment on it i am all years
legendary
Activity: 1190
Merit: 1000

By the way i am trying new timing for nanos, because i think we are loosing some speed with latest 3.10
driver-hexmineru.c line 176
   if (long_sleep) {
         //applog(LOG_ERR, "Long sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
         // 6+ Gh/s Wink
          //cgsleep_us_r (&ts_start, 600*1000);
            cgsleep_ms(200);
       } else {   
         if(!job_switch) {
         //   applog(LOG_ERR, "Short sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
            //cgsleep_us_r (&ts_start, 5*1000);
              cgsleep_ms(5);
         } else {
            //applog(LOG_ERR, "No sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
         }
      }
    }

Cpu goes bit high again but i think nanos work better

If you have time you can play with   cgsleep_ms(200); - increase reduces cpu load but may impact nanos performance - do not go above 800
No need to decrease it


Yesterday i try to change  (&ts_start, 600*1000) to (&ts_start, 800*1000). Got WU around 70/m for 2 nano  Grin
legendary
Activity: 1610
Merit: 1000
Looks like DHL said my miners are coming tomorrow.  I am excite.  Grin
Nice,

Here is the hub i am using for nonstop operation of HEX boards + TpLink host

http://www.amazon.de/dp/B00602C91U/ref=pe_386171_38075861_TE_item - it is MTT hub and working great no issues or whatsoever

I am recommending highly this one and nothing else.

Happy mining:)


It doesn't need a powered hub right?

I got one hub that I still have lying around from my block erupter days.  Grin

I'm just worried I might not have enough molex power plugs. lol (Ordered some of Amazon. Might take a while)
Are you waiting for HEXA2 (Avalon 2) boards?
If yes according to marto post they will feed the hub with 5V. However if you gona use them with tp-link and you have it in hand i suggest to run lsusb when hub is connected and see if tplink supports that hub. It happen to me with some d-links that they work with PC but tp-link is not seeing them

Once again if you live in Germany and want nonstop operation just by the one I suggested. I can guarantee you you will not regret Wink
full member
Activity: 147
Merit: 100
Looks like DHL said my miners are coming tomorrow.  I am excite.  Grin
Nice,

Here is the hub i am using for nonstop operation of HEX boards + TpLink host

http://www.amazon.de/dp/B00602C91U/ref=pe_386171_38075861_TE_item - it is MTT hub and working great no issues or whatsoever

I am recommending highly this one and nothing else.

Happy mining:)


It doesn't need a powered hub right?

I got one hub that I still have lying around from my block erupter days.  Grin

I'm just worried I might not have enough molex power plugs. lol (Ordered some of Amazon. Might take a while)
legendary
Activity: 1190
Merit: 1000
I just open my stupid hub. I found that it used 2 x FE1.1s chip LOL
Did you ever checkout the chip in that 10 port hub loshia? one must be FE2.1. But what about other one?
Yes i did and i posted it couple of times already Smiley

root@HEX16A:~# lsusb
Bus 001 Device 002: ID 1a40:0201 Terminus Technology Inc. FE 2.1 7-port Hub
Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
Bus 001 Device 003: ID 1a40:0101 Terminus Technology Inc. 4-Port HUB



I have not open it and i am not sure if Terminus Technology Inc. 4-Port HUB is tt chip but i am using 10 ports without issues Smiley I have only two tp-links with 10 ports (Actually 9 HEX16B + power for the tplink + data) otherwise my Hex16A are not more than 7 (850W PSu) because of the power  Cry
I guess with 7 boards i am using only one (or maybe 0) port of 4 port chip
Actually tp-link can be hacked to get power from usb data port but i do not have time for this. When i need it (not lose USB port on my hub) i just strip usb power wire spike + crimp/ or solder + tape  and so on you now the story Grin
PS:

In the begging i used to cut USB data red wire to avoid feeding from the hub but as long as i am using single PSU ground i am not doing it anymore

I see, so it's give chip info there  Cheesy
legendary
Activity: 1610
Merit: 1000
Looks like DHL said my miners are coming tomorrow.  I am excite.  Grin
Nice,

Here is the hub i am using for nonstop operation of HEX boards + TpLink host

http://www.amazon.de/dp/B00602C91U/ref=pe_386171_38075861_TE_item - it is MTT hub and working great no issues or whatsoever

I am recommending highly this one and nothing else.

