Bitburner Fury - Hashrate Protection

[Roy Badami]

BTW, someone42, I have a problem whereby if cgminer idles the board due to reaching avalon-cutoff temperature, something about the way cgminer idles the board causes it to lose communcation with the board until cgminer is restarted.  Easily reproducible in my setup by setting avalon-cutoff to a low value.....   But I'm not running latest firmware, so don't worry too much...  Will attempt to repro again after I've flashed, and report back.

roy

[zulover]

The guide on page 13 or 14 by rage was helpful, I followed it had some problwms not recognising but ran zadig a few times to get it right. You should see bitburner in devices in control panel. after that make the script in notepad to run as a .bat file at it worked. Try powering board off and on also . I was having trouble with a bfl single which was connected also but it cleared after running the zadig utility.


Just noticed roadstress your a hero member, you shd be telling me this shouldnt you?

[Roy Badami]

For those who are getting "usb_write" errors and are unable to power cycle the boards, here is some firmware:
https://www.dropbox.com/s/x9zhnm86uuzcn05/bitburner-fury-firmware_1.2.6.hex (BTB identifier, for cgminer < 3.6.0)
https://www.dropbox.com/s/l87xnp5mdfg6f8y/bitburner-fury-firmware_1.2.6_BBF.hex (BBF identifier, for cgminer >= 3.6.0)

This adds a soft watchdog which will auto-reset the microcontroller if the main thread blocks. This doesn't fix the cause of the usb_write errors (I'm still working on that), but it means that if they do occur, your boards won't sit there doing nothing.

Just a reminder to everyone - because I know lots of people don't have time to read every post in this thread:

The BBF firmware - recomended (by me, at least, if not necessarily by anyone else LOL) means that if you are overvolting your Bitburner Fury boards you need to use --bitburner-fury-voltage rather than --bitburner-voltage

BUT: Be aware, the numbers you are used to putting in for --bitburner-voltage are not true voltages in the case of Fury boards; they are Avalon-compatible voltages that are 300mV higher than what the boards really run at.  (Bitfury chips run at 900mV compared to Avalon chips which run at 1200mV).

So if you're used to saying, for example "--bitburner-voltage 1350" to overvolt your Bitburner Fury boards with the old firmware then with BBF firmware you will need to use "--bitburner-fury-voltage 1050" to achieve the same effect - both cause the board to run at 1050mV.  Fortunately I think cgminer will reject "--bitburner-fury-voltage 1350", and even if it didn't the boards are incapable of complying - which is just as well, because if you really put 1350mV through the Bitfury chips that would pretty much certainly fry them.

But still, be careful with your voltages!

roy

[RoadStress]

The guide on page 13 or 14 by rage was helpful, I followed it had some problwms not recognising but ran zadig a few times to get it right. You should see bitburner in devices in control panel. after that make the script in notepad to run as a .bat file at it worked. Try powering board off and on also . I was having trouble with a bfl single which was connected also but it cleared after running the zadig utility.


Just noticed roadstress your a hero member, you shd be telling me this shouldnt you?

It's hard to keep track of everything, but thank you. I just read pages 13 and 14 and helped a bit. Will post an update

[RoadStress]

BUT: Be aware, the numbers you are used to putting in for --bitburner-voltage are not true voltages; they are Avalon-compatible voltages that are 300mV higher than what the boards really run at.  (Bitfury chips run at 900mV compared to Avalon chips which run at 1200mV).

So if you're used to saying, for example "--bitburner-voltage 1350" to overvolt your Bitburner Fury boards with the old firmware then with BBF firmware you will need to use "--bitburner-fury-voltage 1050" to achieve the same effect - both cause the board to run at 1050mV.  Fortunately I think cgminer will reject "--bitburner-fury-voltage 1350", and even if it didn't the boards are incapable of complying - which is just as well, because if you really put 1350mV through the Bitfury chips that would pretty much certainly fry them.

But still, be careful with your voltages!

roy

Finally some good information. I used 1300mV, but in cgminer i see ~1060mV. Shouldn't it be 1000mV?

