GRIDSEED G-BLADE Overclocking 7Mh/s, improvements and repair

[J4bberwock]

this  IRFB7437 can replace all 3 chips?  IRFH5300/5301/5302

Why than they use two  IRF5300
and one IRFH5301?

what is difference with IRF1404 and IRFB7437? less power?

I found IRFB7437 for 1,17€ pcs

here is pic what I think will do with to-220 mosfet
this is look from a side

The orientation of the pqfn isn't the same on the 3 ones, so you won't be able to do it like this.
plus the legs of the TO-220 will be too short to bend them back to the other side

Since there will be 3 TO-220 next to each other, I'll probably go for a heatsink for the 3 of them and I'll let it stay on the PCB without soldering it, maybe glued with non conductive glue.
Look inside of a computer PSU to have an idea of what I mean, except in the PSU, the heatsink is soldered to the ground of PCB.

I have a few dead old PSU that I can take the heatsinks off for this project

[Blisk]

I thought to use heatsink like this
http://www.susan-parker.co.uk/zeus-heatsink-sk158-1a.jpg

about legs I will bend it to work it out or even solder some hard wire to it to be long enough.

so which mosfet you reccomend?
IRFB7437 or IRF1404 for Q6, Q7, Q8?
I ask you because I didn't research about it as much as you do.
Price is almost the same even if you look at ebay

[J4bberwock]

I thought to use heatsink like this
http://www.susan-parker.co.uk/zeus-heatsink-sk158-1a.jpg

about legs I will bend it to work it out or even solder some hard wire to it to be long enough.

so which mosfet you reccomend?
IRFB7437 or IRF1404 for Q6, Q7, Q8?
I ask you because I didn't research about it as much as you do.
Price is almost the same even if you look at ebay

IRFB7437 works, 1404 is ordered, but if you can have both at the same price, take the one that works for now.

I'll test the blade with IRFB7437 and 47K resistor in a day or 2

[Blisk]

ok let me know please in fev days I also will get those and will modify my blades.
So what you change is those mosfets, resistor and ferrite?
And direct wires to powering blades?
Is there anything else?
instead of resistor I will put on this potentiometer
http://www.rapidonline.com/electronic-components/20r-wr3006p-3-4-multi-turn-cermet-pot-68-0266

it is precise one I will use about 68Kohm

can you advise me which part to switch off and blade won't mine anymore and it is low power so I can use dip switch to stop mining?

[J4bberwock]

can you advise me which part to switch off and blade won't mine anymore and it is low power so I can use dip switch to stop mining?

I'd simply use the reset switch for the USB
It should be enough

A switch to disconnect your resistor from the board should also work.

Up to 50K, you should be fine if you use screw terminals (easier than direct wiring), I've already tested it.
47K with the standard connector, and the connector that plugs to the board have melted in a matter of minutes.
Up to 50k, the worst thing I've seen is the board stopping hashing after 1 or 2 minutes because the VRM went into heat protection and cut the line until they cool down.

[Blisk]

Ok, thanks, when you make it please let me know and I will also try what I intent to do and let you know how it works.

[Xer0]

Why than they use two  IRF5300
and one IRFH5301?

this.
should be cleared before we happily throw random chips in

[GenTarkin]

So, I pulled off the bridge mod w/ a tube of electric paint from RS for 8$ ....
Did it on 2x Gblades
Results are 875mhz ... 3 HW in 40mins @ vardiff 256  ... total between the 2 gblades

I would think thats a decent HW rate, less than .2% or so...

The next thing I would like to figure out is a way to overcome the insane voltage drop measured from the first yellow SM cap( in the VRM area) to the second SM cap(in the ASIC area).
On both my gblades there is a drop of .05v between these 2 sections alone.
The drop from the 2nd SM cap to the 4th SM cap(last one in ASIC area) is not nearly as high a drop.

Im wondering, for any of u electronics gurus out there, would soldering a 20 AWG wire from the - / + of the first yellow SM cap in VRM area to the respective legs on the 3rd SM camp(in middle of ASIC area), provide a more stable current flow between the sections of the blade? ... hence resulting in less voltage loss?

Ive seen a mod similar to this performed on avalon gen1 modules. It helped significantly.

Also, Im gonna guestimate that this is why we cant get the same clocks @ same voltages compared to the pods(@ same HW rate).
I highly doubt the pods would have this crazy voltage drop due to their size and layout and small # of chips.

If anyone could chime in on this that would be wonderful! Im dying to try it but I dont wanna destroy my blade....

[J4bberwock]

Why than they use two  IRF5300
and one IRFH5301?

this.
should be cleared before we happily throw random chips in

I can't see any reason for choosing 2 different ones, but one is for a channel, the second one is for a different channel.
Original design used 2 per channel and the same ones
UG and LG going to pin 2 and pin 19 of up1509

All fets were named 09N03 in the schematics.

here is the datasheet for it
http://www.infineon.com/dgdl/IPD09N03LA_Rev2.1_G.pdf?folderId=db3a304412b407950112b408e8c90004&fileId=db3a304412b407950112b429dca44214

On the 5 chips pod, the design shows only 1 for UG (pin2).

