The column after the [4:x] shows how many of the 756 cores have not produced good results. It looks like chips 5 and 12 are half (or more) dead, so you'll need to leave autotune off and just accept that the noncerate on those will be less than half their hashrate. You can try different speeds to see what produces the best noncerate. You'll probably want to clock them as high as the rest, and just expect 50%+ error rate. Chainminer autotune can't handle chips with alot of dead cores and assumes they need to be downclocked or shut off. After running for long enough to get most of the bad-core counts near 0, look at /tmp/.core.log and see what 5 and 12 look like. It'll probably show that one side or the other is dead. Here's one of my half-dead chips: pi@bitfury ~ $ grep ^41 /tmp/.core.log 41 35 36 30 22 31 31 35 35 33 39 21 31 34 0 0 0 0 0 0 0 0 41 31 35 36 38 29 18 42 27 28 26 25 31 25 0 0 0 0 0 0 0 0 41 34 34 36 34 25 28 27 29 26 31 28 26 33 0 0 0 0 0 0 0 0 41 32 31 29 40 33 29 28 26 35 40 31 35 42 0 0 0 0 0 0 0 0 41 19 23 27 29 40 31 40 18 32 22 35 30 38 0 0 0 0 0 0 0 0 41 31 18 26 34 21 25 31 38 39 24 25 34 39 0 0 0 0 0 0 0 0 41 30 24 35 34 32 19 33 26 36 38 30 34 34 0 0 0 0 0 0 0 0 41 43 28 26 36 24 33 36 26 33 44 23 29 40 0 0 0 0 0 0 0 0 41 31 37 32 24 28 31 39 31 22 42 28 37 30 0 0 0 0 0 0 0 0 41 36 20 25 35 28 29 30 25 27 25 35 29 33 0 0 0 0 0 0 0 0 41 28 25 36 38 24 33 32 27 36 25 28 34 34 0 0 0 0 0 0 0 0 41 37 41 36 40 37 24 27 20 39 32 30 34 26 0 0 0 0 0 0 0 0 41 32 26 34 28 32 32 26 33 31 36 35 31 23 0 0 0 0 0 0 0 0 41 27 31 33 26 24 37 30 32 29 28 26 32 31 0 0 0 0 0 0 0 0 41 30 23 25 30 25 23 30 24 30 19 26 39 30 0 0 0 0 0 0 0 0 41 29 38 39 33 30 32 32 37 27 25 41 29 32 0 0 0 0 0 0 0 0 41 29 31 33 30 30 32 39 37 25 35 19 30 51 0 0 0 0 0 0 0 0 41 25 20 22 33 29 39 31 29 30 36 36 34 38 0 0 0 0 0 0 0 0 41 19 30 33 37 23 26 29 37 34 28 35 27 24 0 0 0 0 0 0 0 0 41 41 23 40 27 32 33 28 36 26 28 45 32 30 0 0 0 0 0 0 0 0 41 33 36 29 34 23 36 31 25 27 25 30 37 20 0 0 0 0 0 0 0 0 41 24 33 26 38 23 29 23 30 28 41 36 35 23 0 0 0 0 0 0 0 0 41 35 31 34 34 28 29 38 18 27 33 31 37 37 0 0 0 0 0 0 0 0 41 36 41 30 20 30 35 45 29 37 38 25 26 44 0 0 0 0 0 0 0 0 41 43 33 33 27 41 19 27 33 25 30 23 36 31 0 0 0 0 0 0 0 0 41 26 26 32 30 30 25 39 33 32 32 35 29 37 0 0 0 0 0 0 0 0 41 45 29 34 38 36 25 29 37 34 37 36 26 38 0 0 0 0 0 0 0 0 41 30 43 34 25 35 29 28 30 33 27 27 39 28 0 0 0 0 0 0 0 0 41 32 27 39 36 38 28 34 44 31 32 37 30 35 0 0 0 0 0 0 0 0 41 28 30 37 35 39 16 39 30 25 27 30 46 28 0 0 0 0 0 0 0 0 41 35 27 28 30 39 35 19 23 33 39 35 43 30 0 0 0 0 0 0 0 0 41 30 31 29 34 28 37 33 22 34 43 37 32 45 0 0 0 0 0 0 0 0 41 30 27 39 23 25 24 39 38 28 28 28 29 30 0 0 0 0 0 0 0 0 41 27 28 26 29 37 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 41 27 25 21 32 31 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 41 35 31 40 35 22 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 41 312
|
|
|
I think you need to wait for 7 and 8 to stop flashing before running the software.
|
|
|
Why are you compiling chainminer? If all you're trying to do is change clock speeds, just edit /opt/bitfury/best.cnf and reboot. If that file doesn't exist yet, copy /run/shm/.stat.log to /opt/bitfury/best.cnf first.