Edit: what is the max safe temp? 55? 60?

[Roy Badami]

BUT: Be aware, the numbers you are used to putting in for --bitburner-voltage are not true voltages; they are Avalon-compatible voltages that are 300mV higher than what the boards really run at.  (Bitfury chips run at 900mV compared to Avalon chips which run at 1200mV).

So if you're used to saying, for example "--bitburner-voltage 1350" to overvolt your Bitburner Fury boards with the old firmware then with BBF firmware you will need to use "--bitburner-fury-voltage 1050" to achieve the same effect - both cause the board to run at 1050mV.  Fortunately I think cgminer will reject "--bitburner-fury-voltage 1350", and even if it didn't the boards are incapable of complying - which is just as well, because if you really put 1350mV through the Bitfury chips that would pretty much certainly fry them.

But still, be careful with your voltages!

roy

Finally some good information. I used 1300mV, but in cgminer i see ~1060mV. Shouldn't it be 1000mV?

Edit: what is the max safe temp? 55? 60?

Right... the original firmware (which most of us are running) simulates an Avalon-based board..... so you ask for 1300mV and get (approx) 1000mV

I guess there is some variation.... my boards tend to report about 25-30mV above what you'd expect.... I'm still running the BTB firmware so I set "--bitburner-voltage 1350" to select 1050mv, but cgminer shows the voltage varying between about 1075mV and 1080mV.

The discrepancy you see is more than I do - but ultimately the voltages we are talking about are pretty small and some level of variation is to be expected, I guess....

roy

ETA: There are presumably two separate circuits in there: one that sets the voltage, and one that measures the voltage.  Neither of them will be perfect, so the voltage that the chips will actually run at will not be exactly the voltage you ask for.   And then the voltage that is measured will not be exactly the voltage that the chips are running at.  It's the nature of the world - most all measurements are imperfect.

So it's unsurprising that the voltage measured isn't exactly the voltage set.  And the voltage the chips are running at?  Difficult to say, exactly...  As I said, measuring things accurately is hard...
 

[Roy Badami]

Edit: what is the max safe temp? 55? 60?

My boards run at around 58C give or take a degree or two.  The default setting in cgminer's Avalon driver (which is used to drive all Bitburner boards, including the Fury boards) is to shut down the boards if the stack reaches 60C -  but you can adjust this with --avalon-cutoff.  I set this to 62C on my boards because I was worried about accidental shutdowns - particularly because in my tests I was hitting a bug where the boards didn't auto-restart after they cooled down and I had to restart cgminer to get them hashing again (but that may be fixed in newer firmware and/or newer cgminer - dunno.)

No idea how safe those temps are though....not seen any official discussion of temperatures.... or actually any discussion of temperatures....

roy

[shapemaker]

Here's what I'm getting with 4 boards:

Code:

cgminer version 3.6.4 - Started: [2013-10-19 02:31:44]
--------------------------------------------------------------------------------
 (5s):212.5G (avg):204.9Gh/s | A:140672  R:256  HW:6816  WU:2861.3/m
 ST: 0  SS: 10  NB: 10  LW: 201638  GF: 0  RF: 0
 Connected to stratum.XYZ.com diff 128 with stratum as user XXX
 Block: 000e1b3fcefb9a83...  Diff:268M  Started: [03:19:32]  Best share: 172K
--------------------------------------------------------------------------------
 [P]ool management [S]ettings [D]isplay options [Q]uit
 BBF 0: 46C 290 1113mV | 208.7G/204.9Gh/s | A:140800 R:256 HW:6820 WU: 2862.1/m
--------------------------------------------------------------------------------

 [2013-10-19 03:19:36] Accepted 010c0156 Diff 244/128 BBF 0 pool 0
 [2013-10-19 03:19:47] Accepted 005a76a0 Diff 724/128 BBF 0 pool 0
 [2013-10-19 03:19:50] Accepted 00fab99c Diff 261/128 BBF 0 pool 0

This is with '--avalon-options 115200:64:10:d:290 --bitburner-fury-voltage 1100'. I'm cooling the boards with outside air at 2 deg C currently. HW errors are 4,8%, the lowest I've had them at was 3,3% @280 MHz. However, 290 gives more WU/m (2880 or so vs 2810), so it's still better.

EDIT: at 300, hardware errors rise sharply to over 10%, so it's not worth it without more core voltage, which is impossible without hardware mods.

[zulover]

I see this new cgminer reads them as bbf, does it give some better performance than 3.5.0?

[QuiveringGibbage]

Arrived Monday

http://imgur.com/a/vurFX

Low hash rate?

QG

[zulover]

Mine is 49ghs constant last week

[mtbitcoin]

I see this new cgminer reads them as bbf, does it give some better performance than 3.5.0?

I believe it was the latest firmware that would show them as BBF instead of BTB

[zulover]

And performance wise I see he had a few extra gig

[shapemaker]

And performance wise I see he had a few extra gig

The increase was only because I increased the clock freq from 280 to 290, and started cooling the boards directly with cold air. If the temps creep over 50 deg C, hardware errors again increase a lot, making the frequency increase practically cancel out. I haven't ran the boards alone, but maybe I should. Then I could tune each board separately and maybe gain a few GH/s more. Seeing as the CAN-bus connection is still fragile as a glass vase, using separate USB cables might still be "better"...

[Roy Badami]

I see this new cgminer reads them as bbf, does it give some better performance than 3.5.0?

I believe it was the latest firmware that would show them as BBF instead of BTB

To recognize the boards as BBF, you need both the BBF firmware and a recent cgminer - either 3.6.0+ or the someone42 fork.  (I think the official cgminer is the preferred solution now - should be no reason to use someone42's fork now his changes have been merged back into the official cgminer unless you have problems with 3.6.  ETA: Aparently 3.6 isn't quite stable on Windows or OSX yet)

roy

[juve4v]

For those having trouble reflashing, perform these steps in exact order-make sure board is not powered or conected via USB, first :

1.Start pic32prog.exe
2.Connect the USB cable -if you have already plugged in the mini USB end, leave it and just disconnect/conect the USB end to PC.
3.Power cycle the board
4.Click on connect tab from  pc32prog in 5s
5.Reflash

Edit: So far I'm experiencing some problems with original fw+ cgminer 3.6.1 or  3.6.2, so I 'm on 3.5.1 now testing more . I didn't reflash my board yet cause I'm ok with performance so far, getting arround 51 Gh/s on  280/1390.I also set cgwatcher to monitor and restart the board every 5 h.

[RoadStress]

For those having trouble reflashing, perform these steps in exact order-make sure board is not powered or conected via USB, first :

1.Start pic32prog.exe
2.Connect the USB cable -if you have already plugged in the mini USB end, leave it and just disconnect/conect the USB end to PC.
3.Power cycle the board
4.Click on connect tab from  pc32prog in 5s
5.Reflash

Edit: So far I'm experiencing some problems with original fw+ cgminer 3.6.1 or  3.6.2, so I 'm on 3.5.1 now testing more . I didn't reflash my board yet cause I'm ok with performance so far, getting arround 51 Gh/s on  280/1390.I also set cgwatcher to monitor and restart the board every 5 h.

Will try later and will also post a small review. Have you tried 3.6.4? I was having some issues with 3.6.1 (some kind of memory leak and cgminer was closing when ran out of memory), but it seems fixed with the last version.

[juve4v]

Yes, I got same errors with memory leak in cgminer 3.6.1 to the point the board just froze and cgwatcher couldn't restart it, that's why I went back to 3.5.1.
I'm trying as I type 3.6.4 and see how it goes

[shade88]

I also skip the new cgminer cause it's crash after a few hours running. Still using cgminer from someone42.

Just receive a refund email, in time for another hardware purchase

[RoadStress]

I also skip the new cgminer cause it's crash after a few hours running. Still using cgminer from someone42.

Just receive a refund email, in time for another hardware purchase

Order number?