GridSeed 5-chip USB miner voltage mod

[Kergekoin]

I have wierd anomaly with one gridseed 5chipper. It seems that only one out of 5chips is working. It hashes 80khs, but gives no HW errors.
I checked its PCB and saw no problems.
Any ideas?

Did you do a volt mod ...and which one...?

yeah, basic 2+2 dots solder together voltmod. Only one miner out of 72 is having this problem.

Sounds like a RMA...back to GAW...72 Gridseeds kit...right...?

I got the same kit delivered yesterday...best bang for the money...

modding voids warranty!

Ofc. Im not going to bother with attempting to RMA it.
Anyway. I undid the voltmod and still no improvement. Double checked all PCB components and found nothing wrong again. Going to wash it with toothbrush and cleaner. If theres some physical disruption then washing cleans it up. If theres faulty component, then i will probably never find it. So washing is my last hope.  

[amix]

To everyone that has done this mod, have any of you needed to upgrade cooling on the unit?

I was using thermal compound on mine, and one of mine shorted out and fried the gridseed. Apparently when you remove the pad, the screws compress down and potentially touch the unit.

Are you guys just leaving stock pads on there? I would like to get aftermarket pads, but to upgrade cooling on 50 of my seeds will be like $200 in thermal pads lol.

The thermal interface grease makes me paranoid now because of the 1 unit that fried.

this has been discussed and mentioned about the risks of shortening out componets etc.
there been many discussion about pads or paste or stock! we all know where you stand wolfey
thermal paste is better - but more messy!
the bits that are touching seem to be the crystal - ive dented mine! - and the chip next to it. theres the option to get the rimmel out and take a bit out of the center column on the heatsink.
I say use the thermal paste underside, thats where most heat goes! trim the heatsink, on top, pads - i use a bit of thermal paste as well to help with adhesion to the chips. be warned though

https://bitcointalk.org/index.php?topic=519112.msg6361304#msg6361304 - concerning the underside of the PCB!

to save some pad yher is no reson to cover the whole of the center column in thermal pad. just cover the tops of the chips - 4 small squares! you can make the one near the power supply a bit bigger to cover the crystal and chip.
and i found Zig post as well

EDIT : yeah adding the link to Zig post WOULD help : https://bitcointalk.org/index.php?topic=519112.msg6365619#msg6365619

hope this helps. theres no clean answer on this topic - just lots of advie and recommendaitons - good and bad , its really a person preference, until we get test data on temp and paste vs pads vs stock, but i dont see that happening anytime soon One last thing dont tighen the screw too much risk of short and risk of damage to PCB - i over tighten and dented my crystal!

[amix]

UDATE ON POWERING THE GRIDS WITH 5v
I found it ran nicely at stock settings - 650 mhz .
Pulled 0.77A and had a DVD volatge of 1.46. hash rate of about 250. For some reson reason the though i was only able to get one type of grid running with 5v. The newer type wiht the SMD leds.
Once modded though the hasrate dropped through the floor BIG TIME! went down to about 20-70! ouch! so back to 12 v it was.
So basically anyone runnign these at stock no overcloking or modding - why woud you though! - can run these on 5v wihtout effecting the hasrate and save on enery costs.

summary stock (no modsS!!! ) 650 - never tested 850 sorry - runs fine (SMD LED version only).
DO NOT run a modded grid on 5V your hashrate suffers.Even at 650!

[gbyg]

Does anybody know what this two parts(yellow marked)? my 5volt mode with 49.9k resistor swapped pot is blown(13 out of 20 GS) after 3 days 24/7 run for unknown reason. windows still recognize the com port and 5 chip still visible in cpuminer. if anybody knows the value of this two i might try to change this to see if it works.

I think they are marked as F824 & F823

[nst6563]

*image snipped*
Does anybody know what this two parts(yellow marked)? my 5volt mode with 49.9k resistor swapped pot is blown(13 out of 20 GS) after 3 days 24/7 run for unknown reason. windows still recognize the com port and 5 chip still visible in cpuminer. if anybody knows the value of this two i might try to change this to see if it works.

Ferrite Beads. 
300ohm@100MHz,3A

[gbyg]

*image snipped*
Does anybody know what this two parts(yellow marked)? my 5volt mode with 49.9k resistor swapped pot is blown(13 out of 20 GS) after 3 days 24/7 run for unknown reason. windows still recognize the com port and 5 chip still visible in cpuminer. if anybody knows the value of this two i might try to change this to see if it works.

Ferrite Beads.  
300ohm@100MHz,3A

Thanks for quick replay, any idea why would it might blow? Also is it 0603 ?

[nst6563]

*image snipped*
Does anybody know what this two parts(yellow marked)? my 5volt mode with 49.9k resistor swapped pot is blown(13 out of 20 GS) after 3 days 24/7 run for unknown reason. windows still recognize the com port and 5 chip still visible in cpuminer. if anybody knows the value of this two i might try to change this to see if it works.

Ferrite Beads.  
300ohm@100MHz,3A

Thanks for quick replay, any idea why would it might blow? Also is it 0603 ?

