Pages:
Author

Topic: [GUIDE] GridSeed 5-Chip USB, Blade & Black Miner Support/Tuning - page 77. (Read 308807 times)

legendary
Activity: 3654
Merit: 8909
https://bpip.org
I'm selling 4 pin molex power connectors for gridseeds.  Multiple types available, see here:

https://bitcointalksearch.org/topic/m.5650838

What's the wire gauge in the 10-connector? You are dangerously close to 6A on the standard molex connector with 10 units as it is, and that splitter cable looks awfully thin. I wouldn't connect more than 3-4 miners via molex, and I wouldn't use this kind of splitter unless it is rated for 10+ amps. Is that black electric tape on the joints? Can't see well, the pictures are too blurry.

From what I understand, each GS node runs about 20W which is about 1.7A at 12V. Trying to hook-up 10 of them will draw about 17A.
Probably best to use an ATX supply, take the PCIe connectors and wire those up to the power plugs such that you won't draw more than the rated amps for each 12V rail in the PSU.
I even think someone makes a PCIe female to the 2.5mm (or 2.1mm) barrel plugs. Haven't found the vendor yet though I would be interested in that if anyone has a reference...

In scrypt-only mode it is 7W, so about 0.6A. There is a user "cablez" on these forums, who used to make custom PCIe-to-barrel wiring, but he seems to be overwhelmed with orders and is looking for more manufacturing capacity. I'm planning to make some myself, but I'm terrible at soldering, so will be using something with screws, like these: http://www.amazon.com/dp/B0060L5NM8
full member
Activity: 238
Merit: 100
If you want to run CGminer on Windows, this is your "Rasperry Pi" like Windows solution. These work great. I have 5 of them throughout my house for one reason or another. Not as cheap as RasPi, but not terrible either and very low power.

http://www.amazon.com/gp/product/B00B7I8HZ4?&linkCode=wsw&tag=dv07-20
legendary
Activity: 1109
Merit: 1000
I'm selling 4 pin molex power connectors for gridseeds.  Multiple types available, see here:

https://bitcointalksearch.org/topic/m.5650838

What's the wire gauge in the 10-connector? You are dangerously close to 6A on the standard molex connector with 10 units as it is, and that splitter cable looks awfully thin. I wouldn't connect more than 3-4 miners via molex, and I wouldn't use this kind of splitter unless it is rated for 10+ amps. Is that black electric tape on the joints? Can't see well, the pictures are too blurry.

From what I understand, each GS node runs about 20W which is about 1.7A at 12V. Trying to hook-up 10 of them will draw about 17A.
Probably best to use an ATX supply, take the PCIe connectors and wire those up to the power plugs such that you won't draw more than the rated amps for each 12V rail in the PSU.
I even think someone makes a PCIe female to the 2.5mm (or 2.1mm) barrel plugs. Haven't found the vendor yet though I would be interested in that if anyone has a reference...
sr. member
Activity: 462
Merit: 250
not sure but doubt someone pays that much for a spliced and taped up solution    Undecided
legendary
Activity: 3654
Merit: 8909
https://bpip.org
I'm selling 4 pin molex power connectors for gridseeds.  Multiple types available, see here:

https://bitcointalksearch.org/topic/m.5650838

What's the wire gauge in the 10-connector? You are dangerously close to 6A on the standard molex connector with 10 units as it is, and that splitter cable looks awfully thin. I wouldn't connect more than 3-4 miners via molex, and I wouldn't use this kind of splitter unless it is rated for 10+ amps. Is that black electric tape on the joints? Can't see well, the pictures are too blurry.
full member
Activity: 182
Merit: 100
cambrionix 49 port hub. Each port is isolated. $500. I think it is cheaper to get a bunch of RaspPi
http://www.ebay.com/itm/like/251380675091?lpid=82

why not any of those ?





it`s USB 3.0, rPi don`t like it

I want to order that 49 port hub but am cautious because idk if it will work.  Should but don't know.
sr. member
Activity: 434
Merit: 265
I'm selling 4 pin molex power connectors for gridseeds.  Multiple types available, see here:

https://bitcointalksearch.org/topic/m.5650838


nice work ... but price  Grin
full member
Activity: 230
Merit: 100
I'm selling 4 pin molex power connectors for gridseeds.  Multiple types available, see here:

https://bitcointalksearch.org/topic/m.5650838
sr. member
Activity: 840
Merit: 251
Okay so cgminer 3.8.5 is not working for me for sha 256 any ideas.  Huh


Figured it out in Windows 8.1 you need a double tap start cgminer 3.8.5 then start cgminer 3.8.5. leave the one open thats working.
legendary
Activity: 1855
Merit: 1016
anyone know how to dual mine using cgminer only?
sr. member
Activity: 378
Merit: 250
I've been having a problem with my miners disconnecting and just showing "dispatching new work toGC3355 LTC core" and nothing else. WTF???

