Topic: GRIDSEED G-BLADE Overclocking 7Mh/s, improvements and repair (Read 74129 times)

Nope - write protected, first thing the programming software asks to do is unlock the flash and erase the chip...

Hi - I've managed to destroy one of my g-blades in a manner that I can't easily repair.
I had them on a raspberry pi all powered by a battery / solar set up.
It looks like the power jack on one failed, opening the negative, causing the return current to flow through the RFI bead to the STM32f103 ARM controller ground, down the USB cables to the pi, through the pi ground and back to the solar controller ground.
The pi survived.
Along the way the ARM and 3.3V regulator were destroyed - as in escape of magic smoke and 3.3V shorted to ground.
There must have been over 5 V of drop on the ground there to reverse the polarity on the ARM and regulator to destroy them...
When I power up on 12V, I still have the 1.2 V rail, the 5V rail derived from the 12 V rail, and the 5x 3.3 v rails.
Fingers crossed the GC3355's are all ok. 
I've done some probing tracing of the board to understand how the GC3355's are connected to the ARM.  Looks like the only connects are to the ARM's 5 UARTs and some signals to the uP1509 that genrates the 1.2V

I think I'll replace all the 0 ohm links between the GC3355's and the ARM's UART ports with ~100 ohm resistors to limit the current that can flow into the inputs under fault conditions and put a schottky diode reverse biased  across the 3.3V rail so the power rails across the ARM can't become reversed.

I have a tech background and am able to replace the STM32f103RCT6 ARM controller, but I don't have any firmware for it.   
Back on page 29, I see there was a firmware image from" J4bberwock" but this is now a dead link.
Does any one have a copy of the g-blade firmware any more?
I am familiar with Atmel AVR / Arduino and PIC but have no prior ARM experience.
I'm assuming Gridseed shipped these with flash read protection active?
Surely its not as easy as reading out the flash from the good ARM on the other side of the brick into the programing software and writing it out again to the new blank controller?

Regards,
Nick.

 

On page 22 of this thread there is a great couple of pics of fault diagnosis by j4bberwock. If you've read a bit of this thread you'll get that the most common failure of these boards is the mosfets in the top right of those pics I mentioned overheating.

If you can prod about with a multimeter yourself and establish for sure if it is the mosfets then you can source the parts in my post (2 up) and either fix it yourself or take it to a proper electronics repair guy and they should have it done in few minutes. There's a chance it may not be the mosfets though and it still won't work.

I did mine myself and took that chance.

When I repaired mine, I also bought from china a new hashing board in case I wasn't able to repair my first. I managed to get it for £15 with free shipping to UK. My parts for repair came to just over £10. So I'm not sure if it was really worth it apart from the experience.

For some reason, checking again the price of the boards has jumped again to about £35, not sure why. Perhaps it'll fall again, I don't imagine they get many customers at that price.
https://www.aliexpress.com/wholesale?catId=0&initiative_id=SB_20170904022341&SearchText=gridseed+blade.

So a couple of options there. But either way I would take a look at the fault finding to help weigh up your options.

Hi, recently setup my G Blade again as I had the opportunity of free electric.

Was running fine for a week and then stopped accepting work (still showing hashrate though). When I checked on the unit, the power brick had failed on the board powering the fan, so boards probably overheated.

Is anyone able to check and repair my boards? I'm in the UK.

Thanks

One of my blades started spitting out thousands of hardware errors per second at the stock 600 MHz clock so i did the usual test points to see if there was a fault on the board (https://bitcointalksearch.org/topic/m.7330855) and there was not fault, everything was fine, i don't get it, why did the board suddenly start having so many hardware errors, does anyone have an idea of what's going on?

Hi all, I know this is really a bit of a dead topic but this may be useful for hobbyists still?

So I got one of these units as I got interested in the crypto scene about a year ago with a couple of antminer s3 on the cheep. All was well for about a year until one of the blades started getting too hot, it smelt hot, like a burning. Anyway of all the components the choke seemed the hottest and had dulled in colour along with an area of the pcb around it. Thanks to this thread I've got it going again. It however took quite a few hours of trawling to get all the info I needed. A bulk of the info was on page 21 and 22 of this thread and couldn't have done it without that. So thanks to @J4bberwock for the fault diagnosing voltages and the components lists.

So I've used those and a new choke, I'll list them below along with the store I bought them from, as again, it took a while of trawling to find a store that had all the bits to only pay delivery once. As these are old I assume repair could be on a few peoples minds, hope this helps someone.

I'm in the UK so all in £

Upper gate http://uk.farnell.com/texas-instruments/csd16321q5c/mosfet-n-ch-25v-31a-8son/dp/1791476 £1.42

Lower gate http://uk.farnell.com/texas-instruments/csd17556q5b/mosfet-n-ch-30v-100a-vson-8/dp/2501093 £1.89

Choke http://uk.farnell.com/coiltronics/dr127-1r0-r/inductor-smd-1uh-15-5a/dp/2075607 £1.84

Capacitor http://uk.farnell.com/kemet/a767mu227m1vlae028/cap-alu-polymer-220uf-35v-rad/dp/2614188 £1.63

So with the 2x upper and 2x lower, choke and capacitor replaced, all in £10.03 for parts. I went for the repair to teach myself smd soldering and to get a bit more knowledge on these low voltage and high current circuits which seem to be used in all miners.

And the final assembly...quite pleased with myself...

ok

No, 3.5 per plate, 7.0 total.  Put another way, about 40% above stock.

ok
That 7mh / s per plate?

This is a scrypt miner, not an sha256 miner.  It will mine Litecoin and variants, not Bitcoin.

suport sha256?

Does this board make ash 256?
How much of the income in dollars?

https://www.youtube.com/watch?v=AweqF4AmGwc

First the symptoms...

• Blade 'E9' 34 hash in test period
• Blade 'DD' 28 hash in test period
• Blade 'DB' 15 hash in test period
• Blade 'EB'  8 hash in test period

Something is really wrong with  EB & DB, DD is suspect.
I'm not intending to overclock, just make the ***** things work.

QUESTION:

• What is the tolerance on the supply voltages expected with USB disconnected, and 12V supply plugged in ?

I assume lower voltage = less hash is the root cause.

• What determines the output voltage of AU16..20?

I assume it's meant to be fixed 3.3V out for 5V in and they have degraded over time.

• What should be the current rating of AU16..20?

I can't find a spec for the A8805, assuming 60W for 40 ICs then 1A ought to be sufficient.

• What should be the pinout of AU16..20?

I can't find a spec for the A8805, wild guess is 1-input, 2-ground,3-output,4-input.

• What determines the voltage on AC42,45,47,48 ?

I assume I'm seeing voltage drop across the thin traces on the board.
[/list]

Hardware:
AQ6 IR5802
AQ7 IR5300
AQ8 IR5300

AU5,16..20 A8805


Measurements:
Blade 'DB'
AC42 1.199V
AC45 1.187V
AC47 1.158V
AC18 1.183V

AU5 1.511V
(I think that's because USB is not connected)

AU16..20 5.64V on tab
AU16 3.27V
AU17 3.26V
AU18 3.25V
AU19 3.23V
AU20 3.28V

J21-4 3.23V
J21-5 3.28V


Blade 'DB'
AC42 1.525V
AC45 1.161V
AC47 1.157V
AC18 1.157V

AU5 1.525V
(I think that's because USB is not connected)

AU16-20 5.62V on tab
AU16 3.25V
AU17 3.24V
AU18 3.24V
AU19 3.29V
AU20 3.28V

J21-4 3.28V
J21-5 3.28V

That's probably the fan connection.
Overclocking is just a matter of setting up the right string in the configuration for your miner - but IMO it's a VERY VERY bad idea unless you have a very COLD room to run them in, the PS circuitry for them is very very marginal even at their "stock" clock - and realistically, that "stock" clock was a noticeable overclock to start with compared to the GC3355 "orb" units Gridseed made before they started making the GC3355 "blade" units.

 The mining chip ITSELF has plenty of headroom, it's the buck convertor stuff that gets easily overloaded and dies. Helps a LOT to put heatsinks on the critical parts, though.

