[GUIDE] GridSeed 5-Chip USB, Blade & Black Miner Support/Tuning

[Decagrog]

Hello.
I tested the results of my mods overnight and, so far so good!
I decided a couple days ago to go ahead and try modding one of my pods.
I made the first mod which is to jump the two resistors as shown with the yellow stripes. Done deal.
I got a bit better performance out of it @ 900MHz and 950Mhz but still pulled a red nonce every once in a while and at times they would fill the window (cpuminer) for a short birst then go away.
I thought "how strange, what is causing this? Could it be a timing issue?"
Bingo!
The guy who did the mods and had his GS5 running at between 1000 and 1100MHz stable but with a few red nonce's gave me the tip...
I thought, let me try the pencil mod.
So I tried lowering the PLL voltage first to around 1.01V. Seemed to stabilize running at 950MHz so well that I only saw a red nonce every great once in a while over an hour or so of operation where before it was about 3 times as much.
Okay, so changing the resistor was next.
I changed out the 36K for a 38K 5% resistor.
BINGO!

Now I have stable 'no red nonce' operation at 1000MHz!... It's been running stably for over an hour now with ZERO red nonce's! Perfect!
I'm going to let it run overnight and see how things are in the morning. I have a good feeling this is the trick I was looking for.

Thanks to the author of this mod!

I'll save further congrats for later on if I see stable operation over time.
Oh and one other thing. I decided to go ahead and test my theory that lowering the voltage of the cooling fan to 5V was not a good idea due to the usual problems associated with magnetic DC motors.
I guess the drive electronics of DC fan motors has come a long way since I last farted around with them years ago.

I'm driving it with 5V now without any problems so far. Runs just fast enough to keep air moving at a good enough rate to keep the unit cool to the touch! I mean below 90*F. More like 79^ to 81* currently.
And man, it is quiet! Music to my ears! The pod was cooking at 115 to 120*F at 1000MHz which is understandable. That's crankin, for this little bugger!

Not running hotter or pulling more current like I thought it would, at all.
Sorry for scaring some of you guys out of trying it but I guess I get to make up for it by risking burning up my fan etc.
I'll let you know how that went overnight too.
I will also attempt to run it off of the 5V USB power next and see if it screws anything up comm wise or power wise.

It's pulling less than 75mA running. Pulls 140mA start up for a quick second before it calms back down.
I think it would be easy enough for most folks who end up doing this mod, provided mine works out okay, will be able to compensate for the additional load it puts on USB power. But it's a short cut method I really don't recommend in general. One should use a voltage regulator to drop from 12V to 5V. There is more headroom on that rail too in most cases.

Today I'm going to monitor the performance of my 1000MHz hasher for stability and expected increase in pool results as well as connect the fan to the 5V rail of the USB port and see if it causes any problems.
I don't think it will but if worse comes to worse, I'll use a 7805 voltage regulator to drop to supply the 5V from the 12V rail.

Onward!
Peace
Wolfey2014

I'm pondering about the voltage regulator...since the 7805 is a bit outdated and there's little space I'm considering the  LM2940 in a SOT223 package. It have the same 7805 pinout but a lower voltage drop ( 500mV ) and it's more efficient. The  TO-263 seems perfect to fit in the original fan pins on the pcb ( 2.3mm pin to pin )

[abbeytim]

getting 361.7kh/s cgminer 850 clock ltc mode only

[usao]

getting 361.7kh/s cgminer 850 clock ltc mode only

That's about what I get, but im worried about the HW errors. I do get a lot of them, about 1-2K over 24 hours at that speed.
When I go down to 800, I only get about 10-20 in a 24 hour period.
Do the HW errors cause any permanent or cumulative damage to the pod?

[wolfey2014]

getting 361.7kh/s cgminer 850 clock ltc mode only

That's about what I get, but im worried about the HW errors. I do get a lot of them, about 1-2K over 24 hours at that speed.
When I go down to 800, I only get about 10-20 in a 24 hour period.
Do the HW errors cause any permanent or cumulative damage to the pod?

