|
Please consider selling the heatsinks separately as well. I would definitely be interested in buying a bunch of them for my current boards. How about opening pre-orders for just the heatsinks, so you can see how much interest there is for old boards? Or is there too much uncertainty about when they'll be ready to ship?
|
|
|
|
Try connecting with an irc client to ircnet.eversible.com and join #bfsbsupport |
|
|
|
So the v2.3 m-boards work with the v1.2 h-cards? I'm just about to design these in PADs.
Yes, I believe so. |
|
|
|
I'm actually tempted... In USD terms it'll ROI easily, except that I'd have to sell bitcoins to get that USD, so... Interesting why is the seller selling. What does he know that we do not know  That price is 10x too low. I saw some BFL-based miners sell on this forum recently for about $17/GH, and those use far more electricity. That makes a full BitFury rig worth more than $10,000. He should be selling that lot of 25 rigs for $250,000. |
|
|
|
why do my hashrate diversify so much with "autotune off" and "55" for all chips. between 520 and 440.
You probably have some H-boards with overloaded VRMs. Check /run/shm/.stat.log every 5 minutes and see if the hashrate of some chips are changing significantly. If so, turn down the speeds on all the chips on that H-board until that H-board's VRM can handle it. thank you very much! will try to do so. does this mean the "Voltage Regulator Module" on some boards is not able to supply the power which is needed for the chips to work with this clock settings? The V3 rigs were shipped with the H-board trimpots set to a higher voltage than the original design, so when overclocking you'll usually overload the H-boards' VRMs before getting many nonce errors. |
|
|
|
why do my hashrate diversify so much with "autotune off" and "55" for all chips. between 520 and 440.
You probably have some H-boards with overloaded VRMs. Check /run/shm/.stat.log every 5 minutes and see if the hashrate of some chips are changing significantly. If so, turn down the speeds on all the chips on that H-board until that H-board's VRM can handle it. |
|
|
|
Here are pictures of my adapter kit for mounting the Hyper 212 EVO to a Chili board. Since I don't have any boards in hand yet, I made a simple cardboard model. (click to enlarge)  Notes: * I recommend at least 3 full turns of the screws into the standoffs from each end for stability. If you assemble them as shown, the screws should meet right about in the middle of the standoff, about 4-5 threads (full turns) from each end. * The number of washers inserted in the center of the X bracket affects the pressure on the chips/pads, though I'm not sure by how much. By my calculations, 2 or 3 washers should achieve the same pressure as the cooler would apply to a standard cpu with the stock hardware. The kit includes 4 washers for that location, so you can choose how many to use. * The picture shows 4 pads, each 10x20mm, which you can cut in half to cover 8 chips. You can probably get away with not cutting them and covering 2 chips each if you position them very precisely. Make sure you remove the plastic from both sides of the pads.* The 12mm long M+F standoffs I'm using for the board's feet, and their nuts, are the ones that come with the cooler and normally attach the backplate to the board. I already sold all 17 parts kits I had for sale, but I still have thermal pads for 6 more boards, which I'll sell for $2 per board plus shipping. Is anyone interested in these kits for batch 2? Let me know how many. I figure I can get more parts in as soon as Monday if there's enough interest (at least 25 kits to make it worthwhile). I'm not going to order more parts for making and selling these kits, so I'll just share the parts list: For each Chili: (4 pcs) http://www.digikey.com/product-detail/en/9911-187/492-1088-ND/586408(4 pcs) http://www.digikey.com/product-detail/en/3349/3349K-ND/61924(8 pcs) http://www.digikey.com/product-detail/en/SAEFZ%20004/H777-ND/274988(4 pcs) http://www.digikey.com/product-detail/en/MPMS%20003%200012%20PH/H744-ND/274955(4 pcs) http://www.digikey.com/product-detail/en/24432/24432K-ND/1532164(4 pcs 10x20mm or 1 pc 35x35mm) http://www.frozencpu.com/products/16882/thr-167/Fujipoly_Extreme_System_Builder_Thermal_Pad_-_60_x_50_x_05_-_Thermal_Conductivity_110_WmK.htmlRemoved from the list are the large nylon washers for the board feet, since I discovered that the board's holes are a bit too small for the standoffs that come with the cooler. I just have my Chili's resting on the 4 screw heads that attach the backplate. |
|
|
|
Is the VRM working? Check voltage at the top of the Pulse inductor. You can use the M-board's ground screw for negative if you don't need an accurate reading. |
|
|
|
Is the only way to notch your speed up by making a best.cnf file? Also whats unlcear to me is when I make that best cnf file what all should be in it? Example: 1 AIfDSo 53 <-- is that the only line you need for each chip or should it be everything that the .stat.log has minus the very bottom readout?
If the autotuner isn't giving you the results you expect, you can use best.cnf to turn it off (lowercase a) and manually set speeds. You don't need to put all the stuff from .stat.log into best.cnf, but it's not a problem either. It only reads the first 3 tab-separated columns and ignores the rest. |
|
|
|
And how well does that regulator seem to work?
very well in fact. No overheating issues until max 1V that I have tested, but the bitfury chips get super hot, since I donot have adequate heatsinking for them. Do you know how many amps you're pulling from that 30A regulator? Don't you want something a little beefier like 40A? And is an output side inductor unnecessary for that regulator? Actually I donot have the testing tools to measure > 10A current. This chip is actually a 2phase regulator with 20A OCP limit per phase. So I am drawing less than 40A for sure. There are 2 0.2uH inductors inside the chip  . If bitfury will ever reduce prices of their chips, my plan is to get boards built and quickly solder the chips and couple of decoupling caps thrown in and connect this regulator on the outside. No fuss no muss. Nice! Maybe you could determine the amperage by measuring the resistance of the wire bundle between the reg and card while powered off, and voltage drop while running. Though it might not be possible to isolate it for the resistance measurement without desoldering. Do those wires get hot? Could you make H-boards with that Intersil reg onboard? |
|
|
|
And how well does that regulator seem to work?
very well in fact. No overheating issues until max 1V that I have tested, but the bitfury chips get super hot, since I donot have adequate heatsinking for them. Do you know how many amps you're pulling from that 30A regulator? Don't you want something a little beefier like 40A? And is an output side inductor unnecessary for that regulator? |
|
|
|
Last night my bitfury rig stopped mining on us1.ghash.io:3333 Specifically only my worker1 account.
