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

[DarkKnight]

Try to find replacaing chip for this one.
Here is datasheet.
http://filesave.me/file/39234/uP1509-DS-F0000-pdf.html

Jesus, do you know how many people have been looking for this datasheet all over the internet? Where did you find it? This part is used on GPUs as well, and that community has been looking for it too. As I understand it, the part is covered by an NDA, and is part of the reason the datasheet has been hard to find.

Back up link to Datasheet

Also: http://www.upi-semi.com/sales/sales.aspx to purchase a replacement. Order a sample, or contact one of the companies somewhere in Asia. I'm sure at least one of them will do business with you.

[Blisk]

What is it with you people?
I write to company and asked them polite if they can send me a datasheet and they did it with no question why.
All you need is search for manufacturer and ask them.

[J4bberwock]

Try to find replacaing chip for this one.
Here is datasheet.
http://filesave.me/file/39234/uP1509-DS-F0000-pdf.html

Jesus, do you know how many people have been looking for this datasheet all over the internet? Where did you find it? This part is used on GPUs as well, and that community has been looking for it too. As I understand it, the part is covered by an NDA, and is part of the reason the datasheet has been hard to find.

Back up link to Datasheet

Also: http://www.upi-semi.com/sales/sales.aspx to purchase a replacement. Order a sample, or contact one of the companies somewhere in Asia. I'm sure at least one of them will do business with you.

regarding the sample order, I tried, and never ever got a single answer from Upi. And I believe I was polite.

There is no direct replacement pin to pin compatible for up1509 unless some other UPI products will work.

You can still try to get one from a scrap pod, maybe even the small usb with one chip can carry the up1509 and you should be able to get them for almost free, I hope.

Lack of availability of the UP1509 is why I'm going the route of a custom beefier power board. using easier to source buck controllers, either TPS40090PW (it has legs, easier to check for good soldering) or TPS40140.
After registering on TI's site, I was able to order free samples delivered in 2 days by FEDEX.

I'm currently working on a first proto board for 8 GC3355 chips I got from failed pods.

[Blisk]

Ok than I am just lucky because I get it

[ZiG]

Ok than I am just lucky because I get it

Thanks for sharing, buddy...

ZiG

[Blisk]

Ok than I am just lucky because I get it

Thanks for sharing, buddy...

ZiG

No problem, glad to help

[wolfey2014]

Ok than I am just lucky because I get it

Thanks for sharing, buddy...

ZiG

Yeah, thanks for sharing, buddy....

WOLFEY2014

[kaltar]

could someone tell me exactly what resistor i need to buy to replace the 33k one, i know it`s a 39.9k but there is so many variants that i wan`t to make sure i get the right one. also could i use a axiel one instead of a smd, i would connect it with 24agw solid wire.

thanks

[J4bberwock]

could someone tell me exactly what resistor i need to buy to replace the 33k one, i know it`s a 39.9k but there is so many variants that i wan`t to make sure i get the right one. also could i use a axiel one instead of a smd, i would connect it with 24agw solid wire.

thanks

Hi,

the original ones are 0603 size.
You can use axial ones since the pads are "big", or 0805 size that is easier to solder than 0603 with a fine tip iron.
for 0805, I recommend using the alternate resistor location (that was meant for a parallel capacitor).
You can find this location on one of my first pictures, It's just over the original resistor on the board.

Replacing the power connector is a good idea too.


Last update, replacing ferrite with resistors might not be such a good idea.
On one of my blades, the power mosfet has shorted (as usual), but there were resistors.

The 3 ones for the +12V are blown, as is my 470uF 25V capacitor, and probably other components, since even after replacing the fet, resistors and capacitor, the blade isn't hashing. I'm currently tracking the damages and will report if it happens to someone else.

SMD Ferrite permanently damages the PCB when blowing becaue they are simply melting and they also melt the PCB board/tracks under them.

So for now, I'll try axial ferrite and/or fuses to protect the board.
Axial ferrite won't melt the tracks.

I'm also actively working on a replacement powerboard for the blade that could be piggybacked either to improve reliability, or easily repair a blade with damages to the power board.
I should come with a first prototype in a week or two, depending on how fast I can have all the last components I'm waiting for, and If I can etch the PCB myself or if I need to rely on a third party.

[DarkKnight]

Last update, replacing ferrite with resistors might not be such a good idea.
On one of my blades, the power mosfet has shorted (as usual), but there were resistors.

The 3 ones for the +12V are blown, as is my 470uF 25V capacitor, and probably other components, since even after replacing the fet, resistors and capacitor, the blade isn't hashing. I'm currently tracking the damages and will report if it happens to someone else.

SMD Ferrite permanently damages the PCB when blowing because they are simply melting and they also melt the PCB board/tracks under them.

What voltage/resistor combo were you running when this happened? Did you have active cooling on both sides of the ferrite bead? The 470uF 25v replaced the 220uF 16v?

Using a larger size cap causes somewhat higher current draw during it's charge phase, and may exacerbate the problem. If the charge phase is the same length of time and the capacity is double, it will pull twice as much current to charge it in the same period of time. You might perhaps be better off running (2) 110uF caps in parallel. Same capacitance, half current draw per cap & twice as much surface area for heat dissipation. That should cool them down considerably.

[Blisk]

better is 220uF and 35V or 2x 100uF

[J4bberwock]

Last update, replacing ferrite with resistors might not be such a good idea.
On one of my blades, the power mosfet has shorted (as usual), but there were resistors.

