Topic: BFL board project COINTAMINATION - EU facility - ORDERS OPENED for Chili - page 41. (Read 73054 times)

Hello


Just wanted to give an update. I received this morning a DHL tracking number from our PCB fabrication facility in China. These will likely arrive in 2-3 days, at which time I will post pics on the forum and send a couple boards to form and lucko for testing and I will also send 5 boards to my US based Assembly facility, to have 5 production test boards manufactured.

If all goes well with those test boards (which we expect it will) we will move forward.

- Things left to do:
- Test professionally manufactured production board
- --> Finish web site for ordering and order tracking
- --> Test different cooling solutions
- --> Determine final prices

<-> We will have PCBs in hands in a couple of days
<-> We will have professionally manufactured production boards in 7-10 days (or earlier)

Ill post an update with pics as soon as the boards come in.

-JJ

OK after almost 3 hours of hashing with cgminer 3.4.0 , the 3 jalas seems to be better working

No it has less engines working since chips were poorly cooled. So it is physical damage... And since it has less engines running it is cooler...

This is a "shot in the dark" suggestion for your "Bad" Jala, It seems to be constant at a lower temperature but showing error rate 4.6%
Try removing the cover, the temp may drop a bit more, and then check the error rate.
The other riskier option is to restrict the airflow a "SMALL" amount, and see if the temperature rises to the same as the others, then again check the error rate.

I have found on some of my kit that overcooling made them as unstable as overheating.

Ok I got cgminer working, it was a driver issue, now everything looks great, I am getting an Avg. of 25.3 GH/s I will let it hash for few hours than post a screen shot

Ok so the great jala is even better then we thought... But there is something wrong with good Jala. It looks like it was not autoconfigure OK. You might think of restarting it... And the bad Jala is also a bit higher then we thought... Nice only good Jala is disappointing... It was probably a bad luck during configuration run...

EDIT:
If you don't know which is which...
Grate 0
Bad 1
Good 2

this is very exciting. keep moving forward, i am certain you can outpace bfl if you manage things well.

I am mining now at Eligius, I will give CPPSRB a try, so this is what I see now

Don't use PPS... Use PPLNS. You will also get NMC in this case... Any you have much lower fee(3%) that NMC will cover it... Earning are not that stable but long term it is batter... But if you are a gambler you could look at Bitparking with no fee and merged mining for 6 coins(but 2 of them are more or worthless and 2 are not on normal exchanges)...

EDIT: And that explain why I got same results on 50BTC(I opened account to test) I didn't look at logs but now that I did see the same thing...

I know why I lost that GH/s, I was looking to charts in my pool and it seems that the stratum server had some issues and still having them, it was disconnecting and trying to reconnect when I logged to watch, I have to change that pool now.

 I tried BTCguild and it was working steady at 25.6 GH/s, but fees are 7.5% I will look somewhere else.

I will call him tomorrow since he lost about 600 to 1000 MH just by taking them away from me... He must be doing something wrong... And it looks to be incompatibility with BFGminer. I just did some test. I get less with it too... Pool detects 300MH less on my Jala... I put it on difficulty 1 so I get more stable results... But will run it for longer...

And BTW 8.4 is normal for A grade chips without overclock. And 8,6 isn't that special too. But 8,7 that we did saw on one jala is... It had really good chips...

Boost from 5.ish to 8.ish just from software..... nice.

Compatibility and settings for fine tuning can be a nightmare.

For a while I used GUIMiner with my GPU 7970
One miner on the card gave a steady 520Mh/s
Two miners on the same card settled at 275Mh/s each but the card did run hotter.
I even ran with 8 miners on the CPU (8 core athlon) total of 7Mh/s.

BFGMiner only managed 490Mh/s on the GPU.

The best way to compare miners is to check the results on the mining pool, also give them time to settle into the pool, at least 12 hours, then swap them over and do it again.
Also try on another pool for the same period, the pools do have differences.

CG and BFG are constantly looking to improve their software, so by the time you read this, and or test the systems, there will be a new patch on one of them, and the balance will change again.

My Pi running CGMiner shows 6.378Gh/s and an error count of 1.02%.  The pool shows 6.124Gh/s thats a loss of 4%  (theoretical calculated hash rate 6.365Gh/s)
My PC running BFGMiner shows 9.633Gh/s and an error count of 0.95%.  The pool shows 9.577Gh/s thats a loss of 1% (theoretical calculated hash rate 9.715Gh/s)

In the US even the ISP provider causes differences in the results.

The end result is, you need lots of testing, and make comparisons, and never settle in one place, or you may get left behind.

You should see the chips in that Jala... And it has only 13 and 14 engines working. But still it worked with 8,2ish on my computer...

Also I know from experience with my Jala that BFGminer dosen't work as good as CGminer. But you need 3.3.1 or more. Any lower it will not detect device correctly and start. I guess it is compatibility issue... Or WinUSB drivers speed up things...

I'm mining with asic erupters at the moment. Also with BFGMiner 3.1.4
On my win8 x64 the USB drivers refused to install, SiLabs claim it is because asic erupters all have the same serial number and try to all connect to the first COM port.
However with my Win7 x86 I managed to install and run 55 USB ports and all 55 erupters work perfectly, the drivers even self installed. (SiLabs need to find a better excuse.)

Not all PC's and OS's are the same, If you can try on another PC the drivers may install correctly.

I also run with a Rasberry Pi. This is also quite happy with multiple devices, up to 19, but seems to fall over every time I add more than this.

Your BFL 1 device is showing a lower hash rate 7.55Gh/s, and rather high errors (4.9%)

OK I also have no idea but I guess last one... And it looks like a problem since we had 8.8 to 8.6, 8.7 to 8.6 and 8.3 to 8.2... That is too big of a difference. So you might have problem with some setting... You really need to figure out why you can't install WinUSB drivers with zadig... If anyone can help him with this...

EDIT: can you test on another computer...

I am a new user of bfgminer, I was using cgminer, So i do not know exactly what to look for, I will read the Readme file and let you know which one I have to look to

So we lost 0,6GH? Or is BFGminer showing hashrate modify by HW errors? CGminer was showing 25,6 to 25,0 the short time we run them...

And yes it was just a software... We might increase the voltage next time too... Samples are OK with 1,2V. But I would not push real chips that hi without test since they have more engines running and higher power usage...

And thanks for kind words...

as I know it was only the firmware that Lucko is working on, he said that he will have a newer version that we wish to test again. so yes that is it, no hardware modification .

Wooow, not too shabby at all ... 24.71 GH/s.
This was probably more than a software 'jailbreak' I guess.

     one4many

although I know who is Lucko in person, but I never knew that that guy is the same lucko in this forum till yesterday   , yesterday I PMed him to help me Upgraded my 3 Jalas, he kindly accepted and waited me for so long in his lab and did a Great job (for free ), the fact that it took him some time to do all that stuff, I am considering donating 2 days worth of mining to him, and I would like to ask people to help his project by donating what ever you can, awesome guy, smart and trustworthy.


I wish him best in this project  

edit: I forgot to say that we achieved 24.71 GH/s with only a HW-Avg. of 2.5%  , I had an Avg. of 16.5 GH/s with a HW.Avg. of 2%, so it is a win case of another overclocked Jala.