Pages:
Author

Topic: GRIDSEED G-BLADE Overclocking 7Mh/s, improvements and repair - page 10. (Read 74040 times)

sr. member
Activity: 805
Merit: 250
sr. member
Activity: 805
Merit: 250
joint seems to be good, i get the same value as if i'm measuring the resistor it self.

i sent you a PM btw J4
hero member
Activity: 840
Merit: 1000
On the 4th picture, you see the alternate place for the resistor. You can measure resistance between the 2 pads to check if your joint is good.
https://bitcointalksearch.org/topic/m.6545543
hero member
Activity: 840
Merit: 1000
Well, i'm at a lost, i modded one of my blades with a 39k resistor, and it was working fine for a few hours, i didn;t oc it, was running it at 850, than it stopped hashing, i redid the solder joints, and worked again a bit, but it stopped hashing again, but this time no success, i tried another resistor, no hash and getting HW errors, i even reput the original resistor, ( i add managed to save it. )
and still no hashing, but only getting HW errors, i took a look at the board and nothing seem to be blown or out of place.

Any ideas .?



Have you made the cooling improvements?

Anyway, I suspect a bad soldering. High quality picture would help.

Did you solder with good solder paste/flux?
Clean with isopropyl alcohol around the resistor if this isn't No clean solder flux

Are you using fine tip iron or hot air station?

Try to check the resistor value by measuring it from the alternate soldering place I showed in one of my first pictures.
If it isn't stable or not 39k, you failed the soldering job. Use good quality solder paste. I've made a post with everything I used.

Using 850 clock will drain the same watt than overclocking once you have replaced the resistor. Go up to 975 with all the extra cooking.
sr. member
Activity: 805
Merit: 250
Well, i'm at a lost, i modded one of my blades with a 39k resistor, and it was working fine for a few hours, i didn;t oc it, was running it at 850, than it stopped hashing, i redid the solder joints, and worked again a bit, but it stopped hashing again, but this time no success, i tried another resistor, no hash and getting HW errors, i even reput the original resistor, ( i add managed to save it. )
and still no hashing, but only getting HW errors, i took a look at the board and nothing seem to be blown or out of place.

Any ideas .?

hero member
Activity: 840
Merit: 1000
I've got three modded blades, and I can't seem to get them above 6.1Mh/s each poolside when using clevermining.

They are all at 963Mhz, but I can toggle with the speeds through hashra--anyone have any tips and tricks to squeeze a bit more out fo these guys?

If they are from zoomhash with a single heatsink on the coil, the resistor should be 37.5k. I just repaired one for someone locally that had burnt upper gate mosfet.
It's stable at 950 Mhz, 963 should be fine.

but please, if it is this one, and before it fails, add a nice heatsink 37x37mm around 20mm high with arctic ceramique under the PCB where the mosfets are
You can also add tall custom heatsinks on top of the mosfets with some thermal glue. The heatsink needs to be around 15mm high so it will catch the airflow from the fan.

I'll post a picture of this tall heatsink mod soon.
member
Activity: 70
Merit: 10
What value resistor did you mod with? If you went 39k ohm, you probably won't see much higher than 988mhz.
full member
Activity: 238
Merit: 100
We must become the pitiless censors of ourselves.
I've got three modded blades, and I can't seem to get them above 6.1Mh/s each poolside when using clevermining.

They are all at 963Mhz, but I can toggle with the speeds through hashra--anyone have any tips and tricks to squeeze a bit more out fo these guys?
full member
Activity: 161
Merit: 100
Does anyone have a good working bfgminer.conf or CGminer.conf they wish to share for stock G-blades that'll work under Windows?
Thanks.

...But this works in a .bat file:
bfgminer.exe -S noauto -S gridseed:all --scrypt -o stratum+tcp://gridseed.wemineltc.com:3333 -u testworker.1 -p 123 --failover-only -o stratum+tcp://stratum.scryptguild.com:3333 -u testworker.1 -p 123.  But this won't hash above 8MH/s.

My problem is that they don't go above 3.5MH/s with this string even though they are supposed to hit a minimum of 5.2MH/s.  Not seeing this.  Getting a top hashing of 6-7MH/s with both G-blades.

Under my Raspberry Pi Model-B I get 11-12Mh/s, minimal HW errors but the hashing on there last about 30 minutes before it all comes to a crawling 720KH/s.  Needless to say which is why I'm trying this under windows 7 SP1 now.  I tried getting them to work with Multiminer v3.2.2 but the best I got was also between 5-6MH/s.
So now I'm trying to get this to work from the Window CMD prompt.
Once I can past the basic stuff working stable, then I want to get to mod the crap out of these G-blades.