Had the same problem with my GS5's.
Adjust your comm port FIFO buffer settings.
First try setting Transmit buffer down to 14 - that might do it.
If it doesn't completely handle it,
reset Receive buffer to 8 also.
Then you should be good to go.
After you adjust the FIFO's, disable and enable the port so the settings will take.
Works great for me on my Win 7 laptop via 13 port hub to 6 GS5's.
Good luck!
Wolfey2014

[b0tzy29]

PLEASE HELP!

Every now and then my miner will stop finding shares.  Why does this happen?  The green light will stay on for extended period of times and sometime the red light stays on.  I feel as though this is a driver issue.  The miner has work being sent to it but never finds/accepts any shares and my pool reports hash rate of 0.  If someone could help me this, I would be eternally grateful.  Thanks in advance.

[wolfey2014]

PLEASE HELP!

Every now and then my miner will stop finding shares.  Why does this happen?  The green light will stay on for extended period of times and sometime the red light stays on.  I feel as though this is a driver issue.  The miner has work being sent to it but never finds/accepts any shares and my pool reports hash rate of 0.  If someone could help me this, I would be eternally grateful.  Thanks in advance.

Right. That's typical. Unfortunately.

Set your comm port Transmit buffer to 14.
Reset your comm ports. Disable, then Enable.
If you are overclocking i.e. over 600MHz, you will probably need to set the Receive buffer to 8, as well.

Give this a try and let me know how it goes.

Funny that the Green LED stays lit, solid? Never seen that before.
Usually blinks on and off in 1 second intervals. That's all it does.
The Red LED blinks on and off in 3 second intervals with intermittent flickering while the hasher is hashing.

Wolfey2014

[CartmanSPC]

getting 361.7kh/s cgminer 850 clock ltc mode only

That's about what I get, but im worried about the HW errors. I do get a lot of them, about 1-2K over 24 hours at that speed.
When I go down to 800, I only get about 10-20 in a 24 hour period.
Do the HW errors cause any permanent or cumulative damage to the pod?

I have noticed that HW errors depend on the worker diff. The lower the worker diff the higher the HW errors...but im only talking about 10 HW errors over 24 hours worse case.

[romang]

side view on the rack ... 20 gridseeds, 1 raspberry, 2 usb hubs ... everything powered by the atx psu ..

Love it!

[wolfey2014]

side view on the rack ... 20 gridseeds, 1 raspberry, 2 usb hubs ... everything powered by the atx psu ..

Love it!

Hey, just a heads up you guys, from the gridseed voltage mod thread... Enjoy!
Nice rig darkfriend! Well done!

I am giving out 0.6BTC to the first to get his Gridseed miner stable at 1100 MHz (<10 HW error in 24h) and post the steps to mod the miner. So far I have managed to get it stable at 1013 MHz, but I feel we can push it further.
I have found the only way to push these further is by replacing the 36k resistor with a higher one and playing with PLL voltage, you need to lower the resistance of R211 or R212 by tracing it with a pencil. Measure the voltage across both resistors and it should be around 1.1v total. If you want to increase PLL voltage, trace R212, to decrease it, trace R211 (go back and forth atleast 10-20 times or you will not notice a change). Simply use an eraser to revert the mod. Personally 1.05v seems sweet spot.



Disclaimer: as always, I am not responsible for damage to your miner, do this at your own risk!

What is the test point on the card for checking/monitoring the voltage?

Thanks
Wolfey2014

Core voltage can be measured across the 36k, assuming it's voltmodded (by default it would be the 33k)
PLL voltage can be measured across R211 and R212.

Hello Sandor.
I made the modifications to 2 of my GS5 USB's.
Both jumpers done.
38k @ 5% resistor added in place of the original. I used a 1/8W linear type.
Pencil traced R112 to lower PLL voltage to 1.018 or so -virtually zero red nonce's.
I've never used the pencil trick before. Amazingly effective. Just hope its permanent.
1.05 seemed to high as it still got some red nonce's.
Both GS5's are rock steady for over 40 hours now. No problems at all.

