|
I just took my unit offline and checked the resistance of my pencil mod. immediate resistence was 1050k(ohm). after only 2 minutes it was at 1080k(ohm). I toned back the pencil mod and checked resistance at 1130k(ohm).
bootup voltage: 0.785V
3min voltage: 0.798V - ~35GH
My ideal voltage is about 0.82-0.83V (~40.5GH)
over 0.835V leads to issues like those experienced today, where hashrate constantly struggles around half speed, with high error rates
|
|
|
|
12.7mm height may be a little large for these boards, and may exceed the availale spacing between boards. 5-7mm would be better |
|
|
|
Can you make a photo where to measure ... the current ... ?
beside each chip there is a bank of 8 small capacitors. most of these (seems like half of them in my situation) carry the voltage to the chip (0.73-0.85v = 1.5-2.7GH depending on tuning) and can be easily checked with a multimeter. I usually check at least 2 caps to verify that all chips receive similar voltage and ensure accuracy. I find that on startup from cold pcb i have around 0.805v, which climbs to 0.825v after the system is running and getting toasty over an hour or so. during this time, error rates decrease slightly and hashrate climbs to optimal speeds of about 40-41GH for the board ps: current (amperage) is different from voltage  Does 40GH require you to use heat sinks or just fans blowing across the cards? i have heatsinks in use, and plan to upgrade them to slightly larger ones (35x35x6mm) soon. however, today i have had some issues in the past 3 hours when i lost hashrate at ghash.io and system reboot and start/stop miner ommands seem to only be temporary fixes where it returns to 40Gh, then slowly drops. error rates are within reason (slightly higher than normal), and chip 1 has 3 miso errors that dont seem to be replicating in any other chips. (typically it had 1 per 5min sample before) Its hard to troubleshoot without >20min of data, but i might have to check and/or back down the voltage slightly as im unsure whether the issue is the card or the pool (the initial problem seems to correlate with my blockerupters acting similarly, but they recovered and run fine now) edit: voltage is reading around 0.838V and this seems to be causing the chips to switch off or clock down towards default speeds (how would/does that work?) I think if it does not improve in the next hour i will take the system offline and try to tune down the pencil mod slightly. Ambient room temps are up 2-3 celcius in the last 24hrs now that heat in the building is on, and this may be the cause of lowered resistence=higher voltage Many of my cards are at 20GH. It sounds like I could try to get them to 30 without adding heat sinks? i would say that without adding heatsinks, but having proper airflow (a 120mm or 80mm fan pointed at the board is enough) you should be able to attain 30-35GH on a good board. If yours only runs 20GH by default, it sounds like some chips may be running slow so you may only see 27-30GH. Start at a voltage of 0.77v and see how it runs for 12hours. if it runs well, aim for 0.79v |
|
|
|
Since yesterday evening, I am having an enormous amount of duplicates, about 33% ?!
Any idea why this could happen ?
I have figure it has to do with the empty 4th bank. I placed back the non/hashing boards and the duplicates are back to īnormalī. my understanding was that if you fiull the banks unevenly it can cause issues |
|
|
|
Can you make a photo where to measure ... the current ... ?
beside each chip there is a bank of 8 small capacitors. most of these (seems like half of them in my situation) carry the voltage to the chip (0.73-0.85v = 1.5-2.7GH depending on tuning) and can be easily checked with a multimeter. I usually check at least 2 caps to verify that all chips receive similar voltage and ensure accuracy. I find that on startup from cold pcb i have around 0.805v, which climbs to 0.825v after the system is running and getting toasty over an hour or so. during this time, error rates decrease slightly and hashrate climbs to optimal speeds of about 40-41GH for the board ps: current (amperage) is different from voltage Does 40GH require you to use heat sinks or just fans blowing across the cards? i have heatsinks in use, and plan to upgrade them to slightly larger ones (35x35x6mm) soon. however, today i have had some issues in the past 3 hours when i lost hashrate at ghash.io and system reboot and start/stop miner ommands seem to only be temporary fixes where it returns to 40Gh, then slowly drops. error rates are within reason (slightly higher than normal), and chip 1 has 3 miso errors that dont seem to be replicating in any other chips. (typically it had 1 per 5min sample before) Its hard to troubleshoot without >20min of data, but i might have to check and/or back down the voltage slightly as im unsure whether the issue is the card or the pool (the initial problem seems to correlate with my blockerupters acting similarly, but they recovered and run fine now) edit: voltage is reading around 0.838V and this seems to be causing the chips to switch off or clock down towards default speeds (how would/does that work?) I think if it does not improve in the next hour i will take the system offline and try to tune down the pencil mod slightly. Ambient room temps are up 2-3 celcius in the last 24hrs now that heat in the building is on, and this may be the cause of lowered resistence=higher voltage |
|
|
|
Can you make a photo where to measure ... the current ... ?
beside each chip there is a bank of 8 small capacitors. most of these (seems like half of them in my situation) carry the voltage to the chip (0.73-0.85v = 1.5-2.7GH depending on tuning) and can be easily checked with a multimeter. I usually check at least 2 caps to verify that all chips receive similar voltage and ensure accuracy. I find that on startup from cold pcb i have around 0.805v, which climbs to 0.825v after the system is running and getting toasty over an hour or so. during this time, error rates decrease slightly and hashrate climbs to optimal speeds of about 40-41GH for the board ps: current (amperage) is different from voltage |
|
|
|
on that note, would love to see an update from dave on the october equipment and the upcoming in-stock stuff/sales soon. I have an h-board on order #13X im waiting for, and then would like to start filling up the other 14 slots of my kit when prices make it a reasonable investment
The new stuff will not work in your M boards. You'll have to buy new M cards for November v2 h boards. So if you did not fill them up with October orders, you can forget about it. I wanted to fill up my two August kits (v1 and v2 M boards) but I cancelled orders for 30 h cards because of the price. Now, it looks like v1 and v2 will not be available anymore. So whatever you buy in November, buy the whole kit as you never know what versions will be available in the future (if any). i would be a little annoyed if that was the case. I see no reason why support for gen1 devices would stop so soon. At the very least, im sure a few 3rd parties will show up with h-board designs for sale (i know there was a mini 8-chip design posted in the forums last week) |
|
|
|
Nope Nope Nope.
A 5970 consumes 200 watts and gives 700 megahashes. You lose 24 cents per hour, assuming you pay $0.10 per kWh
A 6990 consumes 200 watts and gives ~750 megahashes. You lose 22 cents an hour, assuming you pay $0.10 per kWh
So.... 50 of those loses you $11.50 every hour you run it, $23.00 for 100 of them. Even with free power, you'd only make $12.50-$25.00 a day.
what the hell are you talking about? your math is off by a factor of 10. a kWh means 1000w used for a full hour. a 200w card uses 0.2 kWh = 2.2c of electricity at $0.10/kWh used for scrypt mining litecoins, a 5970 and 6990 both produce around 700khash/s, which is over a dollar per day, minus roughly $0.30 meas a profit of around 60-80 cents per day per card. Assuming difficulty stays fairly constant (unlikely), it would take a year or more to pay off costs. In reality, it will likely take closer to 1.5-2years if difficulty rises at a minimal pace |
|
|
|
 Still some manual work to be done so that autotune can be turned indefinetly off, but Im fairly certain there is not much more to come out of these. [edit] Oh and thats a stable and constant hashrate. ...and before someone asks, each board is pencil modded individually with trial and error method. Rev. 1 boards as low as 1.15k and rev. 2 at as low as 1.6k. personally, i measure the voltage across the caps rather then pencil mod resistance, since that can change as the unit heats up and its hard to measure the resistor when running and plugged in. |
|
|
|
|
why would a pool operator decide to join another pool that would then charge a fee? It would just mean 1-3% less income for everyone (operator and individual miners). The only benefit would be to increaase the frequency of found block to reduce the variance of payout frequency (payout amounts would still be averaged the same though, minus the fees of the "top pool")
|
|
|
|
Wasn't GHash.io a victim of this attack too?  Were they? They were solving a *ton* of blocks during the attacks. My understanding is their website went down when all of Cloudflare had a major outage, but the website is just a frontend. I read a few here-say posts that said they were being attacked. But I don't know if it was fact. So, is there possibly some truth to this rumor, Eleuthria? the ghash.io hosted stuff went down for an hour or so, and my own equipment using the pool went down for about 2-3 hours. So yeah, seems kind of silly to go pointing fingers at the kid who [also] fell off the playground. |
|
|
|
If any of you have been following my attempts to get a refund issued on an order for a 400GH October kit ...
I would like to report that buzzdave came through, and the whole thing probably would have been resolved sooner had it not be for the forum outage.
So it's done, I got my funds back, and I'm happy.
My October orders were cancelled as well. Now it is up to Dave to send me payment for my August Kit. We supposed to get BTC1.25, right? correct, or 2.25BTC in store credit. My understanding is those who explicitly requested or made clear they want the refund have begun to receive thiers, while those who chose credit or made no choice yet will likely get store credit update/info soon, as they expect to have in-stock sales around november. on that note, would love to see an update from dave on the october equipment and the upcoming in-stock stuff/sales soon. I have an h-board on order #13X im waiting for, and then would like to start filling up the other 14 slots of my kit when prices make it a reasonable investment |
|
|
|
So far out of my full unit I have one totally fucked up board, 5 chips work, 8 are dead, 3 are so dead they dont even show up in the log.
I have 4 other boards performing terribly I need to try and sort out. Hopefully i can atleast get them to 25gh/s via manual tuning.
Sigh...
In my experience, sometimes a little bit of time running will awaken dead chips or improve mediocre chips. My starter kit originally had a chip that responded at about 0.6GH with 20-40% errors for several days and reboots. However, after a pencil mod, it suddenly began working perfectly (now getting >40GH on the board). Try doing a bit of voltage increase (0.78V is a good place to start if you have no heatsinks, 0.8V if you do) and see what happens. for the dead/missing chips, there are some reports on removing its connections to "see" the following chips in the chain. (however, a replacement board is a better alternative if punin can do it quickly) |
|
|
|
Haha I wanted to do the same with a few GPU's but I think the moisture will not do em good.
Is is quite an original idea tbh.
Don't worry, water only condense on an object which is colder than air, miners when running are warmer than the surroundings so there will not be condense. My net is to stop insect/flowers/snowflakes from being sucked into the intake of miners, it has large holes (2mm in diameter) The other day i learnt the [annoying] way that my PSU with a beautiful blue LED glowing fan intake is like a beacon for the only fly in the house. I had to stop the PSU every time the stupid bug got sucked up into it and wait for the fly to leave before restarting. After about 4 of these in an hour, i gave up and turned off the LED functionality  |
|
|
|
|
they are a 200TH private pool, selling/renting something like 3TH on CEX.io
the additional 35+GH is from people adding thier own miners. Ive personally pointed my 40GH bitfury and a 2GH block erupter cluster at the pool. (0% fees, frequent block finds)
|
|
|
|
thats like intel advertising a 3.6Ghz CPU (which everyone knows can be overclocked up higher) but in reality its only manages 1.83Ghz CPU once you put it into a motherboard. and intel replying, thats what customers should expect...
and now i tooo have bit..fury.
by that i mean, darn these bitcoin industry manufacturers... they say one thing, and mean another lol..
no offence to enterpoint as i know its not their chips...
.. still atleast these sticks are gonna be alot better then 333mhash
To be fair to him I don't think bitfury has ever advertised these as 5GH/s chips. It just happens to be written on the package, which presumably came before the testing phase that discovered their actual performance. It is a bit misleading, I admit, but I think **most** manufacturers are being honest about what they will actually achieve. 5GH/s was the idealogical speed of the chips when they went to fab, but in the real-world, the speed can not reach that. This was discovered and made clear as quickly as one would reasonably expect. The fact is thatr the chips are capable of: no heatsinks, air cooling: 1.5-2GH/s small heatsinks and air cooling 2.2-2.7GH/s large cooling apparatus: ~3.4GH/s on a single chip test board. its feasible that 4GH/s might be achieved by water cooling or other similar extreme cooling methods, but it has not been pushed that far. power efficiency at that speed would likely approach/pass 2w/GH |
|
|
|
12V is easy
What do these chips do?
i beleive that at least a portion of them are used to drop 12v down to 1.8v for comms, but im not sure. (there are about a dozen devs that could provide a better answer on this!) |
|
|
|
I would like to connect a h-card directly to my raspberry pi.
Is that easy?
probably not 1) you need to get 12V current into the h-board (the m-board holds these connectors) 2) the m-board has its own chips, its not just a replication of the gpio pins |
|
|
|
I'm confused. You're going to ship V2 H cards but not V3 M boards? V2 H cards work in V1 M boards?
- I have an August Starter with a V1 M board and 2 V1 H cards.
- I have orders for an October Starter Kit and 7 H boards.
What will you be shipping me and when? There are so many versions of M and V boards!
my understanding is that there are 3 revisions of M-board: V1: no PCI-E connectors, V1 H-boards V2: PCI-E connectors, uses V1 H-boards V3: not seen yet, but will use the V2 H-boards H-boards: V1: basic august design. will ship in october as well for returning customers who appear to be "filling" their kits V2: october design, compatible only with V3 M-boards, has adjustable voltage. october kits will almost certainly be the V3M and V2H devices paired together. h-board-only purchases will be as V1H cards |
|
|
|
How would i request: 1klondike...... ?
by going to: https://vanitypool.appspot.com/newWork& small case L is not possible, or upper case i or o so all caps isn't an option either, maybe 1KLoNDike or 1KLoNDiKE or 1kLondike - none of them really look right to me though. ahh, thanks for the heads up. I tried the site, but keep getting told my public key is too short. i tried with 2 different 'receiving addresses', should i be grabbing my key from elsewhere? |
|
|
|
|
Show Posts