It started happening after I pointed them to ScryptGuild.com
Before that, they had pretty much "yay" hashed full time without any interruptions.

Does anyone know if a pool server can cause this to happen? Make them quit running by perhaps, quit sending work to them for a few minutes which causes them to just - hang?

I don't think it's the USB hub causing it.

This happened 6 hours after I signed up with ScryptGuild, then another 2 hours after that, then another 18 hours after that. It's not the USB hub or lack of power causing it.

To get them running again just now, I have re-set all the ports using disable-enable which saves me from having to physically unplug and re-plug each one. I'm going to figure out a way to make this problem 'auto detect this failure and run disable-enable for each port automatically so they keep on hashing.

I have also disabled the power saving feature on each 'root hub' in device manager just to make sure WinBlow$ doesn't cause it to happen too.

I'm using Win7. I don't know if XP allows one to make these types of config changes or not.

I"m not interested in putting up with this disconnect shyt! I am sure no one else is either Wink

Perhaps we can work together to get it solved once and for all.

Wolfey2014

newbie
Activity: 10
Merit: 0
thanx for this guide  Cheesy
sr. member
Activity: 458
Merit: 250
beast at work
member
Activity: 105
Merit: 10
Has anyone figured out how to overvolt yet?  I see this in the gridseed driver...

typedef struct s_gridseed_info {
   enum sub_ident   ident;
   uint32_t   fw_version;
   struct timeval   scanhash_time;
   int      nonce_count[8];  // per chip
   int      error_count[8];  // per chip
   // options
   int      baud;
   int      freq;
   unsigned char   freq_cmd[8];
   int      chips; //chips per module
   int      voltage;
   int      per_chip_stats;
}
Aand later on this method:

static void gc3355_increase_voltage(struct cgpu_info *gridseed) {
   uint32_t reg_value;

   // Put GPIOA pin 5 into general function, 50 MHz output.
   if (!gc3355_read_register(gridseed, GRIDSEED_GPIOA_BASE + GRIDSEED_CRL_OFFSET, ®_value)) {
      applog(LOG_DEBUG, "Failed to read GPIOA CRL register from %i", gridseed->device_id);
      return;
   }
   reg_value = (reg_value & 0xff0fffff) | 0x00300000;
   if (!gc3355_write_register(gridseed, GRIDSEED_GPIOA_BASE + GRIDSEED_CRL_OFFSET, reg_value)) {
      applog(LOG_DEBUG, "Failed to write GPIOA CRL register from %i", gridseed->device_id);
      return;
   }

   // Set GPIOA pin 5 high.
   if (!gc3355_read_register(gridseed, GRIDSEED_GPIOA_BASE + GRIDSEED_ODR_OFFSET, ®_value)) {
      applog(LOG_DEBUG, "Failed to read GPIOA ODR register from %i", gridseed->device_id);
      return;
   }
   reg_value |= 0x00000020;
   //reg_value &= 0xFFFFFFDF;
   if (!gc3355_write_register(gridseed, GRIDSEED_GPIOA_BASE + GRIDSEED_ODR_OFFSET, reg_value)) {
      applog(LOG_DEBUG, "Failed to write GPIOA ODR register from %i", gridseed->device_id);
      return;
   }
}

Looks to me like we pass an int as the argument voltage and it will add it to the base voltage value(which i think is set to 0, if I'm reading the header right), maybe?  But then it also says gpio(general purpose input output?), which is usually a general purpose pin header on the pcb itself, right?  I'm no developer/hardware engineer, anyone want to comment?

You can set voltage=1 in your config, but it turns out that will actually lower the core voltage.

