[CLOSED] Ninja Group Buy #5 - QTY 8 Gridseed LA6M 48Mhash Scrypt [0/160]

[daemonfox]

We are using the lightning asic tp-link controllers and a remote power switch to power cycle when the USB bus hangs.  Does a PC platform give you much of an advantage performance wise or stability wise?  It looks like the tuning options are better, but how is the stability?  The tp-link boots up pretty quick, I'd hate to have to keep cold-starting a server/pc to recover the USB bus. I guess we could probably root and install this on our TP links without too much difficulty.  

We're going to have to put in a couple of servers to do the pool right.  The front end and stratums will be on 2 different VM's on one server in the DMZ, and the database and coindaemons will be on the other on our LAN.

so quick question... you on the modded cpuminer with the power fix that makes these ~8W per unit on scrypt?

Also, I figure you are on Linux so... here ya go...

https://github.com/dtbartle/cgminer-gc3355

nice cgminer 3.7.2 with gridseed support and full stats minus temps (guess there is not a sensor?) and has the same power fix so ~8W per unit.

I am not using a USB hub so maybe it is mostly a hub thing... but I have yet to have a single hang up once I was hashing away in cgminer... but again I have a single unit not a whole stack of them.

My understanding from other posts thus far is people are struggling with your hang up issue because the included hub is shit... SHIT. Apparently... it may not be providing enough STABLE power and when one device goes out, the whole hub has to be restarted. I DID experience something like this in my initial testing playing with the miner applications... the USB must be unplugged and plugged back in (or power cycled) in order for the miner to be picked up again... this arose when I was playing with frequencies to get the best rate. The reason you have to power cycle the HUB is that the board itself on each device draws ZERO power from the PSU itself... the damn things will mine on USB power alone so any fluctuation in temp and hashrate can cause a power spike... overdrawing on the hub... causing the hub to fail and require the disconnect and reconnect of the device before hashing will start.

Some have chosen much higher quality hubs with better peak power performance which seems to resolve these issues. So far since I found my best settings... I was dual mining on Win 7 64 bit with my unit... zero hang ups... 7 GH/s SHA256 and 330 KH/s scrypt. I do not have this on my KILL-A-WATT yet but when I get up in the AM and find I have had no stability issues with the cgminer 3.7.2 for scrypt, I will start them again on it to get a power draw in dual mode. Dual requires one cgminer 3.8.5 miner for SHA256 and then a second minerd launched with the --dual flag for scrypt.

FYI I was able to push mine to 1000 clock in cgminer but didn't like how slowly shares came in... 900 settled back to a good rate and has me at 383 KH/s in cgminer. These little work horses have a TON of headroom... the firmware is sadly very flimsy and can't handle the big guns yet... time will tell if we can double their performance.

[dyland]

We are using the lightning asic tp-link controllers and a remote power switch to power cycle when the USB bus hangs.  Does a PC platform give you much of an advantage performance wise or stability wise?  It looks like the tuning options are better, but how is the stability?  The tp-link boots up pretty quick, I'd hate to have to keep cold-starting a server/pc to recover the USB bus. I guess we could probably root and install this on our TP links without too much difficulty.  

We're going to have to put in a couple of servers to do the pool right.  The front end and stratums will be on 2 different VM's on one server in the DMZ, and the database and coindaemons will be on the other on our LAN.

so quick question... you on the modded cpuminer with the power fix that makes these ~8W per unit on scrypt?

Also, I figure you are on Linux so... here ya go...

https://github.com/dtbartle/cgminer-gc3355

nice cgminer 3.7.2 with gridseed support and full stats minus temps (guess there is not a sensor?) and has the same power fix so ~8W per unit.

I am not using a USB hub so maybe it is mostly a hub thing... but I have yet to have a single hang up once I was hashing away in cgminer... but again I have a single unit not a whole stack of them.