The 3 ones for the +12V are blown, as is my 470uF 25V capacitor, and probably other components, since even after replacing the fet, resistors and capacitor, the blade isn't hashing. I'm currently tracking the damages and will report if it happens to someone else.

SMD Ferrite permanently damages the PCB when blowing because they are simply melting and they also melt the PCB board/tracks under them.

What voltage/resistor combo were you running when this happened? Did you have active cooling on both sides of the ferrite bead? The 470uF 25v replaced the 220uF 16v?

Using a larger size cap causes somewhat higher current draw during it's charge phase, and may exacerbate the problem. If the charge phase is the same length of time and the capacity is double, it will pull twice as much current to charge it in the same period of time. You might perhaps be better off running (2) 110uF caps in parallel. Same capacitance, half current draw per cap & twice as much surface area for heat dissipation. That should cool them down considerably.

It was 43k stable for weeks, main fan failed during the night.
The ferrites aren't the direct culprit. the upper gate mosfet fails, shorting gate to drain, causing the ferrite to blow, if I'm not mistaken.
It also happens on the pods from time to time, so it shouldn't be the load.
upper gate mosfets should have been doubled as we can see on the initial schematic, but it saves a few $$ when manufacturing.

I suspect that with summer temperature, we will see much more blown mosfets on pods especially those overclocked with 5v fan mod.
When T° rises, the efficiency of the mosfets drops, causing more heat to be dissipated, causing the mosfet to short itself.

[Blisk]

that's why I will build in thermal switch if fan fails it will turn off mining
95 oC max

[J4bberwock]

that's why I will build in thermal switch if fan fails it will turn off mining
95 oC max

I'll do it too.

I managed to give a blade artificial breathing . pictures of Frankenblade soon.

A pod is actually providing the power to the blade and it's hashing, but I need to fix the voltage drop. From 1.3 it's dropping to 0.95 when hashing starts. 49.9k resistor user on the pods. 33k gives a voltage drop to 0,75.
BAUAR, can you check that you have the 5v output from the up1707 and 3.3v output from the ame8805?
My 5v was around 2,4v on one pcb. And output at 8805 was 0.1v.
There are a few pin to pin compatible replacement available for 8805 if it's fried. I haven't checked for up1707 yet,

If you can't track the defect, you can still use external power. Output from up1707 on a pods goes to large tab on any of the 5 8805 user for each row of chips. Disconnect the c42 capacitor from the pod and from the blade. Connect both "+" together and both "-" together.
I also removed up0111 from the pods, it gives avdd and dvdd pll, so I expect the chips from the pods won't draw any power.
 49k resistor will give 1v to the blade chips. I'll try 56k tomorrow.

If helps revive your blade, you can donate btc/ltc
If it doesn't, you can still send the blade to me for free

[Brassguy]

Just a question. I have 2 blades (4 boards) running at 838mhz. It's a good problem to have but they are only drawing about 200 watts total (~50w per board) at the wall... doesn't that seem low? The hashrate at the pool is ~11.2mh.

[J4bberwock]

Just a question. I have 2 blades (4 boards) running at 838mhz. It's a good problem to have but they are only drawing about 200 watts total (~50w per board) at the wall... doesn't that seem low? The hashrate at the pool is ~11.2mh.

I don't know how it's possible to draw 50w for 40 chips at this speed, but don't touch anything, let's say you are lucky.

[Brassguy]

Just a question. I have 2 blades (4 boards) running at 838mhz. It's a good problem to have but they are only drawing about 200 watts total (~50w per board) at the wall... doesn't that seem low? The hashrate at the pool is ~11.2mh.

I don't know how it's possible to draw 50w for 80 chips at this speed, but don't touch anything, let's say you are lucky.

edit... you meant 50w per 40 chips right?

It doesn't make any sense.... I have an undervolted antminer s1 using about ~215w and 2 (4 boards) blades on one psu... total usage 411 watts as we speak via my Kill-A-Watt... 11.4mh shown via cgminer 11.5mh showing at pool, so the hashrate is about the same locally and poolside...

edit    I'll investigate further before I overvolt/overclock these...

[J4bberwock]

Just a question. I have 2 blades (4 boards) running at 838mhz. It's a good problem to have but they are only drawing about 200 watts total (~50w per board) at the wall... doesn't that seem low? The hashrate at the pool is ~11.2mh.

I don't know how it's possible to draw 50w for 80 chips at this speed, but don't touch anything, let's say you are lucky.

edit... you meant 50w per 40 chips right?

It doesn't make any sense.... I have an undervolted antminer s1 using about ~215w and 2 (4 boards) blades on one psu... total usage 411 watts as we speak via my Kill-A-Watt... 11.4mh shown via cgminer 11.5mh showing at pool, so the hashrate is about the same locally and poolside...

edit    I'll investigate further before I overvolt/overclock these...

yes, 40 chips.

edited.

[gridsad]

Just a question. I have 2 blades (4 boards) running at 838mhz. It's a good problem to have but they are only drawing about 200 watts total (~50w per board) at the wall... doesn't that seem low? The hashrate at the pool is ~11.2mh.

my blade draw about that with seasonic psu .. measure at wall on killerwatt probably morelso 110 for the blade before killed it

[Brassguy]

More fun facts the upper fet is (AQ6) 16321C, lowers (AQ7 & AQ8) are CSD17556.