The footprint is actually listed as R1206.  I believe those are part of the 12v circuitry on the board, so it could be from a surge of some sort I suppose.  I've seen another report of those beads blowing for someone else but I don't recall what the total damage (beads + collateral) was.

EDIT:  For reference, the information about the board layout and parts list is on github here: https://github.com/gridseed/gc3355-doc or here: https://github.com/gridseed/usb-miner

[gbyg]

The footprint is actually listed as R1206.  I believe those are part of the 12v circuitry on the board, so it could be from a surge of some sort I suppose.  I've seen another report of those beads blowing for someone else but I don't recall what the total damage (beads + collateral) was.

EDIT:  For reference, the information about the board layout and parts list is on github here: https://github.com/gridseed/gc3355-doc or here: https://github.com/gridseed/usb-miner

Thanks for the help!

[nst6563]

The footprint is actually listed as R1206.  I believe those are part of the 12v circuitry on the board, so it could be from a surge of some sort I suppose.  I've seen another report of those beads blowing for someone else but I don't recall what the total damage (beads + collateral) was.

EDIT:  For reference, the information about the board layout and parts list is on github here: https://github.com/gridseed/gc3355-doc or here: https://github.com/gridseed/usb-miner

Thanks for the help!

No problem.  Those parts should be pretty cheap, hopefully that's all that was damaged.  Only way to find out is to replace them with identical spec replacements.  Someone mentioned the possibility of temporarily bypassing those parts to verify functionality of the gridseed - however if it were mine, I'd replace the parts before testing it.

:EDIT:  What thermal paste did you use?  Non-conductive I hope.  You might want to clean that up while you've got it apart.  In my opinion, there's no real point in using thermal paste unless you are positive the heatsink makes full contact with all 5 gc3355 chips.  On all the gridseeds I have there are 2 components which are taller than the gc3355's and prevent full contact.  I milled out part of the heatsink over those two components so the heatsink made full contact.   I posted a pic in this thread somewhere of it.  However, if you do that, then the heatsink will likely make contact with all the capacitors surrounding the gc3355's.  So I covered those with electrical tape to prevent a short.

[wolfey2014]

To everyone that has done this mod, have any of you needed to upgrade cooling on the unit?

I was using thermal compound on mine, and one of mine shorted out and fried the gridseed. Apparently when you remove the pad, the screws compress down and potentially touch the unit.

Are you guys just leaving stock pads on there? I would like to get aftermarket pads, but to upgrade cooling on 50 of my seeds will be like $200 in thermal pads lol.

The thermal interface grease makes me paranoid now because of the 1 unit that fried.

this has been discussed and mentioned about the risks of shortening out componets etc.
there been many discussion about pads or paste or stock! we all know where you stand wolfey
thermal paste is better - but more messy!
the bits that are touching seem to be the crystal - ive dented mine! - and the chip next to it. theres the option to get the rimmel out and take a bit out of the center column on the heatsink.
I say use the thermal paste underside, thats where most heat goes! trim the heatsink, on top, pads - i use a bit of thermal paste as well to help with adhesion to the chips. be warned though

https://bitcointalk.org/index.php?topic=519112.msg6361304#msg6361304 - concerning the underside of the PCB!

to save some pad yher is no reson to cover the whole of the center column in thermal pad. just cover the tops of the chips - 4 small squares! you can make the one near the power supply a bit bigger to cover the crystal and chip.
and i found Zig post as well

EDIT : yeah adding the link to Zig post WOULD help : https://bitcointalk.org/index.php?topic=519112.msg6365619#msg6365619

hope this helps. theres no clean answer on this topic - just lots of advie and recommendaitons - good and bad , its really a person preference, until we get test data on temp and paste vs pads vs stock, but i dont see that happening anytime soon One last thing dont tighen the screw too much risk of short and risk of damage to PCB - i over tighten and dented my crystal!

Where's the ROFL emoticon!??? 

[gtraah]

Guys,

I have just found something out here.... And I am starting to realize why  sometimes the Grids are not working at the full potential on the pool side when you have quite a few seeds..

And Also for people who find CGminer crashing, this is the reason below.

The issue is the USB power, those cheapo Chinese Hubs are really bad and unstable and are hit and miss in terms off good or bad..... FOr example I bought 5 x 10 Port USB hubs only 2 work OK not GREAT but OK and the 3rd is MEhhhhh still trying to work out if it can handle all 10 ports filled up or not..

In-fact even the 2 that work properly sometimes all ports are not detected properly when the mining action starts sometimes I hear the USB disconnect windows sound On-OFF- On - OFF

[wolfey2014]

*image snipped*
Does anybody know what this two parts(yellow marked)? my 5volt mode with 49.9k resistor swapped pot is blown(13 out of 20 GS) after 3 days 24/7 run for unknown reason. windows still recognize the com port and 5 chip still visible in cpuminer. if anybody knows the value of this two i might try to change this to see if it works.

Ferrite Beads.  
300ohm@100MHz,3A

Thanks for quick replay, any idea why would it might blow? Also is it 0603 ?

