Okay, up and running with a couple of good sized tantalum caps on the module 12v rail.  Will catch up on posts while it settles out on 0.94 then see if 0.95 is any better.  Opps, forgot to enablecores as a few were down on 1 die.

Just off the treadmill and a couple of thoughts.  Must get DHCP working on the RPi assigning addresses to the .2.0 network so I don't have to add a static address to whatever is to the RJ45 connector.  But with the hot pepper USB connection, I can just plug it into the RPi and run cgminer on the RPi and it will be handled on the 1.0 network to wlan0.  I would need the RPi net .2.0 routing for the Merc tho as that would be to the RPi RJ45.

Simpler is to really draw the heat out of the BBB.  I have my first native BBB and haven't fired it up yet.  Different by far than Debian/Ubuntu/Rasperian linux, no apt-get to install programs.  Still I'm sure it comes native with whyfri pre-configured.  I have old racks of three fans that would plug into a CD drive opening, 12v, and as they're small, I think I'll put standoffs on the native BBB, add braces to fit the smaller 12v fan, and swap the native BBB for the Merc BBB.  The fan will be blowing down on the wrong side of the BBB pcb to best cool the warmest chip, but should work.  Then bringing out the USB on a cable, put in a USB wifi and wake up the Merc with the ASIC disconnected and no RJ45 ethernet connected and see if DHCP assigns an address to the BBB.

The old WinXP laptop has a wireless that I disabled for testing this RPI router.  Since wasn't all too difficult I might try making it the wireless router instead of the RPI.  It does only have XP home so auto config is out but with the route command it should be possible.

When I go to try the RPI wireless router I'll test with the hot pepper from the unmentionable company.  That should give it a suitable workout.

I bought into high speed wireless switches early and am still using the things.  I've got a TP-Link 8 port and a TP-Link 4 port.  They make great ethernet extenders.  My WDTV can do wifi and I bought the Cisco USB but at the time my wireless routers were only 802.11b and wouldn't accept the Cisco USB adapter logging in.  So, I ran cat5 cable and put the 4port switch there by the TV, added a 2.5" hard drive to the WDTV and I can drag and drop two weeks worth of TV programs onto the WDTV drive.  Using the high speed switch at the end of the cat5 run really improves thruput.

What? spend more money?

This was my first attempt at wireless isolation.  Nothing like the dependability of a cat5 connection so unless I hear thunder....

Bet it's possible to do the reverse.  Say you have a non-wireless router.  Bring up the RPI, plug in a USB wireless, put the wireless on another net, adding routing and forwarding, boot a tablet and assign it a static IP in that new net, and you've converted a non-wireless router into a wireless router.

Darn 60 something unread messages I've got to catch on back there, got no work done outside, will do my mileage on my treadmill rather than on the road so at least I'll get in an hour running.  Didn't get to adding caps to the ASIC module to test if it can then run better on 0.95.  I have one die that keeps losing 1 to 3 cores and I think it's possibly the VRM getting the least cooling, but where does the time go.  Isn't that a song?  ...where does the time go?

And got the RPI working as a router.  It had an eth0 address 192.168.1.50 before.  I added the Cisco USB adapter.  That was given the address 192.168.1.116.  

I added a route to network 2.0 with "route add 192.168.2.0 eth0"

I went into /etc/network/interfaces and changed the eth0 static address from 192.168.1.50 to 192.168.2.50 and changed the broadcast address as well.  Then I restarted networking with /etc/init.d/networking restart which this time didn't disconnect me because wlan0 stayed.

Then I booted my WinXP and gave its internet connection a static address of 192.168.2.6 and changed the gateway to 192.168.2.50 and rebooted.

Could ping 192.168.2.50 and 192.168.1.116 so the WinXP laptop was reaching the wlan0 port of the RPI but when I tried pinging 192.168.1.1 it failed.

I gave the command as root on RPI "sysctl -w net.ipv4.conf.all.forwarding=1" the tried pinging 192.168.1.1 from a dos box on the WinXP home and Bob's your uncle it worked.  Then I pinged yahoo.com from WinXP with complete success.

Now I can use the RPI as a wireless router I'll just have to establish a static address in the .2.0 net on the machine that's going to be forwarded.

