Topic: OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 522. (Read 5806088 times)

Yeah that seems to suggest it also.
But was there a 'BUSY' reply as BFL-Engineer said there is?

15.3355, in fact

You got a link/reference?
Coz that one from BFL-Engineer ... is from BFL-Engineer

When my 2nd FPGA "died" in my BFL I did exactly this.
The result was useless in my case.
Adjusting it up and down (I used a getenv() to set it) appeared to make no difference at all.
(I set it to delay getenv() ms before sending new work - I tried from 100 up to about 1000)
I got the impression that the idle time was doing more bad than good

Though the death was probably something else in it crapping out since:
In my case the final fix was stupid anyway
I installed the 880 bitstream ... and it's been fine ... 2 weeks and counting ...

tl;dr story follows
I got it 4-May, but working OK from 5-May then got problems 11-Jul ... as the weather got colder ... it started to throttle and fail
(It's now the coldest part of winter here)
Then it would only mine on one FPGA.
I'd never touched the firmware coz that meant moving the stupid thing to my kids windows computer and plugging it in there to use the POS EasyMiner program.
So I did this (moved it upstairs to me instead of down in my cold basement garage) and EasyMiner said it was no good (running a test) ... but cgminer would still mine on one FPGA.

I did find a few times that if I pulled the power on it completely it would start up mining normally at 825MH/s for a while and in one case this lasted 2 days (21-Jul to 23-Jul)
After being annoyed at it for 2 weeks (code changes, power cycles etc) I decided to try kill it by putting in the 880 bitstream (26-Jul)
(sending from Aus->USA is not a cheap thing to do, so while it still partially worked I didn't want to do that)
With the 880 bitstream it's been fine ever since, upstairs here next to me, under a chair, reporting 10C higher
(air coming out is only slightly warm, not hot)
My BFL is the version with the new heatsink but only 1 fan on top so not exceptionally loud

Yeah my opinion of the throttling hardware/design is ... not good

If it means anything, I think BFL has specified that a throttle period lasts 15 seconds.

Ah well, then that just means BFL is full of shit as usual:
https://bitcointalksearch.org/topic/m.885143

Now to decide a workaround for more BFL design shit that doesn't involve simply saying "a BFL appearing dead for up to 30s is OK" ...

The MiniRig cards are ~twice this speed, and I have one person with 2xMR running 2.6.2 modified and he's had 29 comm errors out of 4,171,889 accepted shares - so no issue at all.

... sigh - I presume this crap is all due to the fscking stupid throttling design done by BFL where their hardware decides and controls the whole process rather than allowing the software to control it (or know about it)

Also, I started adding the temp to the log when a unit throttles...
I'm thinking about adding code with slows the unit a bit as it reaches it's individual throttle temp. Like waiting 500ms before sending new work or something. Maybe. Dunno.

I've done that, get the following:

ID: BFL0   Dev Max: 0.024049   Dev Min: 0.000366   Dev Av: 0.009301
ID: BFL1   Dev Max: 15.335547   Dev Min: 0.000389   Dev Av: 0.012155
ID: BFL2   Dev Max: 15.335506   Dev Min: 3.8E-5   Dev Av: 0.011484
ID: BFL3   Dev Max: 15.33545   Dev Min: 0.00036   Dev Av: 0.027547
ID: BFL4   Dev Max: 0.02345   Dev Min: 0.00036   Dev Av: 0.009153
ID: BFL5   Dev Max: 0.023163   Dev Min: 0.000404   Dev Av: 0.008941

So I put get times immedately before and after the serial port read function inside BFgets.
Dev max is (surpisingly) the max time a read took
Dev min is (surpisingly) the min time a read took
Dev av is (surpisingly) the overall average, that is all the read times summed, then divided my the read count.

So it's pretty apparent that when the single throttles, comms just stop.

When I did this, I added a 30 delay after startup to start, and a timeout of 10 second in the loop.

I would be careful about this. You may want to add a TIMEOUT or use the ping technique to make sure threads are closed, and that the OS has released handles, otherwise the new cgminer instance won't properly grab them. 5 or 10 seconds should do the trick.

You create a .bat file with a loop that runs CGMiner, and when CGMiner closes itself after 5000 shares, the loop restarts CGMiner. Lather, rinse, repeat.

i dont know how to use window scheduler and

--shares       Quit after mining N shares (default: unlimited) 

you mean the miner will quit after mining example : 5000 share and it will auto restart and mining again ?

It is quite normal to get reject just after new block and/or longpool on any version.

So, is it safe to use 2.6.4 for ltc mining now? I'm on solo and one message I received among the "new block detected on network" was "rejected 040ccc8e blabla Gpu1". Do I have to worry? Should I use an older version? Total noob to cgminer. Used cgeasy to configure (Thanks! great help) to set everything for solo ltc mining.

Edit: Thanks Rav3n ------------->

You can also use:
--shares       Quit after mining N shares (default: unlimited)

in a loop

well you should be able to with task schedule

is really annoying for the 7 days bug !

i wonder if possible to add auto restart feature for windows CGminer

so every 7 days it will auto restart itself !

Thanks. It looks like the code was up to date, but I built it before pulling the version number change on the laptop, sorry about that. Reuploaded packages.

http://ck.kolivas.org/apps/cgminer/cgminer-2.6.4-x86_64-built.tar.bz2

One of my binaries? Which one?