My understanding from other posts thus far is people are struggling with your hang up issue because the included hub is shit... SHIT. Apparently... it may not be providing enough STABLE power and when one device goes out, the whole hub has to be restarted. I DID experience something like this in my initial testing playing with the miner applications... the USB must be unplugged and plugged back in (or power cycled) in order for the miner to be picked up again... this arose when I was playing with frequencies to get the best rate. The reason you have to power cycle the HUB is that the board itself on each device draws ZERO power from the PSU itself... the damn things will mine on USB power alone so any fluctuation in temp and hashrate can cause a power spike... overdrawing on the hub... causing the hub to fail and require the disconnect and reconnect of the device before hashing will start.

Some have chosen much higher quality hubs with better peak power performance which seems to resolve these issues. So far since I found my best settings... I was dual mining on Win 7 64 bit with my unit... zero hang ups... 7 GH/s SHA256 and 330 KH/s scrypt. I do not have this on my KILL-A-WATT yet but when I get up in the AM and find I have had no stability issues with the cgminer 3.7.2 for scrypt, I will start them again on it to get a power draw in dual mode. Dual requires one cgminer 3.8.5 miner for SHA256 and then a second minerd launched with the --dual flag for scrypt.

FYI I was able to push mine to 1000 clock in cgminer but didn't like how slowly shares came in... 900 settled back to a good rate and has me at 383 KH/s in cgminer. These little work horses have a TON of headroom... the firmware is sadly very flimsy and can't handle the big guns yet... time will tell if we can double their performance.

....hoping to pick your brain once I get mine.

[suchmoon]

We are using the lightning asic tp-link controllers and a remote power switch to power cycle when the USB bus hangs.  Does a PC platform give you much of an advantage performance wise or stability wise?  It looks like the tuning options are better, but how is the stability?  The tp-link boots up pretty quick, I'd hate to have to keep cold-starting a server/pc to recover the USB bus. I guess we could probably root and install this on our TP links without too much difficulty. 

We're going to have to put in a couple of servers to do the pool right.  The front end and stratums will be on 2 different VM's on one server in the DMZ, and the database and coindaemons will be on the other on our LAN.

so quick question... you on the modded cpuminer with the power fix that makes these ~8W per unit on scrypt?

Also, I figure you are on Linux so... here ya go...

https://github.com/dtbartle/cgminer-gc3355

nice cgminer 3.7.2 with gridseed support and full stats minus temps (guess there is not a sensor?) and has the same power fix so ~8W per unit.

I now have one LightningAsic with TP-Link, and two "generic" sets with the wiibox. Both controllers are crap, but the TP-Link at least works and is in English, I couldn't get the wiibox work at all. Eventually I hooked up the "generic" sets to a PC and I'm using cpuminer (can't get cgminer to work with Windows COM ports, will try Linux later). Anyway, I can now compare TP-Link vs. computer-attached solution.

TP-Link has a decent UI (with some minor quirks) with quick status/hashrate overview. That's pretty much where it's advantage ends. It restarts miners every 1-2 hours on wafflepool (~330Kh/s per miner avg), and is almost unusable on clevermining or middlecoin (restarts every 10-20 minutes). If you try to connect more than 10 miners to the box it is restarting miners more often and it also seems to become less responsive, just not powerful enough I guess.

Cpuminer on Windows lacks hashrate stats and I need to run a separate cmd window for each miner, so it's a bit of a hassle (cgminer would be better). However the miners now are running stable for nearly 24 hours at ~340 Kh/s on clevermining, 2% reject rate. Handling 20 miners was not an issue, zero CPU usage on a cheap Celeron, just a few megs of RAM per instance of cpuminer, so it can certainly handle a lot more than that. My next step will be to test it with Raspberry + cgminer + a relay to toggle USB hub power remotely (optional - if the next few days show it to be necessary).

[suchmoon]

I'm using this now:

https://litecointalk.org/index.php?topic=9908.msg138204#msg138204

Fantastic so far... rock solid and totally effortless to set up. 360 Kh/s per miner.

[MWNinja]

How many can a Pi handle? This looks great and I can add it to my Pi distro.  I only have a couple of Pi's though, so we will need to order some more of them. We have an ubuntu (sick little i5 microITX rig) machine that is running our private pool, and we will be adding 2 dedicated 1U Xeon servers to take the pool public. For now we need your hashes to help with the variance. Please contact me via PM to help.