1. How i copy /run/shm/.stat.log to /opt/bitfury/best.cnf ? 2. How i get /run/shm/.stat.log ? 1. cp /run/shm/.stat.log /opt/bitfury/best.cnf 2. If /run/shm/.stat.log doesn't exist, let the miner run for at least 5 minutes and it should be there.
|
|
|
You removed the plastic from both sides of the pads, right? The plastic on one side is textured, the other smooth.
I think it's normal for the heatsink to feel just a little warm, even near the board. I thought that was a problem at first, but when I unplugged the fan, it got almost too hot to touch.
correct i removed them from both sides of each strip , also when i remove the fan it jumps up to 80c and drops in GH my chili looks like the one on the bottom left http://i40.tinypic.com/rr8its.jpgNot saying this is your problem necessarily, but I wonder how much difference it makes to orient the heatsink so that some air flows over the VRMs. I'd turn it 90 degrees from how it's shown in that picture. I'm using the Hyper 212 Evo, blowing end to end, which might be creating just a bit of air turbulence around the VRMs.
|
|
|
Mr teal
Any details Like drivers How install the cooler And how install the thermal pad?
If you want to use cgminer, download the Zadig utility that was previously linked to replace the FTDI driver with a WinUSB one. The instructions are in the cgminer directory. It works the same as a regular BFL device. I don't know what cooler you have. Put it between the cooler and the ASICs, and smoosh the two together. Somebody have an alternative link for Zadig drivers for cgminer? I'm getting error message that I don't have sufficient priv. to access at http://ck.kolivas.org/apps/cgminer/zadig/Maybe you need to run as administrator?
|
|
|
Any new firmware please!,,,,,,,,,!,,,,,,
+1 I've lost alot of mining revenue from one of my Chilis because it keeps hanging and needing its power cycled. I'm hoping the firmware update will fix that. Any chance you can give us whatever update you have today even if it doesn't have all the planned features?
|
|
|
OK - not keeping up with all of this. I see there are h-boards available on the website. If I order one, when is the "expected" delivery?
So h-boards are back in stock with 250 count... where's the shipments?
Wow, I literally hit Save about 30 seconds ago...yes I have a few extra H-cards of the old type that arrived. Since I don't the M-boards for them, they are only useful for filling out your V1 or V2 M-boards. Buy them now, and they will ship immediately. Production is cranking for the next-gen boards, but I haven't received any shipping notifications yet. I expect to see something in the next day or two...when product is on the way I'll give an update.
|
|
|
Why are you compiling chainminer? If all you're trying to do is change clock speeds, just edit /opt/bitfury/best.cnf and reboot. If that file doesn't exist yet, copy /run/shm/.stat.log to /opt/bitfury/best.cnf first.
|
|
|
You removed the plastic from both sides of the pads, right? The plastic on one side is textured, the other smooth.
I think it's normal for the heatsink to feel just a little warm, even near the board. I thought that was a problem at first, but when I unplugged the fan, it got almost too hot to touch.
|
|
|
have it on again ... a problem when connecting one pcie after the other when psu is on? 2) is there a switch for a pcie cable? ...
thanx or your help
As the 12+ of the cable comes from one output u could just add a kill switch to it, cut the cable and solder the switch. Any 220V 10A rocker switch would do the trick, dont forget to insulate the soldered terminals. Or you can make an patch cable out of these (link) and the switch, that way you dont have to hack you psu cables. Rather than trying to switch all that high amperage, why not just use the green wire on the 24 pin connector the way it's meant to be used - as an on/off switch? http://www.robotshop.com/en/cytron-atx-power-supply-breakout-board-right-angle.htmlHe wants to power multiple rigs from one PSU but be able to turn off just one of them.
|
|
|
If you're lucky and your chips are very coplanar, you can get away with paste. But usually pad is required because of variance in chip levels.
|
|
|
Even without overclocking, a full rig pulls something like 25A from the 12V input. You'll need a much heftier switch than that.
|
|
|
So basically things turn from bad to worse very very quickly.
what about if all the chips had parallel power, and the m-board (or an external PSU) simply provided 0.85V? That's what the standard design does - you have one power regulator (the big black brick that says PULSE on it) that powers all chips in parallel and all chips get the same voltage. If one chip dies the rest will continue working normally. If the chip that dies actually short-circuits the board the regulator will normally detect the short and just completely cut out power. So in that case the board will still not work, but at least it won't melt the rest of the chips. And with some luck you could insulate the bad chip and get the rest working. Actually the big Pulse component is an inductor. The regulator is the small TI chip below it.