Could be blowing out due to becoming overwhelmed by too high a mean and/or peak current.
It could be (but shouldn't be) caused by the 49.9k mod due to the previous design's FB's having been substituted out for lower value to again, cut production costs or they are perhaps simply faulty chips in the first place. I hope the latter is the case or the 49.9k mod may now require upgrading certain power buss / support components like the FB's, to compensate for the increased power requirements of over-volting/clocking. See if the replacements ones get hot under normal conditions. If they do, perhaps it's due to the above or perhaps there is another yet discovered component issue. Could also simply be a run of bad pods. It happens.

[serpiente12]

Yeah you are absolutely right. Pencil trick just makes things worse. Even with 850mhz and voltage=1 now I get a lot of HW errors. Time to break out the eraser!

Edit: Erased and back to normal. Zero HW errors @ 850mhz. Damn - now to find someone who can solder a bridge that tiny.

It was really annoying to do. What I did was get a gob onto the general bridge area, and then ran the solder tip (I have a fairly fine tip) along both sides of the bridge to collect any stray solder.

EDIT: Going to try using this to solder the bridge: http://www.ebay.com/itm/151169919397

did you have anny luck using this @Andareed? is it as simple as it seems, sorry for the noobis question and it may be in the thread somewere so forgive me if im asking double...BUT is it liquid soldering paste? you just mix and push it out the needle done deal ? thanks in advance great job! ENJOY MINING!

[vipgelsi]

Are u guys getting a lot of he he errors? I get about 12 a day at 460kh

[Kergekoin]

Heres perfomance indication of voltmod1 gridseeds and latest sandor cpuminer with autotune on.
Running for 11 hours.

Sandor has done great work! If you are running autotune on, then dont worry about HW errors. autotune takes out every last bit of perfomance of your gridseeds.  
PS. Red box around the faulty miner with only 1 chip working (mentioned earlier posts here).

[gtraah]

Heres perfomance indication of voltmod1 gridseeds and latest sandor cpuminer with autotune on.
Running for 11 hours.

Sandor has done great work! If you are running autotune on, then dont worrie about HW errors. autotune takes out every last bit of perfomance of your gridseeds.  
PS. Red box around the faulty miner with only 1 chip working (mentioned earlier posts here).

Woww...

Isnt that too much HW errors? I am not happy if I get above 5 HW errors per 24 Hrs and I am getting 510KH

[amix]

Hardware errors are gonna vary from device to device chip to chip, even wiht same mod applied. for starters theres an error in resistor values even at 1%, for the resistor (r52 mod) then there is plenty of other resistors on the board as well - the main one thats gonna effect voltage is r50, connected to DGND from the buck convertor, plus the resistors for the DPLL voltage. All these will give a variance although small, but when measuring voltage in.00 volts it can make a difference.
Secondly these chips are Intel or AMD chips, manufactured to the same constiant and quality control as thoose so each chip itself will have a different optium performance. Not to mention the whole assembly of the grids isnt to the high standards and quality control we would like.

CPUminer i believe uses the hardware errors to adjust the hash rate, no hardware errors over x period hash rate goes up. x hardware errors over x period hasrate goes down. until it finds a balancing point.
When running a batch of grids from the same command file - with setting each one indivdually - a varying amount of hardware errors will be produced, as each device will start of the same and autclocked until it gets the hardware errors. After running CPUminer for say 24hr, if you take a note of the hash rate for each grid, set a batch file wiht autotune off and set each one indivdually youll probably find over the next 24 hours that the hardware errors are a lot lower if none at all.

[michelem]

I will start a new thread when ready but in the meantime I would like to share with you the work in progress behind this new project based on sandor cpuminer.

I'd like to test it with a great bunch of pods (20 or more) but I have only 3 now, may be someone of you could test it? It requires a linux controller like a raspberry with a web server (something like a Scripta img should work very fine, just change the document root to the Minera directory).

If there is someone who wanna try it please leave me a PM, I think I will release a public beta-version next week.

Hope you like it:

[darkfriend77]

I will start a new thread when ready but in the meantime I would like to share with you the work in progress behind this new project based on sandor cpuminer.

I'd like to test it with a great bunch of pods (20 or more) but I have only 3 now, may be someone of you could test it? It requires a linux controller like a raspberry with a web server (something like a Scripta img should work very fine, just change the document root to the Minera directory).

If there is someone who wanna try it please leave me a PM, I think I will release a public beta-version next week.

Hope you like it:

...

I would give it a go :-) ... send me the link

[sandor111]

I will start a new thread when ready but in the meantime I would like to share with you the work in progress behind this new project based on sandor cpuminer.

I'd like to test it with a great bunch of pods (20 or more) but I have only 3 now, may be someone of you could test it? It requires a linux controller like a raspberry with a web server (something like a Scripta img should work very fine, just change the document root to the Minera directory).

If there is someone who wanna try it please leave me a PM, I think I will release a public beta-version next week.

Hope you like it:

Quoted for awesomeness, I applaud your efforts.