Labor Day ? Blades on Strike ?

[dbkeys]

I have Blades that started misbehaving  2 days ago. One Blade's comm modules completely stopped functioning (no yellow link light, no activity at all) and a replacement module is on its way from ASICMiner.

But the other Blades are now only developing less than 1000 MH/s. One is at "0", and it seems to be re-setting or re-booting itself, for no reason,  approximately every 2minutes 12 seconds.  (Link light goes off momentarily, Up Time counter re-sets to 0)

I've tried different pools, stripping down the services on the Linux box running the Stratum proxy, to no avail ....

[wpgdeez]

It's most likely a problem with your power, check that your PSU is not dying. Have you checked the voltage on the different lanes?

[HellDiverUK]

Sounds like a dead/dying PSU to me too.

[dbkeys]

It's most likely a problem with your power, check that your PSU is not dying. Have you checked the voltage on the different lanes?

The 2 power supplies are both new 700W Cooler Master "Extreme Power Plus" (RS-700-PCAA-E3) units, with single 12V rails that can push 52A. One was powering 3 blades and 5 fans(2x 120mm, 3x 100mm), the other 2 blades and a larger 195mm fan.

I did check the voltage at the green supply connector to the blade, one at +11.70v, the other at +11.35 which do seem a bit low, but perhaps this is normal under load ? I haven't yet checked the individual power lanes, nor attempted the voltage tuning.

Since there seems to be consensus this is power supply related I will get a more powerful unit (1000 W or 1200 W) and see if this is the problem.

Thanks for your input !

[candoo]

Gosh 5 Blades are ~ 600 Watt and you have 2 psus = 1400 Watt ?

[wpgdeez]

It's most likely a problem with your power, check that your PSU is not dying. Have you checked the voltage on the different lanes?

The 2 power supplies are both new 700W Cooler Master "Extreme Power Plus" (RS-700-PCAA-E3) units, with single 12V rails that can push 52A. One was powering 3 blades and 5 fans(2x 120mm, 3x 100mm), the other 2 blades and a larger 195mm fan.

I did check the voltage at the green supply connector to the blade, one at +11.70v, the other at +11.35 which do seem a bit low, but perhaps this is normal under load ? I haven't yet checked the individual power lanes, nor attempted the voltage tuning.

Since there seems to be consensus this is power supply related I will get a more powerful unit (1000 W or 1200 W) and see if this is the problem.

Thanks for your input !

I had the same PSU and it has slowly died on me. At first I could power 3 blades, then 2 now 1. The PSU sucks. Try just hooking up 1 blade to it and see how that runs.

[dbkeys]

I bought a "Coolmax" LCD power supply tester and it shows both power supplies within range. I'm still not totally convinced all is OK with the power ... it could be daily variations at the wall 110 VAC due to air conditioning equipment, pool pumps or other large loads kicking in perhaps...
But... as of now, the blades are hashing again at full speed, on the same power supplies. They have done this before, and gone back to crawling at measly 300 MHS, so I don't know if anything I have done has helped or not, but here it is:

* Made sure Slush's stratum proxy is run with higher priority by the kernel:
   sudo nice -n -19 ./mining_proxy.py        
* Eliminated several services which I didn't need (samba, avahi, ipv6 stuff)
* Changed from Firefox web browser (which consumed large amounts of memory and CPU power) to Google Chrome, which seems to have a much lighter footprint on the proxy machine's resources.   (The web browser is needed to monitor, program and operate the blades. It would be nice if there were an API with a native application rather than only HTML/web browser )

db

[wpgdeez]

I bought a "Coolmax" LCD power supply tester and it shows both power supplies within range. I'm still not totally convinced all is OK with the power ... it could be daily variations at the wall 110 VAC due to air conditioning equipment, pool pumps or other large loads kicking in perhaps...
But... as of now, the blades are hashing again at full speed, on the same power supplies. They have done this before, and gone back to crawling at measly 300 MHS, so I don't know if anything I have done has helped or not, but here it is:

* Made sure Slush's stratum proxy is run with higher priority by the kernel:
   sudo nice -n -19 ./mining_proxy.py        
* Eliminated several services which I didn't need (samba, avahi, ipv6 stuff)
* Changed from Firefox web browser (which consumed large amounts of memory and CPU power) to Google Chrome, which seems to have a much lighter footprint on the proxy machine's resources.   (The web browser is needed to monitor, program and operate the blades. It would be nice if there were an API with a native application rather than only HTML/web browser )

db

As the blades hash they got hotter and so do the internal limiters in the PSU. If it goes down again try removing 1 blade from the equation to see if things are more stable. I just fried a 1050 watt trying to run 6 blades off it, 5 work fine. Time for a RMA!

[dogie]

Gosh 5 Blades are ~ 600 Watt and you have 2 psus = 1400 Watt ?

Limit isn't the PSU wattage, is the amperage on the cables

[dbkeys]

Blades are hashing at full speed now, and the culprit  apparently was wi-fi interference  OR excessive network traffic on the wire caused by a chatty or misbehaving wi-fi device.

These posts were very useful:
https://bitcointalk.org/index.php?topic=208212.0
https://bitcointalk.org/index.php?topic=213956.0

Turned off: (2) Netgear wireless access points
                           and...
                  (2) Android phones
                  (1) iPhone
                  (2) iPads

As soon as I returned from the search-and-turn-off patrol, and these devices were inactivated, the blades started rapidly climbing past 10,000 MHS and are now at or near  their 13,000 MHS high clock setting performance limit, all at better than 98% efficiency.

I did
   * improve the cooling on all the blades, (and BTW, it seems like having fans directly on the heatsink, with no gap between the heatsink and the fans, is NOT as effective as having a half inch or so air gap between the heatsink and the fans) and
   * improved the amp carrying-capacity of the power feeds to the blades (bought some PCIe 6 power cables which provide 3 wires each for +12V and ground) so... no chance the blades are amp-limited.

I considered also powering down a set of cordless phones, on the theory that these  operate at the same  frequency range as wi-fi, but that turned out to be unnecessary, which seems to indicate that it's not wifi RF interference per se which was the problem, but rather a "misbehaving" device which gets on the wired network through wi-fi (5 suspects listed above) or possibly a misbehaving wireless access point (2 suspects).

I now have to go through a process of elimination to pinpoint the misbehaver. Possibly use wireshark to examine / monitor the network traffic.

In any case, I'm very happy the blades are working. Turns out they were not on strike, they likely had Attention Deficit Disorder... they were being distracted by some noisy neighbors on the network.