|
|
|
I can sell you the good stuff (Fujipoly 11W/mK) for one 8-chip board for $5.69 including shipping within the US. PM me.
|
|
|
Thanks for the pics. Aren't those heatsinks for the VR really close to shorting something though?
|
|
|
Here's a script I'm running (in screen) to display per-board stats every time chainminer updates the file (every 5 minutes), and copy the full stats to a log dir: pi@bitfury ~ $ cd /run/shm ; if [ ! -e logs ] ; then mkdir logs ; fi ; laststatmod= ; while [ 0 ] ; do statmod=`stat -c %Y .stat.log` ; if [ "$statmod" != "$laststatmod" ] ; then laststatmod=$statmod ; echo ; stat -c %y .stat.log ; grep --color=never -m 1 -A 16 noncerate .stat.log ; cp .stat.log logs/$statmod ; fi ; sleep 5 ; done
2013-10-27 09:34:41.136382762 +0000 speed:8061 noncerate[GH/s]:321.507 (2.233/chip) hashrate[GH/s]:325.651 good:22457 errors:273 spi-errors:5 miso-errors:0 jobs:323 (record[GH/s]:327.649) 0: 900 34.646 36.381 2420 18 0 0 1: 897 35.591 36.064 2486 9 0 0 2: 900 35.620 35.535 2488 51 0 0 3: 901 35.849 35.747 2504 12 1 0 4: 910 35.978 36.169 2513 45 1 0 5: 894 36.035 36.085 2517 14 0 0 6: 891 36.035 36.888 2517 13 0 0 7: 873 34.947 35.842 2441 79 0 0 8: 895 36.808 36.941 2571 32 3 0
2013-10-27 09:39:41.054781143 +0000 speed:8067 noncerate[GH/s]:318.558 (2.212/chip) hashrate[GH/s]:326.380 good:22251 errors:307 spi-errors:4 miso-errors:0 jobs:322 (record[GH/s]:327.649) 0: 899 36.020 36.455 2516 14 0 0 1: 899 35.691 36.391 2493 25 0 0 2: 902 34.331 35.969 2398 138 0 0 3: 904 36.035 36.148 2517 11 0 0 4: 909 35.663 35.757 2491 8 0 0 5: 896 34.417 36.085 2404 12 1 0 6: 891 35.047 36.930 2448 16 2 0 7: 873 33.730 35.673 2356 63 1 0 8: 894 37.624 36.973 2628 20 0 0
Watch out for running out of ramdisk space. It will fill up in less than 2 weeks with a full rig. I'll need to think of a way to send the files to another location for storage if I want, or just prune the old data occasionally. I'm doing this logging so I can notice if performance changes over time and which chips are the culprit.
|
|
|
Just to share some numbers...
So far, without heatsinks, my BitFury rigs are up to a total of 743 GH/s, for 21 H-boards (minus 3 bad chips on one), so an average of 2.23 GH/s per chip or 35.7 GH/s per full board. Ambient is 26C, and I have 3 Delta AFB1212VHE (148 CFM each) fans blowing through both rigs. With my cheapie IR thermometer, I didn't find any spot on the boards hotter than 70C, but I wan't able to focus closer than probably 1" diameter. Average voltage as measured between the top of the Pulse inductor and the M-boards' ground terminal is 0.864v. I'm running the original v1 chainminer on one board and the latest chainminer on the other, both on auto. Total power usage at the wall is 782W including probably 30W for the fans. The PSU should be about 86% efficient.
|
|
|
It's putting a heatsink on the back of the voltage regulator that looks like a challenge because of the components around it. Has anyone succeeded with that?
|
|
|
EDIT: I have heatsinks on back of all boards in regulator area, and 120 CFM fans!
Does the backside of the regulator even get hot? I have felt the top of the chip and its very hot but I don't feel much on the backside. Wouldn't putting a small heatsink on the chip do more? For the hashing chips those thermal vias work well as even touching the vias themselves gets hot, even better with a heatsink. Yes, it gets quite hot. I believe they are also constructed do dissipate the heat to the board, but I did not do any research in that matter. In my opinion the regulator heatsink is much more important than chip heatsinks. +1, the most important task before serious overvolting of the h-boards (0.8V+) is to first stick heatsinks on the back side of the board under the regulator. If possible, stick a heatsink on top of the regulator and the inductor. This way you can take them as high as 1V. The board will dissipate close to 70Watts at 1V. Can you recommend a source for suitable heatsinks?
|
|
|
|