Happy mining:)


I just open my stupid hub. I found that it used 2 x FE1.1s chip LOL
Did you ever checkout the chip in that 10 port hub loshia? one must be FE2.1. But what about other one?
Yes i did and i posted it couple of times already Smiley

root@HEX16A:~# lsusb
Bus 001 Device 002: ID 1a40:0201 Terminus Technology Inc. FE 2.1 7-port Hub
Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
Bus 001 Device 003: ID 1a40:0101 Terminus Technology Inc. 4-Port HUB



I have not open it and i am not sure if Terminus Technology Inc. 4-Port HUB is tt chip but i am using 10 ports without issues Smiley I have only two tp-links with 10 ports (Actually 9 HEX16B + power for the tplink + data) otherwise my Hex16A are not more than 7 (850W PSu) because of the power  Cry
I guess with 7 boards i am using only one (or maybe 0) port of 4 port chip
Actually tp-link can be hacked to get power from usb data port but i do not have time for this. When i need it (not lose USB port on my hub) i just strip usb power wire spike + crimp/ or solder + tape  and so on you now the story Grin
PS:

In the begging i used to cut USB data red wire to avoid feeding from the hub but as long as i am using single PSU ground i am not doing it anymore

PS:
By the way i am trying new timing for nanos, because i think we are loosing some speed with latest 3.10
driver-hexmineru.c line 176
   if (long_sleep) {
         //applog(LOG_ERR, "Long sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
         // 6+ Gh/s Wink
          //cgsleep_us_r (&ts_start, 600*1000);
            cgsleep_ms(200);
       } else {   
         if(!job_switch) {
         //   applog(LOG_ERR, "Short sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
            //cgsleep_us_r (&ts_start, 5*1000);
              cgsleep_ms(5);
         } else {
            //applog(LOG_ERR, "No sleep %s%d",hexmineru->drv->name, hexmineru->device_id);
         }
      }
    }

Cpu goes bit high again but i think nanos work better

If you have time you can play with   cgsleep_ms(200); - increase reduces cpu load but may impact nanos performance - do not go above 800
No need to decrease it

legendary
Activity: 1190
Merit: 1000
Looks like DHL said my miners are coming tomorrow.  I am excite.  Grin
Nice,

Here is the hub i am using for nonstop operation of HEX boards + TpLink host

http://www.amazon.de/dp/B00602C91U/ref=pe_386171_38075861_TE_item - it is MTT hub and working great no issues or whatsoever

I am recommending highly this one and nothing else.

Happy mining:)


I just open my stupid hub. I found that it used 2 x FE1.1s chip LOL
Did you ever checkout the chip in that 10 port hub loshia? one must be FE2.1. But what about other one?
legendary
Activity: 1610
Merit: 1000
Looks like DHL said my miners are coming tomorrow.  I am excite.  Grin
Nice,

Here is the hub i am using for nonstop operation of HEX boards + TpLink host

http://www.amazon.de/dp/B00602C91U/ref=pe_386171_38075861_TE_item - it is MTT hub and working great no issues or whatsoever

I am recommending highly this one and nothing else.

Happy mining:)
full member
Activity: 147
Merit: 100
Looks like DHL said my miners are coming tomorrow.  I am excite.  Grin
legendary
Activity: 1190
Merit: 1000
..............
Does those nonces = 0 mean they are not working or damaged?
Maybe you can help Marto
Thanks!
Smells like dead chips to me

Yeah sound like dead chip. Or you can try to visual check if there is any missing component or unclean solder.
I mark the chip base on component numbering in schematic but you can make sure by measuring the chip temperature or by touching the chip. Working chip will be very hot  Cheesy


If all clean then you need to send back to marto.
Pages:
Jump to: