BTCMiner - Open Source Bitcoin Miner for ZTEX FPGA Boards, 215 MH/s on LX150

[Turbor]

Will there be an option to change the frequency manually in the future ? After 48h or so i'm down to 180 to 190 MH/s but that's at 184 MHz. Would be nice to be able to play around with the frequency  

[sadpandatech]

Will there be an option to change the frequency manually in the future ? After 48h or so i'm down to 180 to 190 MH/s but that's at 184 MHz. Would be nice to be able to play around with the frequency  

  if I understand his software correctly, it automaticly adjusts the freq up/down based on the error rate. With that in mind if we 'force' the unit to 200MHz and it is getting a theoretical 10% error rate the hash rate would still be 180MHs.  It would be nice to have the option to manually adjust the error tolerance though to set the optimum limit of error/freq ourselves.

  Have you done anything to try and further cool the chip to see if that helps?

[Turbor]

if I understand his software correctly, it automaticly adjusts the freq up/down based on the error rate. With that in mind if we 'force' the unit to 200MHz and it is getting a theoretical 10% error rate the hash rate would still be 180MHs.  It would be nice to have the option to manually adjust the error tolerance though to set the optimum limit of error/freq ourselves.

  Have you done anything to try and further cool the chip to see if that helps?

I have an extra 92mm fan blowing at the board. Your right about self adjustment. It changes to 192 MHz on startup but then always goes back to 184 MHz. At 184 the error rate is below 0.3 %.

[trilby]

if I understand his software correctly, it automaticly adjusts the freq up/down based on the error rate. With that in mind if we 'force' the unit to 200MHz and it is getting a theoretical 10% error rate the hash rate would still be 180MHs.  It would be nice to have the option to manually adjust the error tolerance though to set the optimum limit of error/freq ourselves.

  Have you done anything to try and further cool the chip to see if that helps?

I have an extra 92mm fan blowing at the board. Your right about self adjustment. It changes to 192 MHz on startup but then always goes back to 184 MHz. At 184 the error rate is below 0.3 %.

my one is different it starts at 200 then it changes to 192 and holds stable

[sadpandatech]

if I understand his software correctly, it automaticly adjusts the freq up/down based on the error rate. With that in mind if we 'force' the unit to 200MHz and it is getting a theoretical 10% error rate the hash rate would still be 180MHs.  It would be nice to have the option to manually adjust the error tolerance though to set the optimum limit of error/freq ourselves.

  Have you done anything to try and further cool the chip to see if that helps?

I have an extra 92mm fan blowing at the board. Your right about self adjustment. It changes to 192 MHz on startup but then always goes back to 184 MHz. At 184 the error rate is below 0.3 %.

my one is different it starts at 200 then it changes to 192 and holds stable

  Your's sounds more what I would expect from the latest production he has out.  I am wondering, do you guys know what voltage your units are? I.E. what resistance is on R12/R13?

  I'm just tossing things out, since I unfortuantly do not have one of these units to test. But I am wondering if it could be a thermal compound issue. Do we know what compound was used to set the heatsinks and what its cure time is?

[Dexter770221]

In GPU/CPU world each piece overclocks differently. You may have a worse piece. I've wonder what hashrate could be possible with peltier cooling I have two modules (50W & 70W) from an old days when they cooled my celeron 333@560. Temp was -18C...

[Turbor]

Can this be a thermal thing ? I placed the heatsink myself. Should be perfect. I'm happy so far with the board.

[ztex]

I have an extra 92mm fan blowing at the board. Your right about self adjustment. It changes to 192 MHz on startup but then always goes back to 184 MHz. At 184 the error rate is below 0.3 %.

The minimum guaranteed hash rate is 176 MH/s. But in fact I never saw a 1.15x board which is as slow as yours.

If your board can only do 184 MHz at an error rate of 0.3% it will hold the 192 MHz only for a few seconds. If your board holds it longer you have a cooling problem.

First check the heat sink. If it feels hot their is not enough airflow. Then check the temperature of the PCB below the FPGA. If it feels warm there is an air gap between heat sink and FPGA. More thermal grease should solve this problem

[Turbor]

I went with the thermal pad that came with the heatsink. The heatsink gets a bit warm but not hot. There is no problem for me with 180 mh/s 24h. Just interesting to see that others run other frequencies. I have almost zero error rate. Will the device only clock (up) at startup ?

[ztex]

If I understand his software correctly, it automaticly adjusts the freq up/down based on the error rate.

 With that in mind if we 'force' the unit to 200MHz and it is getting a theoretical 10% error rate the hash rate would still be 180MHs.  It would be nice to have the option to manually adjust the error tolerance though to set the optimum limit of error/freq ourselves.