So that's why I was asking if I could get someone to display either their BFGminer.conf file, preferably under v4.2.0 or their CGminer.

Here's my .bat file

If you don't specify frequencies bfg miner will default to 600... hence you're only getting ~2MH per blade.

Code:
bfgminer --scrypt -o pool URL -u Username -p password --api-listen -S noauto -S gsd:all  --set-device gridseed0:clock=825 --set-device gridseed1:clock=825 --set-device gridseed2:clock=825 --set-device gridseed3:clock=825

my problem with bfg miner is I can't figure out how to specify frequency by serial number string like cpuminer Cpuminer fails for me after about an hour, even with fifo disabled in windows for those ports.

So I'm currently using jmordica's cgminer.  Anytime I have to restart or change settings, I have to run Zadig to get the blades to be detected again. Sometimes have to reset blades over and over and run zadig 5 times to get all my blades running.

I'm going to try this as soon as I get home. Thank you.
full member
Activity: 139
Merit: 100
hero member
Activity: 840
Merit: 1000
1404 also works as upper gate with 5300 as the lower. I'll try 7437 7430 and 1404 743x pairs later.
Higher rds on as upper gate.
sr. member
Activity: 412
Merit: 250
ok thanks, let me know if it works so I will mod it my blade
hero member
Activity: 840
Merit: 1000
I have now IRFB 7437 and IRF 1404 for lower and upper gates.
Will it work If I use it and also change ferrites, did you test it?
Irfb 7437 works for upper gate. I still haven't tried the 1404
I noticed that when switching to the dualcool MOSFET, Gridseed also added a resistor on the board. Probably to boost the signal for switching the gates.
I'll add some pics later and finally try to use 7437/7430/1404 on a pod.
sr. member
Activity: 412
Merit: 250
I have now IRFB 7437 and IRF 1404 for lower and upper gates.
Will it work If I use it and also change ferrites, did you test it?
hero member
Activity: 840
Merit: 1000
ok 2,7 MH is default blade and you get 3,1 out of one board. Isn't cheapper to just buy another gridseed and get that 400Khash more? Cheesy

On one of the panels, the ferrites melted the tracks so much it was impossible to recover.
Without external power, it wouldn't hash at all.
I know, but if you have a new blade, is it worthed?
voltmod with the resistor if you have a new one. and cool it correctly.
But external power supply is fun and can be used on other asics.
I can even power the Zeus Blizzard with a gridseed pod if I want.
sr. member
Activity: 412
Merit: 250
ok 2,7 MH is default blade and you get 3,1 out of one board. Isn't cheapper to just buy another gridseed and get that 400Khash more? Cheesy

On one of the panels, the ferrites melted the tracks so much it was impossible to recover.
Without external power, it wouldn't hash at all.
I know, but if you have a new blade, is it worthed?
hero member
Activity: 840
Merit: 1000
ok 2,7 MH is default blade and you get 3,1 out of one board. Isn't cheapper to just buy another gridseed and get that 400Khash more? Cheesy

On one of the panels, the ferrites melted the tracks so much it was impossible to recover.
Without external power, it wouldn't hash at all.
full member
Activity: 210
Merit: 100
sr. member
Activity: 412
Merit: 250
ok 2,7 MH is default blade and you get 3,1 out of one board. Isn't cheapper to just buy another gridseed and get that 400Khash more? Cheesy
hero member
Activity: 840
Merit: 1000
As requested, pictures of Frankenblade (copyright to me  Grin ) have been stable at 3.1MH/s, 925Mhz per panel for a few days with very low HW errors.

Piggybacked gridseed pods to get their 1.5 and 5V outputs to the blade.
Taking the 1.2v from where the big yellow capacitor was, routing it to to the blade on 2 of the above AC42 AC45 AC47 AC46.
Taking the 5v at the output of the up1707Q, connected on the blade to 2 off the 5 AME8805 feeding each line of 8 GC3355 chips
voltmodded 56K resistor on the pods gave me stable 1.35V on the blade.
There is a huge voltage drop, maybe also caused by my 0.75mm² wires, so a stock pod won't give enough voltage to hash at full speed with the blade. You should get something around 0.7-0.88v on the blade with a pod that isn't voltmodded.

I'm currently working on a standalone powerboard to power any asic PCB up to 300 watt (150A from 0.8v to 2v)

Fans on the pods put on the other side to also directly cool the top of the mosfets













Pages:
Jump to: