Avalon ASIC users thread

[Sitarow]

Has anyone tried 375mhz air cooled?

I've been running at over 24 hours stable at 350mhz (105Gh-120Gh), and I'm getting that itch to try and bump it up a little more. I'm not sure if Avalon changed something to make it more stable overclocked or not and wondering you batch 1/2 guys experience with pushing over 350mhz.

I've been considering the best way to determine a sweet spot for my avalon...  currently running at 343 (it just "feels" like a nice balance between hashrate and HW errors), but now I'm getting that overclocking itch once again.

I'm trying to devise a method that is slightly more scientific than my gut feel. Perhaps it could even be integrated into some kind of algorithm that could be added to the firmware to improve the auto speed setting? I don't know if that's even possible. Also I'm not really sure what a lot of the numbers mean in the web interface cgminer status tab, so I'm just guessing at a lot of this. Corrections to my assumptions/mistakes would be appreciated.

So, I'm just guessing at what all those numbers really mean, but here's what I'm thinking.

if:
HW = Hardware errors
Diff1Shares = equivalent number of shares if submitting at difficulty 1 (not sure whether to use this or DiffA ? )
MHSav = MH/s average

then:
EHR = Effective Hash Rate = MHSav - ( MHSav * ( HW / Diff1Shares ) )

So, what I'm gonna do is just keep increasing or decreasing my Frequency until I can find maximum EHR. Makes sense?

DiffA is what work you get paid for as in accepted shares.

[1714029464]

1714029464

[1714029464]

1714029464

[1714029464]

1714029464

[1714029464]

1714029464

[1714029464]

1714029464

[fracta15]

How would one go about buying one of these bad boys?

[shmadz]

Has anyone tried 375mhz air cooled?

I've been running at over 24 hours stable at 350mhz (105Gh-120Gh), and I'm getting that itch to try and bump it up a little more. I'm not sure if Avalon changed something to make it more stable overclocked or not and wondering you batch 1/2 guys experience with pushing over 350mhz.

I've been considering the best way to determine a sweet spot for my avalon...  currently running at 343 (it just "feels" like a nice balance between hashrate and HW errors), but now I'm getting that overclocking itch once again.

I'm trying to devise a method that is slightly more scientific than my gut feel. Perhaps it could even be integrated into some kind of algorithm that could be added to the firmware to improve the auto speed setting? I don't know if that's even possible. Also I'm not really sure what a lot of the numbers mean in the web interface cgminer status tab, so I'm just guessing at a lot of this. Corrections to my assumptions/mistakes would be appreciated.

So, I'm just guessing at what all those numbers really mean, but here's what I'm thinking.

if:
HW = Hardware errors
Diff1Shares = equivalent number of shares if submitting at difficulty 1 (not sure whether to use this or DiffA ? )
MHSav = MH/s average

then:
EHR = Effective Hash Rate = MHSav - ( MHSav * ( HW / Diff1Shares ) )

So, what I'm gonna do is just keep increasing or decreasing my Frequency until I can find maximum EHR. Makes sense?

DiffA is what work you get paid for as in accepted shares.

OK, what I'm unsure about is whether it is more accurate to use accepted shares or total submitted shares...

my thinking is that total shares would remove the variance of the latency or other problems with the pool, so they might be more accurate? But I am unsure of how the total shares are calculated... perhaps they include some types of hardware related problems and that is what causes them to not be accepted?

Still unsure on this, maybe I will run numbers on both, maybe the difference is trivial in the end?

[shmadz]

How would one go about buying one of these bad boys?

Simply jump in your Delorian, dial the flux capacitor to 1.21 Gigawatts, set your target to Feb 18, 2013 and floor it.

[bitdude]

How would one go about buying one of these bad boys?

Simply jump in your Delorian, dial the flux capacitor to 1.21 Gigawatts, set your target to Feb 18, 2013 and floor it.

or check Bitcoin Forum > Economy > Marketplace > Auctions (https://bitcointalk.org/index.php?board=73.0)

[organofcorti]

How would one go about buying one of these bad boys?

Simply jump in your Delorian, dial the flux capacitor to 1.21 Gigawatts, set your target to Feb 18, 2013 and floor it.

or check Bitcoin Forum > Economy > Marketplace > Auctions (https://bitcointalk.org/index.php?board=73.0)

I can get a Delorian there?

[lastbit]

Read this:

https://en.bitcoin.it/wiki/Avalon#About_.5Busb_1-1:_clear_tt_1_.288030.29_error_-71.5D

Thanks mitac64.
Wireless was disabled (I'm using ethernet), but after your post I have checked and noticed I have a lot of WiFi related modules running. I've rmmoded them and now I'll make another try @325.

16h uptime @325 with wireless properly disabled. No need to replace TP-link it seems.
Thanks mitac64 for reminding me to read more carefully!

[xiangfu]

Hi

This is a late firmware release message , as you may know there is a new case design in batch3 machines. new case design come with new temperature sensor, this improvement make the temperature more close to the real chip temperature.

This new machine have to use this new firmware 20130723, please read the ChangeLog below for more information
  https://en.bitcoin.it/wiki/Avalon#20130723

This release works just fine on old type. if you are in summer, I would advise setup:
  Target Temperature to 50℃ and Overheat Cut Off Temperature to 70℃ (this is my configure)

Also we have more overclock options in this release. just pay more attention on heatsink and PSU when you want do static overclock.

BTW:
  Some people report that some fpga controller not working. you may want read this: https://en.bitcoin.it/wiki/Avalon#Issues
  the QFN re-work is not easy, you have to find the professional people doing this. (the ft232 have to clean before soldering again)

Again, please donate to cgminer developers.
  https://en.bitcoin.it/wiki/Avalon#Donation

Xiangfu

[el_rlee]

Any ideas on how to troubleshoot this?

Not.

[rudrigorc2]

Does anyone know a way to tell if I have any bad chips on my modules?

Two of my Avalons run at 82Gh/s at 350, and the third maintains only 79Gh/s after several days, so variance is not the issue.  Thinking it was a bad PSU, I upgraded the power supply on the slow one to an OCZ ZX1250.  I alsodisabled wifi on all of them, and have 0703 firmware on all three. I set the fans to run 100% and temps are low to mid 40s.  I can't figure out why one is 3Gh/s slower.  Any ideas on how to troubleshoot this?

If you go to cgminer API log you can see the performance for each one of the 24 miners(with 10 chips)
this might give you a hint.

[pikeadz]

Any ideas on how to troubleshoot this?

Not.

Can you elaborate?  Are you saying don't bother troubleshooting because it's normal?  Or something else?

[BenTuras]

Does anyone know a way to tell if I have any bad chips on my modules?

Have a look at your cgminer API log for out of average values for match_work_count, like this:

   [match_work_count2] => 242475
   [match_work_count3] => 14287
   [match_work_count4] => 95
   [match_work_count5] => 242116
   [match_work_count6] => 242801

Next challenge will be to figure out which module that is, get a replacement and install it.

Or just leave it as it is and accept the minor difference!

[BenTuras]

Does anyone know a way to tell if I have any bad chips on my modules?

Have a look at your cgminer API log for out of average values for match_work_count, like this:

   [match_work_count2] => 242475
   [match_work_count3] => 14287
   [match_work_count4] => 95
   [match_work_count5] => 242116
   [match_work_count6] => 242801

Next challenge will be to figure out which module that is, get a replacement and install it.

Or just leave it as it is and accept the minor difference!

The lowest number is within 3% of the highest.  Nothing crazy like what you have shown above.  I am leaning towards calling this worker "theslowlearner" in my mining pool, and just accepting it.  But i did want to see if it was something obvious I was missing.  thanks!

Yep, saw your numbers after I posted my message. It's clear that I have 2 of the 24 boards out of wack. I have accepted that this machine is performing a little less than average(~8%).

One day I will replace the 2 boards, when the availability and price are 'normal'.

[el_rlee]

Any ideas on how to troubleshoot this?

Not.

Can you elaborate?  Are you saying don't bother troubleshooting because it's normal?  Or something else?

Imho it's normal and you have little chance of changing anything. Don't give too much about the numbers, this is for a 2-module Avalon:

  [match_work_count1] => 93877
   [match_work_count2] => 93617
   [match_work_count3] => 94072
   [match_work_count4] => 93691
   [match_work_count5] => 93578
   [match_work_count6] => 93646
   [match_work_count7] => 93869
   [match_work_count8] => 93004
   [match_work_count9] => 93233
   [match_work_count10] => 93724
   [match_work_count11] => 93743
   [match_work_count12] => 93655
   [match_work_count13] => 94141
   [match_work_count14] => 94080
   [match_work_count15] => 93962
   [match_work_count16] => 93753
   [match_work_count17] => 109
   [match_work_count18] => 141
   [match_work_count19] => 151
   [match_work_count20] => 107
   [match_work_count21] => 133
   [match_work_count22] => 109
   [match_work_count23] => 132
   [match_work_count24] => 162

means miners 17 to 24 are not even connected...

Also how would you manage to exchange modules on a board? Cut it?

[pikeadz]

Imho it's normal and you have little chance of changing anything. Don't give too much about the numbers....

Thanks.  I guess I will leave it alone then. 

The only other thing I am noticing is that the slower one produces 5% HW errors, and the other two are around 2%.  But even using avalon_auto doesn't change that, nor does running it at a lower frequency. 

[Bogart]

Imho it's normal and you have little chance of changing anything. Don't give too much about the numbers....

Thanks.  I guess I will leave it alone then.  

The only other thing I am noticing is that the slower one produces 5% HW errors, and the other two are around 2%.  But even using avalon_auto doesn't change that, nor does running it at a lower frequency.  

You could swap the modules around to isolate which is the slow one(s).

[Ytterbium]

Thanks.  I guess I will leave it alone then. 

The only other thing I am noticing is that the slower one produces 5% HW errors, and the other two are around 2%.  But even using avalon_auto doesn't change that, nor does running it at a lower frequency. 

I guess you could theoretically try buying raw chips and trying to de-solder and re-solder them yourself.

[RobbieMor]

I received my Batch 3 unit a few days ago and was hoping someone might be able to recommend someone to help me set it up. I've been mining with a BFGFPGA miner for some time and also the new Asic USB sticks so I do have a little experience. When I got the unit I discover that I need Unix so I installed Ubantu on one of my computers but am not up to speed on using it yet. I was a software test engineer until I retired so I do know a little about computers and software and I learn fast. My first computer was the Sol-20 by Processor Technology. If you haven't heard of it do a google search.

 

I'm willing to pay for help. I'm a disabled Vietnam Veteran and want to get my unit up and mining. Phone help would be a definite plus.

[ProfMac]

I received my Batch 3 unit a few days ago and was hoping someone might be able to recommend someone to help me set it up. I've been mining with a BFGFPGA miner for some time and also the new Asic USB sticks so I do have a little experience. When I got the unit I discover that I need Unix so I installed Ubantu on one of my computers but am not up to speed on using it yet. I was a software test engineer until I retired so I do know a little about computers and software and I learn fast. My first computer was the Sol-20 by Processor Technology. If you haven't heard of it do a google search.

 

I'm willing to pay for help. I'm a disabled Vietnam Veteran and want to get my unit up and mining. Phone help would be a definite plus.

I can try to walk anyone through the setup.  For pay, I would want all the mining output that the unit produces between now and Noon CDT on Monday, or about 3 days from the time I start.  If I set it up soon, I get more.  If I don't get it working soon enough, I don't get anything.

The risks are:
1.  Your unit is defective and it is not possible for me to succeed.
2.  You become impatient and do things that I have not requested, and I am no longer working on the configuration that I think I am.
3.  You don't have a proper router available.

Do you or a helper have the ability to move the unit, connect and disconnect cables quickly, and the like?


Oh, and I don't believe that you need to use Unix.  Any laptop or other computer that will connect to the internet can be used to do the configuration.  If you are using Windows, you do have to be logged in as an administrator.

[gyverlb]

I received my Batch 3 unit a few days ago and was hoping someone might be able to recommend someone to help me set it up. I've been mining with a BFGFPGA miner for some time and also the new Asic USB sticks so I do have a little experience. When I got the unit I discover that I need Unix so I installed Ubantu on one of my computers but am not up to speed on using it yet. I was a software test engineer until I retired so I do know a little about computers and software and I learn fast. My first computer was the Sol-20 by Processor Technology. If you haven't heard of it do a google search.

 

I'm willing to pay for help. I'm a disabled Vietnam Veteran and want to get my unit up and mining. Phone help would be a definite plus.

I'm surprised you found it difficult to configure. The wiki currently has difficulties for me but I could get it from Google cache (just google "Avalon setup", should be in the first entries).

The only thing you need is any system capable of web browsing you can setup on a specific IP address (192.168.0.101). Then you only have to use a navigator to connect to http://192.168.0.100.

Edit: you need Ethernet too. It may help us if you described your setup and what you tried.