DPoS
|
|
January 15, 2014, 08:00:54 AM |
|
So they only have one prototype? That would mean that they may think hardware changes still have to be made.. if it was just firmware they would make a couple so you would get some 24 hour performance testing in
Also, that 1.62Th value is not pool side. Looks more like ~1.5Th. But without any screen shots from the pool who knows
Sure, lots of other miner companies are terrible, but Cointerra was touted as the Gods and they didn't give a detailed engineering report on what they think the troubles are.. HashFast at least did that even though they were doomed to fail
|
|
|
|
aerobatic
|
|
January 15, 2014, 09:06:28 AM |
|
So they only have one prototype? That would mean that they may think hardware changes still have to be made.. if it was just firmware they would make a couple so you would get some 24 hour performance testing in
Also, that 1.62Th value is not pool side. Looks more like ~1.5Th. But without any screen shots from the pool who knows
Sure, lots of other miner companies are terrible, but Cointerra was touted as the Gods and they didn't give a detailed engineering report on what they think the troubles are.. HashFast at least did that even though they were doomed to fail
a previous screenshot showed they had plenty of asics so i doubt they're limited to the one prototype.. but since they're making tweaks to it, maybe they can't spare any to leave hashing right now... especially if they make improvements every day. leaving it hashing serves no purpose at this time.. they're still in development... not yet in marketing mode... they're not trying to pacify existing customers... and they've stated they're not yet at their final performance numbers... so there is no reason at this time to leave a machine hashing (until they're done with the tweaking) we have to assume that we're not just talking about firmware tweaks as if it was only that, they probably would've shipped it and done their updates in the field like knc did... albeit knc did release a second gen board that hashed 20% faster using the same asics as the 1st gen board so thats a very real example of the benefit of hardware and board improvements.. which their 2nd batch customers were happy to receive (different dc/dc converters) and their 1st batch customers missed out on. my jupiters are batch 1 and hash between 450 and 550 gh. the gen 2 models hash at around 650 i believe. that 2nd gen board must be capable of improved power delivery. presumably thats what cointerra is doing right now... as the chips already exist, and the cooling system seems over specced and both have been proven to work so the main board change they can make are improvements to power delivery. -- Jez
|
|
|
|
Ytterbium
|
|
January 15, 2014, 12:09:28 PM |
|
So they only have one prototype? That would mean that they may think hardware changes still have to be made.. if it was just firmware they would make a couple so you would get some 24 hour performance testing in
Also, that 1.62Th value is not pool side. Looks more like ~1.5Th. But without any screen shots from the pool who knows
Sure, lots of other miner companies are terrible, but Cointerra was touted as the Gods and they didn't give a detailed engineering report on what they think the troubles are.. HashFast at least did that even though they were doomed to fail
a previous screenshot showed they had plenty of asics so i doubt they're limited to the one prototype.. but since they're making tweaks to it, maybe they can't spare any to leave hashing right now... especially if they make improvements every day. leaving it hashing serves no purpose at this time.. they're still in development... not yet in marketing mode... they're not trying to pacify existing customers... and they've stated they're not yet at their final performance numbers... so there is no reason at this time to leave a machine hashing (until they're done with the tweaking) we have to assume that we're not just talking about firmware tweaks as if it was only that, they probably would've shipped it and done their updates in the field like knc did... albeit knc did release a second gen board that hashed 20% faster using the same asics as the 1st gen board so thats a very real example of the benefit of hardware and board improvements.. which their 2nd batch customers were happy to receive (different dc/dc converters) and their 1st batch customers missed out on. my jupiters are batch 1 and hash between 450 and 550 gh. the gen 2 models hash at around 650 i believe. that 2nd gen board must be capable of improved power delivery. presumably thats what cointerra is doing right now... as the chips already exist, and the cooling system seems over specced and both have been proven to work so the main board change they can make are improvements to power delivery. -- Jez That doesn't make any sense at all. Maybe they could get a 20% performance boost by changing the board, but why on earth would they delay for months when they could get devices in the hands of their customers. I made way more money with my B2 Avalon then I have with my KnC Saturn, even though I got them just a few months apart. If they have chips and aren't shipping, they must have some clearly fatal flaw. Either with the chips or the boards.
|
|
|
|
testerx
|
|
January 15, 2014, 03:48:36 PM |
|
Well I'm hoping that once they've made whatever hardware level tweaks they feel they need to that they just ship it and then patch the firmware like KNC did to further optimize power draw. I'd rather have 1.7thash in hand ASAP then wait for the perfect 2thash. I doubt that this is am insurmountable issue, worst case they may have worse yields than anticipated but they can always just add more boards/chips in since the marginal cost is pretty minimal.
|
|
|
|
VolanicEruptor
|
|
January 15, 2014, 04:19:06 PM |
|
It takes about 10 days to decrease profitability by 20% I agree with the notion that they should just settle for 1.6 and give us firmware updates as they tweak. That makes good sense..
|
|
|
|
lucazane
Legendary
Offline
Activity: 1198
Merit: 1000
|
|
January 15, 2014, 04:19:58 PM |
|
It takes about 10 days to decrease profitability by 20% I agree with the notion that they should just settle for 1.6 and give us firmware updates as they tweak. That makes good sense..