I'm using this now:

https://litecointalk.org/index.php?topic=9908.msg138204#msg138204

Fantastic so far... rock solid and totally effortless to set up. 360 Kh/s per miner.

[suchmoon]

I have 20 on it now, does not seem to have any issues so far. I'll try to daisy-chain a couple more hubs over the weekend and see if it can do 30 or 40.

How many can a Pi handle? This looks great and I can add it to my Pi distro.  I only have a couple of Pi's though, so we will need to order some more of them. We have an ubuntu (sick little i5 microITX rig) machine that is running our private pool, and we will be adding 2 dedicated 1U Xeon servers to take the pool public. For now we need your hashes to help with the variance. Please contact me via PM to help.

I'm using this now:

https://litecointalk.org/index.php?topic=9908.msg138204#msg138204

Fantastic so far... rock solid and totally effortless to set up. 360 Kh/s per miner.

[lostering]

Can you remove (payment due) after my name on owner's list when you finalize the list.
And are we expecting payout tonight?

Bests 

[Flep182]

Problem with Raspberry is that it can only go 2 hubs down and that it doesn't do well with usb 3.0.

Have 22 on one now but so far except for one that doesn't send shares it's looking well. I've ordered better hubs from china and expect them next week. Currently fixing my last set of 10 and see whether I can get the Pi up to 32 with my setup. It doesn't work well with hotplugging 10 at a time though

I've ordered two more Pi's and that should fix most of the problems

[tntdgcr]

Problem with Raspberry is that it can only go 2 hubs down and that it doesn't do well with usb 3.0.

Have 22 on one now but so far except for one that doesn't send shares it's looking well. I've ordered better hubs from china and expect them next week. Currently fixing my last set of 10 and see whether I can get the Pi up to 32 with my setup. It doesn't work well with hotplugging 10 at a time though

I've ordered two more Pi's and that should fix most of the problems

22 ? I've heard the PI can do about 50 in perfect setup
My distro has 200+ coming in so we'll see how they work soon in that config

[MWNinja]

Well my Pis were crashing more than the Tp Link, and then they trash the SD card some of the time, so remote reboot is risky.  They did run really fast on the Pi.  I tried running them on a beaglebone and it worked and was very stable, but the performance was lower for some reason.  I'll try on a linux PC next, so far my experience with these Gridseed units continues to be frustrating.

[suchmoon]

Problem with Raspberry is that it can only go 2 hubs down and that it doesn't do well with usb 3.0.

Have 22 on one now but so far except for one that doesn't send shares it's looking well. I've ordered better hubs from china and expect them next week. Currently fixing my last set of 10 and see whether I can get the Pi up to 32 with my setup. It doesn't work well with hotplugging 10 at a time though

I've ordered two more Pi's and that should fix most of the problems

22 ? I've heard the PI can do about 50 in perfect setup
My distro has 200+ coming in so we'll see how they work soon in that config

After some experimenting I think the theoretical maximum is 98 perhaps... two levels of 7-port hubs, 49 on each port. Unfortunately the "stock" 10-port hubs internally already have two levels, so can't daisy-chain those to the Pi. I left my Pi running with the 20 miners it started with, and will probably buy another one for the other 20.

[suchmoon]

Bought 2 (making it 4 of each GB4 and GB5).

1BhvKqYiew1ofQPg8L2QdN1W713oPScNG7

https://blockchain.info/tx/1c88deea0faba482d23db0d613cb1e2c0fd3b68325909a7d7959399bfb200c47

Ninja, could you please check this. I should have 4 shares in GB5. Owner's list shows only 2 and I believe my weekly payment is also for 2. Thanks!

[daemonfox]

Just noticing some rounding discrepancies so wanted to ask how that is being handled... what are the rules for rounding dividend payments?

See Lines 22-25 for inconsistencies.
https://docs.google.com/spreadsheet/ccc?key=0AsVvuskew8TmdFFrLWZXbi1SQUk3SzFhcWFXbXVRd3c

[MWNinja]

Just noticing some rounding discrepancies so wanted to ask how that is being handled... what are the rules for rounding dividend payments?