Did someone find a measurement point on the board to check?  What happens if we pass -1?  I'm surprised more people aren't asking about this, surely with as much cooling as these chips have on them we can get away with a bit more voltage in scrypt only mode, and hence, higher stable clocks!  I don't want to sacrifice a miner if I can avoid it by researching instead... Smiley

Yes, it should be possible to replace a resistor to increase the voltage. I haven't had a chance to try this yet.
Why replacing a resistor, do you remember the pencil mod on those bitfury miners ??   Wink
full member
Activity: 151
Merit: 100
http://cryptomining-blog.com/tag/gridseed-asic/

I have been scrypt mining with my gridseed for a couple days and wanted to try out dualmining so I followed the guide above. Everything went well and both instances open, but neither are actually doing any work. Usually others run into the same issues but after reading through everything I'm still lost. If anyone knows what's wrong please help a brother out.

legendary
Activity: 784
Merit: 1000
For those of you running Windows and CGminer for Gridseed ASICs:

Long time lurker, but new member here.  I usually just sit back and read to solve my problems. I registered here today just to make this post.

I noticed when using Gridseed 5 chip miners with Windows/CGminer that miners would randomly drop off. When you close CGminer or reboot, this was especially true. The fix? Unplug the USB cable to your miner or hub and set hotplug to a higher number than 0 (most seem to like 5).

Great, but I have a job/life/kids/wife/car/hobbies/etc and can't always be next to the computer waiting to unplug/plug a USB connector.

I did a little experimenting and found out that disabling/enabling them in device manager brought them back without having to physically unplug them. This got me thinking...

I was able to use devcon.exe (from Microsoft) and write a quick batch file that enables/disables the miners. Then I scheduled the batch file to run ever 60 seconds (maybe overkill). The good part is that devcon.exe does not disable devices if they are in use, it waits for a reboot. So, your active miners never go down they just get "flagged" to be disabled on next boot and your inactive miners get disabled/enabled and start working again. This also requires hotplug to be greater than 0 (I used --hotplug=10)

It's quick and dirty and by no means a great solution. But it works. If someone wants to take this and expand on it, be my guest.

To make sure its working:
Run CGMiner, make sure it works and miners start.
Close CGMiner.
Run CGMiner, no miners (or less miners) are found.
Close CGMiner.
Run batch found below.
Run CGMiner, all miners working again.

You may need to edit the batch file to include match your miners hardware ID (*5740* is wildcard-part_of_hardware_ID-wildcard). I'm not sure if all hardware IDs are all the same, but all mine are. I just shoved it in the root of my C:\ drive. It's very basic and self explanatory.

Zip contains:
devcon.exe
resetusb.bat
https://www.dropbox.com/s/ih0eszwbf6bucl4/resetusb.zip

I hope this helps some usb plugging, nail-biting, BCTer get out and grab a little sunlight instead of hiding in the basement waiting for that inevitable USB failure.  Grin

Nice!  I'll test this out, if it works you have a bit of coin coming your way Smiley
full member
Activity: 238
Merit: 100
For those of you running Windows and CGminer for Gridseed ASICs:

Long time lurker, but new member here.  I usually just sit back and read to solve my problems. I registered here today just to make this post.

I noticed when using Gridseed 5 chip miners with Windows/CGminer that miners would randomly drop off. When you close CGminer or reboot, this was especially true. The fix? Unplug the USB cable to your miner or hub and set hotplug to a higher number than 0 (most seem to like 5).

Great, but I have a job/life/kids/wife/car/hobbies/etc and can't always be next to the computer waiting to unplug/plug a USB connector.

I did a little experimenting and found out that disabling/enabling them in device manager brought them back without having to physically unplug them. This got me thinking...

I was able to use devcon.exe (from Microsoft) and write a quick batch file that enables/disables the miners. Then I scheduled the batch file to run ever 60 seconds (maybe overkill). The good part is that devcon.exe does not disable devices if they are in use, it waits for a reboot. So, your active miners never go down they just get "flagged" to be disabled on next boot and your inactive miners get disabled/enabled and start working again. This also requires hotplug to be greater than 0 (I used --hotplug=10)

It's quick and dirty and by no means a great solution. But it works. If someone wants to take this and expand on it, be my guest.

To make sure its working:
Run CGMiner, make sure it works and miners start.
Close CGMiner.
Run CGMiner, no miners (or less miners) are found.
Close CGMiner.
Run batch found below.
Run CGMiner, all miners working again.

You may need to edit the batch file to include match your miners hardware ID (*5740* is wildcard-part_of_hardware_ID-wildcard). I'm not sure if all hardware IDs are all the same, but all mine are. I just shoved it in the root of my C:\ drive. It's very basic and self explanatory.