Turns out my worker1 no longer mines on us1.ghash.io:3333.
it mines on nl1.ghash.io:3334[/b]
I ran into that same issue recently trying to mine on btcguild using the MPBM software. It seems that not all mining software handles the stratum "client reconnect" message, where the pool asks the client to change hosts. I'm guessing the included ChainMiner software also does not handle that message correctly. (that's just a guess). ok, this is weird. i too am finding one of my miners won't mine on BTCGuild, Slush, or Ozcoin all of a sudden. i think it's Chainminer. What's chainminer doing? sudo screen -r `sudo screen -r | grep -m1 Detached | cut -b2-7`
|
|
|
|
i haven't heard anybody talk about bad slots. for instance, i have an m board where slots 1 & 3 in bank 1 appear to be dead. i've swapped multiple good h boards that previously were hashing @ 32-36 into these slots and they won't even boot up beyond 0. how do you deal with that? another weird thing is, given those facts, why does Xian's strategy of swapping good cards to the front of the m board even work? does anyone understand the organizational structure of how voltage flows thru the 4 banks and 16 slots? Dave insinuated the concept of a bank doesn't apply for newer boards (v3 m boards) here: "Don't pay attention to the Bank 1, Bank 2, etc. Those line up for V2 M-boards. The numbers of the slots do line up as 1-16 map to 1 - 9, A - F. There, you can see the output of each H-card."https://bitcointalk.org/index.php?topic=288109.0Just forget about what Dave said there. V3 M-boards do have 4 banks of 4 slots each. If a chip or card is not forwarding messages properly, all chips after it within that bank will have a problem also. This is a communication issue. Power distribution is not chained like that; each card taps into the same power planes on the M-board, and each chip on a card taps into the same power planes on the card. Here's the mapping from the slot numbers printed on the M-board and listed in the summary section of .stat.log, to the web interface: M-board
.stat.log | |Web interface
| | | 0 | |Bank 1, slot 1 | | 1 | |Bank 1, slot 2 | | 2 | |Bank 1, slot 3 | | 3 | |Bank 1, slot 4 | | 4 | |Bank 2, slot 5 | | 5 | |Bank 2, slot 6 | | 6 | |Bank 2, slot 7 | | 7 | |Bank 2, slot 8 | | 8 | |Bank 3, slot 9 | | 9 | |Bank 3, slot 10 | | A | |Bank 3, slot 11 | | B | |Bank 3, slot 12 | | C | |Bank 4, slot 13 | | D | |Bank 4, slot 14 | | E | |Bank 4, slot 15 | | F | |Bank 4, slot 16 |
(How can I put gridlines in a table in bbcode?) |
|
|
|
*Facepalm* Someone please tell me they have a copy of the Vanilla V3 image, or that one of the older ones works? I was shipped a blank SD card.
And, yes, I did have the sense to plug it into a PC to try to back it up prior to inserting it in the Pi.
Edit: Bah, V2 won't fit on the card and GandalfG's image, while booted, isn't generating any hashes. I don't know if it's my rig or not, yet.
Login with ssh and check what chainminer is doing: sudo screen -r | grep -m1 Detached | cut -b2-7
i tried that command and got back a 4377.. what does that mean? Oops, I left off an important part. Here's the complete command: sudo screen -r `sudo screen -r | grep -m1 Detached | cut -b2-7`
|
|
|
|
Now that I got BFG running, I went through the voltages and adjusted them. The ones in the back ran away up to .88 and I trimmed them back down to .85-.87. I adjusted the ones in bank 0 up to .86. Bank 1 is at .83-.835. Bank 2 is at about .86.
You are now running the back cards at a much lower voltage than the front ones. Unless you accounted for the voltage rise on the ground plane from front to back of the M-board? The voltage between the M-board's ground screw and the top of the inductor, will vary depending on how close that card is to the M-board's power input. But the chips only care about the voltage between the card's ground and power planes. The correct way to measure voltage is between the top of the inductor and the back of the card's VRM (the exposed silver pad). That will remain the same regardless of which slot the card is in (except for airflow and temperature affecting the VRM, trimpot, or pencil-modded resistor). |
|
|
|
*Facepalm* Someone please tell me they have a copy of the Vanilla V3 image, or that one of the older ones works? I was shipped a blank SD card.
And, yes, I did have the sense to plug it into a PC to try to back it up prior to inserting it in the Pi.
Edit: Bah, V2 won't fit on the card and GandalfG's image, while booted, isn't generating any hashes. I don't know if it's my rig or not, yet.
Login with ssh and check what chainminer is doing: sudo screen -r | grep -m1 Detached | cut -b2-7
EDIT: I omitted an important part. Here's the complete command: sudo screen -r `sudo screen -r | grep -m1 Detached | cut -b2-7`
|
|
|
|
|
I, and a few others, have now received our payouts. Thanks Graet. Everyone check your wallets now.
|
|
|
|
|
Show Posts