See Lines 22-25 for inconsistencies.
https://docs.google.com/spreadsheet/ccc?key=0AsVvuskew8TmdFFrLWZXbi1SQUk3SzFhcWFXbXVRd3c

Theres some rounding (set to 5 digits) in my Google payout spreadsheets. The rounding is a little weird, as it will show a rounded per share value, but then use the full precision for calculating the actual payout, which is then rounded to 5 digits. 

[suklee]

I'm still a bit confused after our previous PM exchanges... the GB 5 page says 0.01483 per share * 4 shares = 0.05932, but the payment that came in was 0.0593.

On the other hand for GB 4, each share = 0.0186598190 * 20 = 0.37319638, but I got 0.3732 which is slightly more than I'm supposed to receive.  

[suchmoon]

Well my Pis were crashing more than the Tp Link, and then they trash the SD card some of the time, so remote reboot is risky.  They did run really fast on the Pi.  I tried running them on a beaglebone and it worked and was very stable, but the performance was lower for some reason.  I'll try on a linux PC next, so far my experience with these Gridseed units continues to be frustrating.

Have you tried this?

https://litecointalk.org/index.php?topic=9908.msg139447#msg139447

I don't know if it works since my Raspberry didn't seem to crash before, but others are reporting it fixes the issue.

[MWNinja]

Bought 2 (making it 4 of each GB4 and GB5).

1BhvKqYiew1ofQPg8L2QdN1W713oPScNG7

https://blockchain.info/tx/1c88deea0faba482d23db0d613cb1e2c0fd3b68325909a7d7959399bfb200c47

Ninja, could you please check this. I should have 4 shares in GB5. Owner's list shows only 2 and I believe my weekly payment is also for 2. Thanks!

You are correct, thankfully there's a blockchain that has all the history! I will fix it today and send your missing payment.

[suchmoon]

Bought 2 (making it 4 of each GB4 and GB5).

1BhvKqYiew1ofQPg8L2QdN1W713oPScNG7

https://blockchain.info/tx/1c88deea0faba482d23db0d613cb1e2c0fd3b68325909a7d7959399bfb200c47

Ninja, could you please check this. I should have 4 shares in GB5. Owner's list shows only 2 and I believe my weekly payment is also for 2. Thanks!

You are correct, thankfully there's a blockchain that has all the history! I will fix it today and send your missing payment.

Got it, thank you!

[MWNinja]

Well my Pis were crashing more than the Tp Link, and then they trash the SD card some of the time, so remote reboot is risky.  They did run really fast on the Pi.  I tried running them on a beaglebone and it worked and was very stable, but the performance was lower for some reason.  I'll try on a linux PC next, so far my experience with these Gridseed units continues to be frustrating.

Have you tried this?

https://litecointalk.org/index.php?topic=9908.msg139447#msg139447

I don't know if it works since my Raspberry didn't seem to crash before, but others are reporting it fixes the issue.

That seems to have done the trick, my miners at home are now stable.  We will be procuring 8 Raspberry Pis for this group buy.  Hopefully we can recover the cost of the Pis by selling off the TP Link controllers.

[daemonfox]

Well my Pis were crashing more than the Tp Link, and then they trash the SD card some of the time, so remote reboot is risky.  They did run really fast on the Pi.  I tried running them on a beaglebone and it worked and was very stable, but the performance was lower for some reason.  I'll try on a linux PC next, so far my experience with these Gridseed units continues to be frustrating.

Have you tried this?

https://litecointalk.org/index.php?topic=9908.msg139447#msg139447

I don't know if it works since my Raspberry didn't seem to crash before, but others are reporting it fixes the issue.

That seems to have done the trick, my miners at home are now stable.  We will be procuring 8 Raspberry Pis for this group buy.  Hopefully we can recover the cost of the Pis by selling off the TP Link controllers.

In an effort to maybe increase total earnings since you have had to come out a bit more in cost to get these stable... I have found these gridseeds can easily run at 850 frequency and hit 383 KH/s without a fan and no worries of overheating as long as their is some ambient airflow. We seem to be running at about 300 kh/s now per device on average... as long as you have either the optmized cpuminer that runs ~8w per device or cgminer, you should be able to get those kinds of results easily.

That would be a jump to over 61 MH/s on average. A 27% increase just by changing to 850.