Topic: Fury/Blizzard tuning and mods - page 16. (Read 115281 times)

This one from here: https://litecointalk.org/index.php?topic=20477.0

There are also some good tips in there for optimizing your performance.

Bfgminer 4.4.0-32-gac4e9b3, compiled with --enable-scrypt

What versions / builds of cgminer or bfgminer? 
The cmds are different.  Could be -S zus:all or -S zeus:/dev/ttyUSB2 (notice zus & zeus) depending on which version of bfgminer for example.

swapped usb ports?  swapped cables?  swapped power supplies?  (sorry....obvious and dumb questions but have to ask).

All of my Fury's use the Silicon labs USB chipset.  My Falcon uses the FTDI.  I've never had a problem with any of them being detected in a linux environment.  I did have one fury that would just stop working at times when using the power supply it shipped with.  After some testing of that power supply, it was putting out varying voltages of 12.5v all the way to almost 14v.  Tossed that power supply and used an atx psu for everything and it all returned to normal.

As for board markings, I'm not sure.  I think there are some different markings that are on the board, but I'm not positive.  One of my Fury's had a few more test points.

How do you determine which batch your Blizzard is from? I can't get cgminer or bfgminer to detect it. Have tried everything I can think of; I spent literally 8 hours yesterday trying to get it to work. I'm on Mint (Ubuntu).

If anyone is willing to take a look it would be greatly appreciated. Here is the thread on Stack Exchange with all of the details of my setup.

Contents of cgall.conf

I actually think it's a software problem and here's why.  If I run 3 fury's all is fine.  As soon as I add the 4th all hell breaks loose - it doesn't have to be the falcon added to the mix.  I've tried different difficulty settings, from 256 on up, not much change. 
And here's the real kicker.  If I leave out the "scan-serial" section and let cgminer auto-detect them all with libusb I get thousands of hw errors within the first couple minutes and it continues.  If I add the scan-serial  section and specify the ports then I get FAR less hw errors. 

I just changed the scan time to 15 this morning to see if that made any difference.  Likely not, but it's worth a shot.  It's just weird that 3 fury's is fine, but add more devices to the mix and I get walls of hundreds of hw errors and even more if I omit the scan-serial section of the config.

What are your exact startup parameters?

nst6563, if I understand correctly, some of the "unavoidable" HW errors are job discarded because one chip already found the hash.
They should be counted as duplicates or even ignored, but that's the way Zeus implemented them, and I don't think anything can be done software side.
Could  it be that sometimes, the falcon finds a hash and there were many duplicates?
But this should also happen with a single falcon, and it looks strange that so many identical shares are worked on with different cores at the same time.
On a falcon with 1024 cores, it would be like 30% duplicates at a given time. Maybe a work distribution error.
A way to have falcon and fury working at different diff settings would help. The difficulty looks really low for the falcon.

Anyone know how to run 4 Fury's and a Falcon off of a single instance of cgminer?  

Also...anyone else get a problem of like 100 hw errors right after launching cgminer?  And then seemingly random "walls" of hw errors here and there (like 300+ in a row)?

I can't get bfgminer to work on my setup at all.  It randomly declares my miners dead and doesn't revive them.  I get lower hashrate local and poolside with it than cgminer...and that's on the pools it works on.  It doesn't work on manicminer (merge mining pool), I get tons of rejects on nicehash with it, coinking and coinmine just get very inconsistent results and I never seem to get the full hashrate even if I point it to a specific coin's port.  Not sure why I have crap luck with it.  

:EDIT: In cgminer I can get all the zeus devices recognized correctly...however not all the speeds.  They're all set to a uniform speed.

:EDIT2: Got them all to their correct speeds....just running into the walls of hw errors.  Here's a screenshot:


Pic of the "wall of errors":


And before anyone says "your clocks are too high!"....They ALL work perfectly fine either alone, or in pairs, and all stay within a 5% margin of hw errors.  Put them all together and I randomly get these huge "walls" of hw errors on any one of them.  I even overclocked the rpi to 1Ghz thinking maybe it wasn't fast enough...made no difference. 

I am DL'ing it now.
I am looking forward to it.

From my testing of the latest minera & fury.

There seems to be a 5-7min detect delay. So be patient till it is detected.

For dmaxl cgminer, all you need is the freq cmd and your freq. "--zeus-clock 328" or w/e freq you use. Reboot the RPI by the gui, go in and start your engines miners, wait 5-7 min. You should be golden.

Good Luck!

Thanks michelem....1st post updated with the link

Hello guys,
I'd like to tell you the Minera project now supports Zeus devices too. Please give a try and let me know!

* v0.3.0 announcement
* Official site

Thanks!

Nice job on the cable

But how do you type with that keyboard

Uh,I run 3 Fury's per PCIE cable :



I made that cable myself  

Running 5 Fury's on CX 600 now,soon to be 7.Each PCIE cable can handle 200 watts.I used that same PSU on my Antminer S1 & it ran with no issues for 2.5 months  

rant - a lengthy discourse by a single performer, especially if irritated or upset.

If it walks like a duck....

My blizzard decided yesterday to stop working. It no longer has any accepted shares, but it still seems like it runs ok. CGWatcher shows it at the regular 1.35 MH per second but nothing is ever accepted. I have 3 setup, doing them individually ICA 0 and 1 are fine, ICA 2 is the problem. ICA 2 changes obviously if it's the only one connected to ICA 0, I have tried all combinations to get it to work. I am under the impression now that it is busted and I should be seeking a warranty replacement if there exists such a thing.

For PCIe on these you will only want 2-3 fury's per PCIe cable. I'm currently running 3 off of one pci-e

2 Boards, each has it's own power draw, this keeps it within the window of the PCIe capabilities

If you're about to go 3 fury's, you need a power supply that has atleast 4 pcie cables coming from the psu. And rule of thumb, one fury per pcie cable. so I was told. Otherwise you may overload the pcie adapters and psu.

This a basic question I made about the power the pcie pulls.
http://www.overclock.net/t/1495905/evga-supernova-g2-super-flower-leadex-psu#post_22418213

And I am wondering how the hell gaw miners are powering the

http://www.gawminers.com/products/13-mh-s-the-black-widow-by-gaw-miners-asic-scrypt-miner-free-zencontroller-week-8/

off two pcie cables...

http://hoosierminer.com/gawminer-black-widow-unboxing/

https://www.youtube.com/watch?feature=player_embedded&v=Roo_xzT_CDQ

Thanks for the input jekv2.  I was actually looking at that barrel tip power cable earlier this week.  Has all good reviews on it except one which semi scared me.



"One of the connections causes my gridseed to spark...
Gawminers did offer to replace it...
I would still buy from them
what can I say Sh#t happens..."

Defective products happen and it seems to be only one person (who cared to review which usually the people with bad experiences review products more than ones with good experiences) so far who had issues with the cable.  Also, the cable would be perfect for me since I'm about to go to 3 Furies.