When the last digit is a zero it is the network.  A gateway on 192.168.1.1 is on network 192.168.1.0. 

Drat, dropped my whole network again.

Tried bridging RPI with wlan0 and eth0 on same network 192.168.1.0/24.  I had been using the RPI for a mining proxy but shifted that load to another.  Stock RPI.  Added a Cisco USB 802.11g adapter and it came right up and was assigned an address.  Added route "route 192.168.1.x (the eth0 IP) wlan0".

Then I woke up an old WinXP Home laptop having an address on the 192.168.1.0 network but plugged into the RPI eth0 port.  On that I added a route "route add 192.168.1.y (RPI wireless IP address) eth0" and I could ping that address.

When I ran "route" no options on the RPI it showed a route with the computer name of the WinXP!  So, the RPI had communicated with the WinXP laptop.  But when I tried pinging the gateway of network 192.168.1.0, which of course is 192.168.1.1, it failed.  I wasn't getting forwarded on the RPI.  

So on the RPI I tried "sysctl -w net.ipv4.conf.all.forwarding=1" which I got from here (http://www.raspberrypi.org/phpBB3/viewtopic.php?f=36&t=52246) but it still didn't forward.  so I tried going to a separate net as he had done, changed the eth0 port on the RPI to 192.168.2.x and tried to add a route between the two subnets on the Cisco E900 router but that's when my network crashed.  Have it up now and still trying.

Would ROFL except it made me sad.

Would really appreciate it if anyone getting a miner with fans knocked off in transit mark those ASIC modules some way that if later there's one or more modules with failing cores it can be noted as previously had the fan knocked off or not.

Theory: if a shock separated a solder point on the ball grid array of the ASIC module but still makes physical contact tho not flowed-solder contact, it will make a somewhat higher resistance point under current and raise temperature and that raised temperature will cause cores with gates at the hot spot to have more errors.

Yes, I think it's go big or go home.  Will have to keep adding to hash power to stay up while those faint of heart or broke fold.

My module took a T10.

Might try to get bridging working on an RPi but there's work to do on the truck, on the car, on the pumphouse roof, and a cold front is heading south this way, not to mention adding tantalum caps to the Merc Asic module and testing.

The thread is at page 633 so I didn't read it searching for a solution tho did search "duplicate share" and didn't find a solution.  I'm running a KnC Mercury and seeing a lot of these:

[2013-10-21 03:10:08] Accepted 019ac0d5 Diff 159/113 KnC 0 pool 1
 [2013-10-21 03:10:11] Accepted 01aeaa0e Diff 152/113 KnC 0 pool 1
 [2013-10-21 03:10:14] Accepted 0207bcfd Diff 126/113 KnC 0 pool 1
 [2013-10-21 03:10:14] Rejected 0207bcfd Diff 126/113 KnC 0 pool 1 (Duplicate share)
 [2013-10-21 03:10:15] Rejected 0207bcfd Diff 126/113 KnC 0 pool 1 (Duplicate share)
 [2013-10-21 03:10:26] Accepted 00341792 Diff 1.26K/113 KnC 0 pool 1
 [2013-10-21 03:10:26] Accepted 0052d1a1 Diff 791/113 KnC 0 pool 1

I tried using a mining_proxy but the firmware has a minimum difficulty set that is too high for the proxy I think.  And, trying mining_proxy.py --real-target 100 fails.

Is there a solution to the Duplicate share problem?  Each gives me a reject and that figure is below:
KnC 0:  | 149.1G/137.0Gh/s | A:307173 R:20267 HW:25192 WU:2183.1/m

Thanks.

soy

I'm not going to add the caps now as after a 7+mile jog and that it's late afternoon and that I'm getting on in years, I'm more likely to make a mistake.  Maybe tomorrow after lunch before a run.

Using 0.95 or 0.96 will save me more than $5.50 per month only if I get the same hashrate and WU.

Okay, whether the KnC caps are there or not, the VRMs have internal caps on the 12v rail, and some of those are assuredly low ESR, and those are not there when the VRM is not installed so adding low ESR caps to the 12v rail should improve it.