Topic: GRIDSEED G-BLADE Overclocking 7Mh/s, improvements and repair - page 18. (Read 74047 times)

I think I found out the problem.
I read on some froum that COM ports over 10 can be problem, so I put my Blades under 10 com ports and I will see.

Could be MOSFET or main capacitor wearing. The powerboard is already close to the limits, even non overclocked.

No fan works fully and I even tried with hand how hot board is but not that hot.
I assume maybe can be a power plug, one stops to work something is disconnected in it.
After a while and high current maybe connection broke, I will see tomorrow I will change it.
Because it is no reason why after more than one month working with no problems suddenly not working anymore

I haven't got any problems with my two Gridseed's I bought three weeks ago from monster-asic.com. Fans working fine, no problem here. But yes, 7 MH is a bit optimistic, I managed to get mine to 6.5 MH

regarding your trouble with blade stopping to send shares, I also have similar trouble when for some reasons the fan stops to work.
mosfets overheating, and closing.

irfb7437 and 1404 could work.
7437 and 7430 should work, I can't see any reasion they won't, but I'm not an electronics guru, just experimenting and learning very fast.

if your blades stop with cpuminer too, it should be hardware related.
try at least to cool the 3 mosfets over their junction temperature, they do stop to work and the blade stops hashing.
If you have the libusb error with cgminer, direct connecting to the computer with a short usb cable, could help too.

irfb7430 ordered, I should receive them next week.
In the meantime, I'll prepare a board to fast swap the to-220 mosfets and post a few pics of the easiest soldering points for upper and lower gates external fets
I'll also look for good solder points for a complete external powerboard.

so I already got IRFB7437
and I am waiting to get 1404
and that should work?

let me know if you test it if it works.
I need first to find out why my blades stops to mine after that I will change mosfets.

I will add pics.

As far as I now understand the mosfet used with the buck converter, you need lower Rdson for the lower gate ones, and higher Rdson for the higher gate one(s)

5300 is 1.3mohm (dual lower gate)
CSD17556 is 1.4mohm (lower gate)
5301 is 1.85mohm (some of you have seen it instead of the 5002 higher gate)
5302 is 2.1mohm (single higher gate)
CSD16321 is 2.4mohm (higherr gate)

7437 is 2.0mohm, that's why it works as replacement for 5302
7430 is 1.4mohm with TO-220 package. it should work as the lower gate with either 5302 or 7437 as the higher gate

7004 is 1.4mohm, should work as lower gate, that's why it wasn't working when I put it in place of 5302 (blown higher gate)

The last one I'm still waiting for delivery is
1404 with 4.0mohm rdson.
it should work ok as a higher gate.

So I have 2 options

1404 higher gate and 7437 lower gate (lowest rdson)

ordering 7430 to have 7437 higher gate and 7430 lower gate

great I will also check which can be good replace for lower gate.
Don't know what you mean with...

update on the blade, I can confirm that irfh7004 don't work on upper gate when you have dual 5300 on lower gate.
but CSD16321 works with dual 5300 on lower gate. I received a pod with this configuration, and I tried it on the blade.
It's happily hashing again.

next step will be external quad parallel fets on each gate to feed the up1509 before trying complete external powerboard thaqt could be piggybacked to a blade

we still need a good TO-220 candidate for lower gate

as a side note, if you do replace the power plug with screw terminal like me, put them on the back of the board, that will allow for more space to solder the upper gate mosfet when it will die.
I've melted half the screw terminal with my hot air station when desoldering and resoldering the fet.
It will also be easier to screw/unscrew.

yes I tried cpu miner from link you post It makes a little lower hashrate but works great. But not for blades because works fev hours and than stops to mine. my blades are in room at 20 oC so overheating can't be reason and it must not happend on not overclocked ones.
I will try with some extra fans blowing from a side but I think it won't solve a problem, why is this happening now after more than a month working ok? And I didn't do any changes.

Could be overheating on the power mosfets.
when too hot, there is continuous connection between source and gate, same as when they are dead.
even when they are cooled down, the blade needs restarting.
try heatsinks on the fets and/or a secondary fan blowing by the side of the blade on the powerboard, just to check.
Try cpuminer.
cpuminer solved every trouble I had with cgminer

So you suggest to use a  CSD 17556 with IRFB7437?
is there any to220 instead of CSD 17556

I still didn't overclock mine because I have strange things with my not overclocked blades.
Both stops to work after fev hours, cgminer works but hashrate drops to under 1 Khash.
Had someone the same thing? What can be wrong?
I already changed PSU

Quick update after reading hundreds of pages on buck converters, it looks like upper and lower gate mosfets can't be tne same ones.
There must be difference in a few parameters to allow the gate to work.
I'll give more details once I understand it correctly.

In the mean time, basic powerboard redesign has started.
I'll try to use mostly through hole components and when it isn't possible, pqfp of similar components with legs to allow easier rework/replacement.

buck converter will be tps40140, well documented and that will allow any voltage from 0.7 to 2.2v easily.
That will allow the powerboard to be re-used for newer asics later.
That mean independant powerboard that will connect to the chips board and probably independant communication board for usb.


I've also blown a mosfet on one of my blades, as usual, it's the upper gate one that is alone. I suspect that shrinking the design from 2 upper + 2 lower to 1 upper + 2 lower fet puts stress on the single when overclocked. I'll solder a replacement one tomorrow. IRFH7004 doesn't seem to work in this place or something else is shorted. I'll try irfb7437 that is proved working on a pod or I'll replace with the newer components used by gridseed as I have some.
experiments so far:

Upper gate, pin 19 -> 5302 or CSD16321 or irfb7437 (works on pod) irfh7004 doesn't work on blade?
Lower gate, pin 2 -> dual 5300 or dual CSD 17556

I've found alternate soldering points on the blade for the gate of lower fets that will make experiments easier with fets soldered outside of the blade

I don't use raspberry pi.

Yeah you MUST use a powered HUB, and cgminer is a shit on windows for blades... Now running 5 blades on Hashra img, individual clock tuning, smooth interface, nice as fuck !

Thank you. I tried using cpuminer and a powered Hub and its working 

Of course i connect the blades first and then run cgminer. But thanks for the hint with cpuminer. I haven't tried tis yet.

first connect blades and than run cgminer.
I use cpuminer it works much better

Hey, i tried to run multiple G-Blades on Win 7 as well. But it only works with one Blade at once. As soon as i connect the second Blade cgminer crashes at atartup, or isn't finding any of the Blades. I am plugging both G-Blades directly into the Computer USB Ports (so 4 Ports). Have you encountered similar Problems? Or maybe do you think it will work if i use a powered USB-Hub?