Zip contains:
devcon.exe
resetusb.bat
https://www.dropbox.com/s/ih0eszwbf6bucl4/resetusb.zip

I hope this helps some usb plugging, nail-biting, BCTer get out and grab a little sunlight instead of hiding in the basement waiting for that inevitable USB failure.  Grin
member
Activity: 84
Merit: 14
Has anyone figured out how to overvolt yet?  I see this in the gridseed driver...

typedef struct s_gridseed_info {
   enum sub_ident   ident;
   uint32_t   fw_version;
   struct timeval   scanhash_time;
   int      nonce_count[8];  // per chip
   int      error_count[8];  // per chip
   // options
   int      baud;
   int      freq;
   unsigned char   freq_cmd[8];
   int      chips; //chips per module
   int      voltage;
   int      per_chip_stats;
}
Aand later on this method:

static void gc3355_increase_voltage(struct cgpu_info *gridseed) {
   uint32_t reg_value;

   // Put GPIOA pin 5 into general function, 50 MHz output.
   if (!gc3355_read_register(gridseed, GRIDSEED_GPIOA_BASE + GRIDSEED_CRL_OFFSET, ®_value)) {
      applog(LOG_DEBUG, "Failed to read GPIOA CRL register from %i", gridseed->device_id);
      return;
   }
   reg_value = (reg_value & 0xff0fffff) | 0x00300000;
   if (!gc3355_write_register(gridseed, GRIDSEED_GPIOA_BASE + GRIDSEED_CRL_OFFSET, reg_value)) {
      applog(LOG_DEBUG, "Failed to write GPIOA CRL register from %i", gridseed->device_id);
      return;
   }

   // Set GPIOA pin 5 high.
   if (!gc3355_read_register(gridseed, GRIDSEED_GPIOA_BASE + GRIDSEED_ODR_OFFSET, ®_value)) {
      applog(LOG_DEBUG, "Failed to read GPIOA ODR register from %i", gridseed->device_id);
      return;
   }
   reg_value |= 0x00000020;
   //reg_value &= 0xFFFFFFDF;
   if (!gc3355_write_register(gridseed, GRIDSEED_GPIOA_BASE + GRIDSEED_ODR_OFFSET, reg_value)) {
      applog(LOG_DEBUG, "Failed to write GPIOA ODR register from %i", gridseed->device_id);
      return;
   }
}

Looks to me like we pass an int as the argument voltage and it will add it to the base voltage value(which i think is set to 0, if I'm reading the header right), maybe?  But then it also says gpio(general purpose input output?), which is usually a general purpose pin header on the pcb itself, right?  I'm no developer/hardware engineer, anyone want to comment?

You can set voltage=1 in your config, but it turns out that will actually lower the core voltage.

Did someone find a measurement point on the board to check?  What happens if we pass -1?  I'm surprised more people aren't asking about this, surely with as much cooling as these chips have on them we can get away with a bit more voltage in scrypt only mode, and hence, higher stable clocks!  I don't want to sacrifice a miner if I can avoid it by researching instead... Smiley

Yes, it should be possible to replace a resistor to increase the voltage. I haven't had a chance to try this yet.
newbie
Activity: 39
Merit: 0
Do you know of any hubs that keep all the ports on a single level?  I already purchased and installed a 2nd RaspPi and have 20 miners on each, but in the future it might be helpful for others that want to control more from a single unit (which I bet it can since 20 miners only results in around 10% utilization).

Probably too expensive but here is a 49 port designed for continuous enterprise use:

How about 2 of these?:  http://www.monoprice.com/Product?c_id=103&cp_id=10307&cs_id=1030702&p_id=6436&seq=1&format=2

One for each of the Raspberry Pi's 2 ports?  We should be able to conceivably connect up to 48 miners per RPi...  Anyone had any experience with these or know its internal topology?

Or you could try two of these? I don't like the port layout though:
http://www.microcenter.com/product/393329/28-Port_MondoHub_SuperSpeed_USB_30_Hub

The one from monoprice looks nicer.

Yeah, and a bit spendy (could get 2 monoprice ones for price of 1 MondoHub), plus I could have sworn i've read in one of the many threads floating around that the gridseed miners don't play well with USB 3.0 devices.  Unfortunately the price on the hubs is roughly the same as a RPI so there is little incentive to not just run 2x 10-port cheap hubs per Pi and scale accordingly.

-Eric
Pages:
Jump to: