My board PIC FW and heat sink and fans were upgraded yesterday. New photo of new heatsink/fans please! 45A = 540W is high for 28nm though! 0.5GH/1W is like an AntMiner but that has a huge heatsink to dissipate that heat. I'm not a fan of the Sunon 4500rpm leaf blowers. Can't wait to get my hands on the Hex8A1 and see if I can make it quiet. I like a challenge! I guess you can always throw more chips at the GH problem like AntMiner did if you can make them cheap enough. It's a shame it wasn't a 16x chip board then it would really have rocked.  Gator I am the kind who like to be insured. So 45a is at least 20% on top If you want 1w/GH you can always have it undervolting/clocking the board no problem Yeah, if I can get 200GH out of them with quiet fans somewhere around 0.7GH/1W (like the Bitfury) I'll be very happy.  I'll post in "hands on" with the results when I get mine. A friendly advise start looking for two quiet fans for every board you have ordered.  |
|
|
|
My board PIC FW and heat sink and fans were upgraded yesterday. New photo of new heatsink/fans please! 45A = 540W is high for 28nm though! 0.5GH/1W is like an AntMiner but that has a huge heatsink to dissipate that heat. I'm not a fan of the Sunon 4500rpm leaf blowers. Can't wait to get my hands on the Hex8A1 and see if I can make it quiet. I like a challenge! I guess you can always throw more chips at the GH problem like AntMiner did if you can make them cheap enough. It's a shame it wasn't a 16x chip board then it would really have rocked. Gator I am the kind who like to be insured. So 45a is at least 20% on top If you want 1w/GH you can always have it undervolting/clocking the board no problem |
|
|
|
Morning fellows, My board PIC FW and heat sink and fans were upgraded yesterday. What I can tell you it rocks. Here are my results after 14+ hours nonstop mining  Some things to point out: 1. My stats are not settled yet, because I am mining at 512 diif + filtering out nonces inside ASIC chips which do not meet it. The board is making 270GH stable and there is some room for overclocking. BUT I do not feel brave enough to do it and I suggest to all of you not to push it too much. We are hiiting the limits of hardware especially voltage regulators so overdriving them for a long period is not good idea at all. This is my personal opinion. I do hope that there will be precise measurement of power consumption very soon. I am using single rail psu rated 55A/12V. I think the bottom line is at lest 45A, to be on a safe side 2. Tplink+cgminer worked rock stable without an issue. My pool reject rate is 0.12% only which is golden! I am not commenting my HW error rate at all Technobit team did their job just perfect  3. What I can tell you that all of you should be able to get the 260Gh advertised at clock around 260 and voltage around 1000 mv. You have to experiment a little bit to find your sweet spot though. More updates to come |
|
|
|
@ckolivas as promised my impression, the new cgminer runs very well but the hardware error remain at approximately around 1%, the clock rate does not matter if 1.6 or 2.0!  greets 1% and you are not happy wired  It is absolutely normal dude. You are golden. Hw errors are always there no matter what you do. |
|
|
|
No more locking issues works perfect Thank you Con! |
|
|
|
Uploaded 3.12.3 with the correct fix now, reused same announce post.
Thank you Con! I will give it a shot and report back if get work locks are gone. However I see that you still write in last_device_valid_work get_work without a lock. Is it ok? Every where you use locks when writing to it but not in get work Thank you For that particular statistic it's pretty harmless and it's in get_work() for which the code needs to be as lean as possible. The lock protecting it elsewhere, stats_lock, is a global lock potentially contended by quite a few threads so I decided against it. Thank you! I will post back my results in a couple of hours |
|
|
|
Uploaded 3.12.3 with the correct fix now, reused same announce post.
Thank you Con! I will give it a shot and report back if get work locks are gone. However I see that you still write in last_device_valid_work get_work without a lock. Is it ok? Every where you use locks when writing to it but not in get work Thank you |
|
|
|
Hi,
If the Klondike 16 PCB takes 16 Avalon chips, and Avalon state 1.5-1.6 GH/s (Overclocked) why aren't people doing this as surely they could be getting 25.6 Gh/s (16 x 1.6) per board , whats stopping them?
Thanks
They do for fun not for business that is what I think. And you need to know a lot to be able to do it even though all the info is public. Besides bkk was the inspiration and hart of Klondike an unfortunately he stepped out of the project |
|
|
|
You have a chance for a refund, and even gains now. Just got to sell the thing you asked them to create for you, when you made the purchase, after it is delivered. I am sure you can get $12,000 for it, once it is sitting in your hands. I see people have sold BFL's $1,000 units as high as $10,000. So anything is possible.
I agree here, if you don't think you can recover the money by mining BTC, then sell the unit the moment you get your hands on it. You will get more in return. That is likely what I will also do. See, 1.2 THs is state of the art, the only thing better out there is Cointerra and KNC. But you can't get those until March or April. And bitfury dude in terms of power efficiency. We are not discussing prices at al |
|
|
|
Unless you prefer eu based company technobit I can assure all of you that I am having eight chip coincraft board in hand hashing like I beast 240-250 gh And pulling around 3W/GH I believe, albeit overclocked, but without even accounting for the PSU inefficiency. Based on the same chip, AMT are still claiming 0.5-0.75 W/GH at the wall, and thats up from their previous ludicrous claim of 0.25-0.5W/GH. To quote AMTs favorite shill: Yes, I do believe the unit will produce up to 1.2THs, running around 600Watts.
Dirty lie dude it is same chip. If you wanna efficient miner go for bitfury miner or under lock coincraft. If they do something at all it will be a trade of between power and hashing rate. The math is very simple dude 1. Chip supplier gives you a figure of power at one volt or whatever chip power is 2. The above figure is taken from simulations not real chips and differs with let say 10 % from reality. When chip is Overclocked the difference is quite big 3. Add at least 10% for cd/cd 12v to 1 and on top 15at least psu loss There is one slight chance for them to meet their figures if they putt wise as much chips as planned and under clock them. It is too late to do the math right know but they will roughly loose 60 chips costing 100 used each to make it happen and so on and on Again it will be not their fault the chip needs power that is all And they are aware of the fact they speak nonces unless instead of 220/120 v in your plugs you happen to have ONE VOLT dc It is likely the 1.2 THs units will be around 1.1 THs instead of the published. Cointerra TerraMiner IV was advertised at 2 THs and only delivered like 1.8 THs. A 10% drop from original estimates. This of course depends if the manufacturer fixes the design to add more chips. Cointerra has much larger chips so it is unable to adjust the design, however bitmine and AMT have a lot of smaller chips, so if they add 10% more, they can hit their targets. Dude, 1.2THs, running around 600Watts.If the quoted 600Wats are at the wall (220V better scenario) I can make 10 BTC bet that they will not make it!!! never! I am talking about mass shipped units. If they put 300 chips (instead of 40-50) undervolted and underclocked there is a slight chance this to happen. But they will not do it on production units for sure because this costs a lot of money! Just find a reputable third party person like john to hold our coins Who has the guts to earn some quick money Anyone? The best dreamed scenario will be arround 1.2 KW at the wall for 1.2T unit PERIOD!!!!!!!! At 20GHs per chip, they will need 60 chips. So if they claim .6 W per GHs they are talking around 720 watts. So 20 chips on top means around 1500 USD i am wandering how many chips will they need 600Watts Any way....I am still taking bets Some info: http://bitmine.ch/?page_id=863Power usage of 0.35 W/GH in low power, 0.6 W/GH in nominal and 1 W/GH in Turbo mode Supply voltage of 0.65V in low power, 0.765 V in nominal and 0.85 V in Turbo mode Please note that while we took the highest possible care in doing accurate and conservative simulations, the real performances of the final product could vary by a +/- 20% You know that 0.6 W/GH is at 0.765 V in nominal voltage right. Best case scenario 0.6w*1.15 (PSU loss 220v to 12V) *1.1 (DC/DC loss from 12V to .765 V i ) * 60 chips as you suggest * 20 GH each EQUALS TO 910 WATS And the above calculation does not take into account PCB losses at all. Then tell me about GOLD PSU, SUPER efficient dc/dc converters and so on The beauty of it with each added chip they "increase the power efficiency" decreasing the performance of ALL chips They can not do it or they can but they will be not able to cover their costs because of additional chips added and more complex design. Which means guess what |
|
|
|
WARNING: There may be a problem with cgminer 3.12.1 that makes it stop retrieving work. I will investigate further and possibly release a fix in the next 24 hours. In the meantime I suggest users do not upgrade.
Sigh.
Super! I have noticed same. Good news are that commenting commit is solving the problem for me for last hour or so. PS my hanged again  the only thing from what i can see since my stable build which have changed and may be related https://github.com/ckolivas/cgminer/commit/c4267e663aa7b4e7c2cd997fbe3c45bf87365f13but the above commit seems good to me |
|
|
|
Unless you prefer eu based company technobit I can assure all of you that I am having eight chip coincraft board in hand hashing like I beast 240-250 gh And pulling around 3W/GH I believe, albeit overclocked, but without even accounting for the PSU inefficiency. Based on the same chip, AMT are still claiming 0.5-0.75 W/GH at the wall, and thats up from their previous ludicrous claim of 0.25-0.5W/GH. To quote AMTs favorite shill: Yes, I do believe the unit will produce up to 1.2THs, running around 600Watts.
Dirty lie dude it is same chip. If you wanna efficient miner go for bitfury miner or under lock coincraft. If they do something at all it will be a trade of between power and hashing rate. The math is very simple dude 1. Chip supplier gives you a figure of power at one volt or whatever chip power is 2. The above figure is taken from simulations not real chips and differs with let say 10 % from reality. When chip is Overclocked the difference is quite big 3. Add at least 10% for cd/cd 12v to 1 and on top 15at least psu loss There is one slight chance for them to meet their figures if they putt wise as much chips as planned and under clock them. It is too late to do the math right know but they will roughly loose 60 chips costing 100 used each to make it happen and so on and on Again it will be not their fault the chip needs power that is all And they are aware of the fact they speak nonces unless instead of 220/120 v in your plugs you happen to have ONE VOLT dc It is likely the 1.2 THs units will be around 1.1 THs instead of the published. Cointerra TerraMiner IV was advertised at 2 THs and only delivered like 1.8 THs. A 10% drop from original estimates. This of course depends if the manufacturer fixes the design to add more chips. Cointerra has much larger chips so it is unable to adjust the design, however bitmine and AMT have a lot of smaller chips, so if they add 10% more, they can hit their targets. Dude, 1.2THs, running around 600Watts.If the quoted 600Wats are at the wall (220V better scenario) I can make 10 BTC bet that they will not make it!!! never! I am talking about mass shipped units. If they put 300 chips (instead of 40-50) undervolted and underclocked there is a slight chance this to happen. But they will not do it on production units for sure because this costs a lot of money! Just find a reputable third party person like john to hold our coins Who has the guts to earn some quick money Anyone? The best dreamed scenario will be arround 1.2 KW at the wall for 1.2T unit PERIOD!!!!!!!! |
|
|
|
|
Con,
There might be an issue caused by fa26f8df82f9213cffd8a715ac1450d3c191d459 commit
mutex_lock(&stats_lock);
cgpu->last_device_valid_work = time(NULL);
mutex_unlock(&stats_lock);
I got some locks recently. One possible reason might be.....that everywhere last_device_valid_work was changed under lock, but commit does not lock it
if (diff_t > 0) {
applog(LOG_DEBUG, "Get work blocked for %d seconds", (int)diff_t);
thr->cgpu->last_device_valid_work += diff_t;
}
I am commenting it to see if locks will appear
If i wrote something stupid pls do excuse me
Best
|
|
|
|
I know this is a bit off topic but still relates to the mining off the S1 but why is the BTC price crashed ?
Why do exhanges like local bitcoins under cut the bloody market we should stay at a steady level. Maddness
uhhh... russia banned btc! Ops....Will they increase the BTC price for their miners My bet is yes!!!!! |
|
|
|
Unless you prefer eu based company technobit I can assure all of you that I am having eight chip coincraft board in hand hashing like I beast 240-250 gh And pulling around 3W/GH I believe, albeit overclocked, but without even accounting for the PSU inefficiency. Based on the same chip, AMT are still claiming 0.5-0.75 W/GH at the wall, and thats up from their previous ludicrous claim of 0.25-0.5W/GH. To quote AMTs favorite shill: Yes, I do believe the unit will produce up to 1.2THs, running around 600Watts.
Dirty lie dude it is same chip. If you wanna efficient miner go for bitfury miner or under lock coincraft. If they do something at all it will be a trade of between power and hashing rate. The math is very simple dude 1. Chip supplier gives you a figure of power at one volt or whatever chip power is 2. The above figure is taken from simulations not real chips and differs with let say 10 % from reality. When chip is Overclocked the difference is quite big 3. Add at least 10% for cd/cd 12v to 1 and on top 15at least psu loss There is one slight chance for them to meet their figures if they putt wise as much chips as planned and under clock them. It is too late to do the math right know but they will roughly loose 60 chips costing 100 used each to make it happen and so on and on Again it will be not their fault the chip needs power that is all And they are aware of the fact they speak nonces unless instead of 220/120 v in your plugs you happen to have ONE VOLT dc |
|
|
|
|
Remove enable bitfury and avalon also
|
|
|
|
how do I do it exactly for me it does not work can I HEXMINER 2 or 3 start times and different pool register?
If I understand correctly Yes it is possible to start 2 or 3 copies of HEXminer and to run 3 different boards as an example to mine in 3 different pools. use cgminer with load balancing or/and quota dude  |
|
|
|
A few question for Loshia and Marto:
#1 Its possible to ship it out on 10. feb?
#2 Can you say me a configuration for cgminer? Is there something like this: --hexminer8-options x,x,x ?
Thanks!
First question is for marto about second one: --hexminer8-set-diff-to-one 1 --hexminer8-chip-mask 255 --hexminer8-voltage 955 --hexminer8-options 8:240 One by one --hexminer8-set-diff-to-one 1/0 the diff when set to 1 like mine - am testing it now brings a lot pressure between UB-HOST and PIC/ASIC communication Coin crafts can filter their response based on difficulty configured. When you set to one you got all nonces match diff 1 which has to be filtered and not send to the pool amusing you mine at diff greater than one. The only real advantage setting it to one is the cgminer 5 secs stats not moving up/down like crazy and average stats are settling down more quickly. Assume that you mine at diff 128. Cgminer sets it individually for each work passed to asic so it stays in sync with your pool always. Even if it is 1! When you set it to zero - ASIC send to the pic only nonces matching the diff 128 and back to host so you have about 128 times less data transfer, and cgminer shorts the time for submitting the nonce not checking against pool diff if it is a HW and all nasty stuff inside. the only disadvantage is that each nonce is counted not as one like before but as work difficulty itself which causes the stats variation. I suggest to use --hexminer8-set-diff-to-one 0 and configure your worker to mine at some constant higher difficulty. for instance when btcguild changes difficulty some old work may become stale oldddiff even though it is valid. In general if you happen to be in between two pools difficulties and your pool changing it up and down frequently you can have higher pool reject ratio - not good --hexminer8-chip-mask 255 - this was for development you can enable/disable chips based on binary mask bin 11111111 - 255 dec all chips enabled. Obviously zero is not a good choice here --hexminer8-voltage 955 the core voltage in milivolts. Do not go above one volt 1000 or do it on your risk --hexminer8-options 8:240 8 is chip count - used for some internal usb timing calculations , 240 is the frequency. The actual chip frequency is multiplied by four so 240 means 880 actually. the maximum value as far as know is 275. But try to stay below 260 in general And finally use --quiet when you have seen enough of course |
|
|
|
The first prototype with one direction of both heatsinks do not have enough surface for the top.
So we had to switch, sorry Gator Another reason for aligning the heatsinks in the same direction, not just that I like it that way, data centers who will want to run them in 4U rack cases would also need it that way too. They blow air in one side of the case and out the other side. True, But this works only in case. In open air as boards are it is useless. In the case you can force air to flow. For the data centers i am sure that they will approach marto and he will do as requested |
|
|
|
|
Show Posts