Hi,

I have few questions on those gridseed 80 chip blade minesr:

1) Upon seeing product, I noted there red and black wire connected to one blade and the other don't
Is there a reason for that or is it recommend connecting another similar wire on other blade?

2) I read several comments on overclocking those units:
How is this possible on my part?
Is overclocking good or bad by using BFgimer for example

Again, It'll be great knowing some answers before using those units

Thanks

Yes, they're very linear on power/hash when you downclock them - and since the "stock" clock is actually a fairly hefty OVERCLOCK in reality, it's a GOOD IDEA.

Have anyone underclock these miners?

Zeus or Gridseed vs. A2 based gear though is an even more impressive win for the A2.

Zeus was too inefficient when it was released to be competative, the ONLY thing it was better than was the GC3355-based gear - and that was more because Gridseed's board-level design skills were horrible, and their insistance on their stupid "dual mining" concept, than anything GOOD about the Zeus designs.


 Who the heck is / was "Joe Bloggs"?


 Also, your figures on the Gridseed blades are off based on the ones I had. I was routinely seeing under 100 watts for mine at anything up to about 5 Mhash, a bit over (110 perhaps) at 5.4

 Dunno why you are talking about burnout on USB ports, the gridseeds did NOT soak a significant amount of power from their USB connections - and their power ground was via that shitty barrel connector for 12VDC, not via a USB port that had ZERO connection to the +12vdc circuit (unless you were powering the blade from the same PS as you were running the miner software on).


 I never saw an issue of the Gridseeds struggling to get through a block - you must not have been on a well-designed pool or had them badly misconfigured to see that sort of issue.


 The REAL issue with the Gridseed blades was the shit design on the power circutry - if they'd had a decent design on that, they'd have been quite reliable.

Reading upwards regarding Zeus vs Gridseed Blade argument.

Unequivocally, Zeus takes the podium here.
I've had the pleasure of dealing with both of these units, Zeusminer 7Mh blades and Gridseed 5.4Mh blades.

While the Gridseed Blades actually hash correctly, they are an absolute nuisance otherwise.
Issues with ferrite beads, with the standard PSU's running uncomfortably hot (not that you'd use them, but think about Joe Bloggs), the choice of a barrel power plug for 70W+ per pcb, major hotspots on the pcb's and the fact that when you release a product to the public - you DO NOT EXPOSE PCB's.
Not to mention you only need to look sideways at these units and they stop working.

The only good point about these units is perhaps the overclockability of the GC3355 IC's themselves; however the previous points hinder any improvement you'd make.

As for the Zeus; albeit being noisy and using 28% more power (Gridseed 5.4Mh/150w, Zeus 7Mh/250w), it's a delight.
Not having to worry about burning out your USB ports, even after investing in a heavily powered USB hub, is fantastic as well.
Or the Gridseed 'syndrome' of having the unit ground it's powersupply THROUGH THE USB CABLE (people will pretend this has never happened, because basically people are shit and also because electricity always follows path of least resistance), either due to shoddy cablework or a cable not being pressed in *just* enough.

Onto performance, having the entire Zeus show as one single device means this thing chews through HUGE blocks, like your mum after doing food shopping on welfare day. The amount of times i've watched the Gridseeds struggle to spit one share out, then find the stratum has already announced a new block and all that work for nothing.

Looking past others anecdotes of overheating and other issues, i'll throw my own in for fair share.

Having setup my Zeus Thunder X3 with two Corsair PSU's (one later found to be a bit dodgy), left them running all day at 342Mhz (from 328 standard) at approximately 32-34Mh. Only had the fans wired to one PSU; which happened to be the dodgy one (woops). Noticed around 2PM (whilst at work) that my hashrate had dropped right off on Nicehash.
Got home to find the fans had stopped and two of the blades had been sitting at upwards of 80 degrees for SEVERAL HOURS.
Like the troopers they were, replaced the PSU, turned it back on, no hassle. Like Zeus himself.

If these units were Gridseeds, there would be a puddle of tears on the floor and someones face buried in a pillow (interpret that however you like). Gridseeds do an ok job if you can pick them up cheap enough and micromanage them; better than a GPU i guess.

I've a Zeus Blade and 6 Gridseed Blades; guess which ones give me the most hassle?