So we have
Upper gate, pin 19 -> 5302 or CSD16321
Lower gate, pin 2 -> dual 5300 or dual CSD 17556

We should be able to use the same fets everywhere.

[xcooling]

ive done direct wiring on motherboards to fix vdroop and stabilise memory. So in theory, your idea could work.

Anyone up to test it ?

Remember to remove the 2 ferrite beads, and replace them with thick solder. Otherwise you will damage your blade.

[racebyu]

ive done direct wiring on motherboards to fix vdroop and stabilise memory. So in theory, your idea could work.

Anyone up to test it ?

Remember to remove the 2 ferrite beads, and replace them with thick solder. Otherwise you will damage your blade.

Anyone have a picture of the removed and bypassed ferrite beads?
I assume its the rectangular part beside the two power fets.

[racebyu]

Another idea for best cooling of Q6, Q7, Q8 is position your Blade so the circuit boards are vertical.
Mount 2 L brackets on the existing heat sink as there are holes on the ends for self taping screws.
Place a flat heat sink across the two L brackets 1/2 way between the Blade circuit boards and the
outside of the heat sink (Best airflow for cooling).
Mount Q6, Q7, Q8 on the heat sink (not sure of TO220 tabs are common?) if not film mica isolators
are needed and isolation hardware.
Run wires from Q6, Q7, Q8  other side of board where surface mounted chips where originally located.
This should get the best cooling to the FET's with the massive airflow of the current fan.

Revised quick picture below.

[racebyu]

If the TO220 tabs have to be isolated you can use Mica Insulators + hardware or make the heat sink out of plastic.
Then add 3 finned TO220 individual heat sinks and space them out more, then the back tabs are isolated.

Top view of heat sink.



TO220 Finned Individual Heat Sink.


Winged Larger suface area.

[Blisk]

this way you close fans path at the end and mosfets are really alot far away from soldering pins.
and i plan to use way larger heatsink to coll down mosfets, this what you suggest is a question if it will be enough.

[GenTarkin]

ive done direct wiring on motherboards to fix vdroop and stabilise memory. So in theory, your idea could work.

Anyone up to test it ?

Remember to remove the 2 ferrite beads, and replace them with thick solder. Otherwise you will damage your blade.

Why do they have to be removed? just curious. Mine seem to be running warm, but not scorchingly hot. Also, I would like a pic of the ones being talked about. The hottest ones seem to be next to the side of the DC jack. Do they just overheat? if so, couldnt putting little heatsinks on them / backside help?

Do they have to be removed when just doing the bridge mod?

BTW, update, over the course of 11hrs, the bridge mod is going well.... I have them at 875mhz , .04% HW rate!

[Blisk]

some people make it to run on over 1000 Mh and than it makes them blown

[J4bberwock]

some people make it to run on over 1000 Mh and than it makes them blown

ferrite are still alive at 1063Mhz for weeks on my first blade, but I have forced cooling on the powerboard area.

Here is the newer version from Gridseed with ferrite replaced by 0 resistors and the vrm replaced by gridseed too

[GenTarkin]

oh! mine has the newer AQ6 vrm as well, wonder if the FB's are gone on mine... lets see...
hrm... I dont show a 0 on top of mine, just blank... guessing they are still FB's then...

I have little tiny 6x6mm heatsinks, could I put them on the FB's safely?

Or would it risk shorting the contact points on the FB's?

Im wondering, could I just short the tops of the ferrite beads ... to achieve same thing as removing & shorting the pads?

[J4bberwock]

oh! mine has the newer AQ6 vrm as well, wonder if the FB's are gone on mine... lets see...
hrm... I dont show a 0 on top of mine, just blank... guessing they are still FB's then...

I have little tiny 6x6mm heatsinks, could I put them on the FB's safely?

Or would it risk shorting the contact points on the FB's?

Im wondering, could I just short the tops of the ferrite beads ... to achieve same thing as removing & shorting the pads?

Do you have CSD 17556 for the other ones too or still the old IRFH ones?

[GenTarkin]

oh! mine has the newer AQ6 vrm as well, wonder if the FB's are gone on mine... lets see...
hrm... I dont show a 0 on top of mine, just blank... guessing they are still FB's then...

I have little tiny 6x6mm heatsinks, could I put them on the FB's safely?

Or would it risk shorting the contact points on the FB's?

Im wondering, could I just short the tops of the ferrite beads ... to achieve same thing as removing & shorting the pads?

Do you have CSD 17556 for the other ones too or still the old IRFH ones?

Hrm... dont recall, they are under heatsinks now =/