They run sufficiently warm enough now that I put the fan back on them but am running them off the 5V USB with good results. Only pulling about 78mA. 140mA surge at start up but it lasts less than 2 seconds. They are whisper quiet and provide enough air volume to keep them nice and room temp cool. They were running at nearly 118*F fanless as expected after the mods.
I'm sure they can run hotter than that without problems but I also didn't want to just let the fans go to waste

I'm very pleased with their stable performance at 1000MHz!
Running it solo, I'm seeing nearly 600KH's out of it at the pool.
NICE!

So, what's next? How do we squeeze more performance out of them? Any ideas?
I'm game!

Peace!
Wolfey2014

[maxzilla]

I could really use a tip here. I have followed the instructions or lack of and am still not able to view the interface. I'm sure I'm missing something . . .

I have the latest firmware for the raspberry pi. It boots up to a command prompt that prompts for

hashracontroller login:

then password.

I cannot find the default settings anywhere in the manuals or online. Can someone please help me out here? Am I doing something wrong?

Thanks in advance!

Fyi, using the controller I'm running at 3.0
3.6 would be nice

If i where you, i would use user:pi pw:hashra  

Thanks, I'm curious . . . Where did you come up with that? Lucky guess? Should I have known, lol?

Just trial and error.....  

  That worked! Thank you very much. I'm still not able to view the interface though after following the instructions in the manuals or lack there of. After logging into hashracontroller command prompt I see this screen. I have tried re-booting and removing the line from the text file that I added to change password.

What am I doing wrong?  

Do I need to install Raspian from Noobs before writing the hashra image file to the disk?

Thanks!

Did you enter the IP in your webbrowser ? You are not able to see the interface via ssh-connection. There is no connection necessary, you don't need the ssh password to connect to the hashra gui. Btw. i figured out, that the hashra firmware hashes with really good 7200kh, BUT, the pool only sees around 3000kh.......

You probably plugged in 20 miners on the PI, this is a problem with CGMiner.
When you click on the UPDATE CONTROLA link, it will upgrade your firmware to version 1.2.0, which has BFGMiner for scrypt hashing This solves the Cgminer problem. After this, your pi will be mining at a solid 7200mhs (and this time, it will also reflect on the pool)

[darkfriend77]

...

You probably plugged in 20 miners on the PI, this is a problem with CGMiner.
When you click on the UPDATE CONTROLA link, it will upgrade your firmware to version 1.2.0, which has BFGMiner for scrypt hashing This solves the Cgminer problem. After this, your pi will be mining at a solid 7200mhs (and this time, it will also reflect on the pool)

I'm using for my 20 miner rig on one rasbian a clean weezy img + cgminer .... this works stable and reports 7.1 gh at the pool ... is there a cgminer problem? for more then 10 miners?

[darkfriend77]

...

You probably plugged in 20 miners on the PI, this is a problem with CGMiner.
When you click on the UPDATE CONTROLA link, it will upgrade your firmware to version 1.2.0, which has BFGMiner for scrypt hashing This solves the Cgminer problem. After this, your pi will be mining at a solid 7200mhs (and this time, it will also reflect on the pool)

I'm using for my 20 miner rig on one rasbian a clean weezy img + cgminer .... this works stable and reports 7.1 gh at the pool ... is there a cgminer problem? for more then 10 miners?

Here a 4gb Hashra image ... be sure to change pools before mining ... it's set to my setup right now ... as I used to test it ...
will try it out with 20 miners in a few minutes ...

Hashra 4 gb image

[miner765]

Did anyone try making/made a autostart script for pi in case in restarts? e.g. when power restarts -pi starts mining automatically. I think it would be really helpful. And I personally would tip a few coins

[maxzilla]

...

You probably plugged in 20 miners on the PI, this is a problem with CGMiner.
When you click on the UPDATE CONTROLA link, it will upgrade your firmware to version 1.2.0, which has BFGMiner for scrypt hashing This solves the Cgminer problem. After this, your pi will be mining at a solid 7200mhs (and this time, it will also reflect on the pool)

I'm using for my 20 miner rig on one rasbian a clean weezy img + cgminer .... this works stable and reports 7.1 gh at the pool ... is there a cgminer problem? for more then 10 miners?