The software automatically chooses the frequency which delivers the highest effective hash rate.  This calculation is based on the maximum error rate (maxErroRate) in the output.

A better optimization goal would be maximize the income based on error rates and considering the energy costs. But we are talking about average improvements of fee per mill ...

[ztex]

Your's sounds more what I would expect from the latest production he has out.  I am wondering, do you guys know what voltage your units are? I.E. what resistance is on R12/R13?

All boards that were delivered out have a core voltage of 1.23V.  The first lot with 1.25V (2% to 3% faster) will be produced end of December.

[ztex]

I went with the thermal pad that came with the heatsink.

I never tried out this thing. Check the temperature of the (bottom side) of the PCB below the FPGA. If it feels quite warm after a runtime of at least 30min
the heat conductivity of the pad is to bad.

There is no problem for me with 180 mh/s 24h. Just interesting to see that others run other frequencies. I have almost zero error rate. Will the device only clock (up) at startup ?

The boards are clocked up at start-up until the effective hash rate decreases and then are clocked down by one step. Since the maximum error rate is used for frequency optimization the boards usually hold this frequency. BTW, nextMaxErrorRate denotes the maximum error rate of the next frequency step.

If your board holds the 192 MHz for more the a few seconds and then clocks down to 184 MHz at an error rate of 0.3% this indicates a cooling problem.

[Turbor]

Tnx for the information. I'll check this. But as long as it runs at 184 with no error there is nothing to fear right ?

Edit: Here is a screen when i start the miner. Why does it set the initial frequency to 184 MHz in my case and 200 MHz for someone else ? The board was running more than a week when i took that pic. Not that i want to complain ! I'm just trying to understand  ...

[ztex]

Edit: Here is a screen when i start the miner. Why does it set the initial frequency to 184 MHz in my case and 200 MHz for someone else ? The board was running more than a week when i took that pic. Not that i want to complain ! I'm just trying to understand  ...

The software always starts with 184 Mz.

Your board seems to hold 192 MHz for about 20s. Since the error rate increases very quickly at maximum frequency it should do 184MHz with (almost) no errors (<0.1%).

If the error rate increases over time (0.3% as you wrote in a previous possting) this indicates a cooling problem.

[Turbor]

Thanks for clarification. It is as you said. The thermal pad is sh!t !!!  . Please tell your customers not to use it. I put some termal grease on and now it works at 192 MHz with max 0.68 % errorRate so far.

F.cked up one heatsink holder but i was able to replace it with a M3 screw. I could not be happier. Great product !

Edit: Here is a screen when i start the miner. Why does it set the initial frequency to 184 MHz in my case and 200 MHz for someone else ? The board was running more than a week when i took that pic. Not that i want to complain ! I'm just trying to understand  ...

The software always starts with 184 Mz.

Your board seems to hold 192 MHz for about 20s. Since the error rate increases very quickly at maximum frequency it should do 184MHz with (almost) no errors (<0.1%).

If the error rate increases over time (0.3% as you wrote in a previous possting) this indicates a cooling problem.

[DeathAndTaxes]

Thanks for clarification. It is as you said. The thermal pad is sh!t !!!  . Please tell your customers not to use it. I put some termal grease on and now it works at 192 MHz with max 0.68 % errorRate so far.

Good to know.  What brand thermal grease did you use?

[Turbor]

Good to know.  What brand thermal grease did you use?

I have a spare Scythe Katana 3 CPU cooler. First the idea was to replace the heatsink with the Katana but that would have been a massive overkill in almost every way. The grease (Scythe) seems to work fine. I never did that before and was a bit afraid to f.ck up the board but it's really a nice piece of hardware. The stock heatsink pad is just too thick imo.

[ztex]

Thanks for clarification. It is as you said. The thermal pad is sh!t !!!  . Please tell your customers not to use it. I put some termal grease on and now it works at 192 MHz with max 0.68 % errorRate so far.

Good to know.  What brand thermal grease did you use?

Remember it is not a 140W CPU. Any thermal grease can be used. Thermal grease is included in the heat sink kit delivered with the board.

I dissected the pad today. It is some kind of textile with sticky grey stuff and it quite thick. IMHO it is not usable for anything.

[DeathAndTaxes]

Remember it is not a 140W CPU. Any thermal grease can be used. Thermal grease is included in the heat sink kit delivered with the board.

Good point.
Just wondered because the cooling improved so much.  Guess that was more a case of the pad sucking then the grease being super good.

[Turbor]

Guess that was more a case of the pad sucking then the grease being super good.

Exactly