Here a 4gb Hashra image ... be sure to change pools before mining ... it's set to my setup right now ... as I used to test it ...
will try it out with 20 miners in a few minutes ...

Hashra 4 gb image

yes CGMiner is not stable with the GD5 chips, It causes them to "zombie out" after a while. When 20 miners are plugged in to 1 CGMiner instance, only half of them are actually hashing properly, while still showing maximum local hashrate, hence the much lower hashrate on the pools. BFGMiner (installed on CONTROLA 1.2.0) works perfectly with 20 miners (or more? haven't tried more). I would love to see the performance of CONTROLA with 30 - 40 miners on 1 PI...anyone want to try it?

[maxzilla]

Did anyone try making/made a autostart script for pi in case in restarts? e.g. when power restarts -pi starts mining automatically. I think it would be really helpful. And I personally would tip a few coins

CONTROLA starts mining automatically when it boots up.

If you're not using CONTROLA,
you can add the mining command in /etc/rc.local
This command will be called on boot time.

[darkfriend77]

...

You probably plugged in 20 miners on the PI, this is a problem with CGMiner.
When you click on the UPDATE CONTROLA link, it will upgrade your firmware to version 1.2.0, which has BFGMiner for scrypt hashing This solves the Cgminer problem. After this, your pi will be mining at a solid 7200mhs (and this time, it will also reflect on the pool)

I'm using for my 20 miner rig on one rasbian a clean weezy img + cgminer .... this works stable and reports 7.1 gh at the pool ... is there a cgminer problem? for more then 10 miners?

Here a 4gb Hashra image ... be sure to change pools before mining ... it's set to my setup right now ... as I used to test it ...
will try it out with 20 miners in a few minutes ...

Hashra 4 gb image

yes CGMiner is not stable with the GD5 chips, It causes them to "zombie out" after a while. When 20 miners are plugged in to 1 CGMiner instance, only half of them are actually hashing properly, while still showing maximum local hashrate, hence the much lower hashrate on the pools. BFGMiner (installed on CONTROLA 1.2.0) works perfectly with 20 miners (or more? haven't tried more). I would love to see the performance of CONTROLA with 30 - 40 miners on 1 PI...anyone want to try it?

trying out with more then ...

[pauldulong]

I have some problems connecting these to a Windows PC, I tried to follow the guides provided. The problem is, when I connect the HUB all devices show up in the device manager. I can't install the Zagig driver of the HUB (because 10 devices show up - see screenshot) and therefore I can't start CGminer. CPUminer works if I specify a specific COM port for a single unit. My preference strongly goes to get this working on CGminer since I want to over clock and monitor via the API.

Any suggestions??

https://photos-5.dropbox.com/t/0/AAAwtlJIH4Fafe8y6nYQseqlgqwgvPVKNTh2Tt1PIOGdOg/12/3583623/jpeg/320x320/1/_/0/2/screenshot4.jpg/TqlTOiZhhe8w0IEBWvqJb7HpanjD-J1xbvsT_Kyefx8
Sorry for the shitty screenshot
https://www.evernote.com/shard/s12/sh/80639127-52a0-4a86-9cd5-d83d5428662b/3a42e2298dcad63e7a72e5cb3fbbaf2a/res/95578a05-e15a-4545-8510-a3144a32474c/skitch.png?resizeSmall&width=832
https://www.evernote.com/shard/s12/sh/bbf78403-dd53-4111-99c8-56c0c9b7c1c8/def3ce5c8b05ab7558d4302fa9f8717c/res/cfdf990d-d617-42e8-84d4-c196e6567a16/skitch.png?resizeSmall&width=832

[miner765]

I'm interested - could they be hashing slower or reporting "blank shares" if they have absolutely no fan around them? (at 850MHZ)

[miner765]

https://www.youtube.com/watch?v=28qHmaZXdmI&feature=youtu.be

So you do have two fans blowing onto them. I was talking about a completely fanless setup.

[romang]

getting 361.7kh/s cgminer 850 clock ltc mode only

How long has it been running like this? Any crashes?