Maybe it's hardware tweak ^^
|
|
|
|
VolanicEruptor
|
|
January 15, 2014, 04:21:32 PM |
|
It takes about 10 days to decrease profitability by 20% I agree with the notion that they should just settle for 1.6 and give us firmware updates as they tweak. That makes good sense..
Maybe it's hardware tweak ^^ I sure hope not, that would be a lot more work wouldn't it?
|
|
|
|
miahallen
Newbie
Offline
Activity: 48
Merit: 0
|
|
January 15, 2014, 04:56:22 PM |
|
It takes about 10 days to decrease profitability by 20% I agree with the notion that they should just settle for 1.6 and give us firmware updates as they tweak. That makes good sense..
Maybe it's hardware tweak ^^ I sure hope not, that would be a lot more work wouldn't it? Agreed!
|
|
|
|
newguy05
|
|
January 15, 2014, 07:12:59 PM |
|
those guys are really dragging their feet, it doesnt take 2+ weeks with still no end in sight. KNC did it in a few days, and any team with cointerra's experience on paper (if it is true) can easily do it in the same time. It's just trial and error unless they screwed up the design which i doubt.
/thumb down
|
|
|
|
Puppet
Legendary
Offline
Activity: 980
Merit: 1040
|
|
January 15, 2014, 08:31:56 PM |
|
those guys are really dragging their feet, it doesnt take 2+ weeks with still no end in sight. KNC did it in a few days, and any team with cointerra's experience on paper (if it is true) can easily do it in the same time. It's just trial and error unless they screwed up the design which i doubt.
/thumb down
Chip bring up and characterization, PCB and firmware tuning typically takes months. With a bitcoin miner Im sure that can be done faster, but what KnC did was exceptional and literally, an exception, not the rule. (The result was also a lot of DOA's and at least one customer seeing 100% failure/issue rate on a batch of 10 miners. Thats not what you want as a company). Also keep in mind how long it took Avalon, BFL, Bitfury how long it is taking HF and Bitmine. If I were a customer, I would be disappointed that they are not only a few weeks behind schedule, but also still below target atm, but considering they received their first chips only 2 weeks ago, I wouldnt have expected miracles.
|
|
|
|
Syke
Legendary
Offline
Activity: 3878
Merit: 1193
|
|
January 15, 2014, 11:44:08 PM |
|
It's very strange that no one has been able to product the working stats page from Eligius. This is a strong indication that there are problems and they are unable to leave the unit running.
|
Buy & Hold
|
|
|
gmaxwell
Moderator
Legendary
Offline
Activity: 4242
Merit: 8684
|
|
January 16, 2014, 01:37:34 AM |
|
It's very strange that no one has been able to product the working stats page from Eligius. This is a strong indication that there are problems and they are unable to leave the unit running.
They didn't leave it running. Though it doesn't look like problems— it looks like they ran it there just for the demo. http://eligius.st/~wizkid057/newstats/userstats.php/1CTtm4iiwqt35Rgew1DQW6YoSwN4bKNopf
|
|
|
|
VolanicEruptor
|
|
January 16, 2014, 01:43:24 AM |
|
Very interesting that they wouldn't just point it to the same place they have been testing the whole time. I wonder why!
|
|
|
|
gmaxwell
Moderator
Legendary
Offline
Activity: 4242
Merit: 8684
|
|
January 16, 2014, 01:51:24 AM |
|
Very interesting that they wouldn't just point it to the same place they have been testing the whole time. I wonder why! Why do you even assume they are testing (much) on the production network? Up-thread I was arguing that doing so is a poor choice both for technical (not a great test case, noisy and doesn't test things like big blocks) and professional (competing with your customers is bad mojo) reasons.
|
|
|
|
VolanicEruptor
|
|
January 16, 2014, 01:55:03 AM |
|
Very interesting that they wouldn't just point it to the same place they have been testing the whole time. I wonder why! Why do you even assume they are testing (much) on the production network? Up-thread I was arguing that doing so is a poor choice both for technical (not a great test case, noisy and doesn't test things like big blocks) and professional (competing with your customers is bad mojo) reasons. Maybe do half and half? That way they make millions of dollars more profit for themselves. Wouldn't you?
|
|
|
|
Syke
Legendary
Offline
Activity: 3878
Merit: 1193
|
|
January 16, 2014, 02:54:05 AM |
|
They didn't leave it running.
Yeah, why? Overheated? Chip fried? Hashing failed? Caps blew? For a device that needs to run 24/7/365, the fact that they only had it running for mere minutes on Eligius is highly troubling.
|
Buy & Hold
|
|
|
plato14
|
|
January 16, 2014, 03:19:07 AM |
|
So what is the consensus here, should we be concerned about late machines that also underperform?
|
|
|
|
JoseSan
Member
Offline
Activity: 117
Merit: 10
|
|
January 16, 2014, 06:22:57 AM |
|
So what is the consensus here, should we be concerned about late machines that also underperform?
Speculating about what the timing and numbers behind the latest news release mean by a bunch of yahoos like us isn't going to produce much. I'll bite, though. I'm guessing this was a guy taking a little time out of the normal stress testing electronics go through: http://en.wikipedia.org/wiki/High-temperature_operating_lifeKNC probably skimped on this a little, and thus had failure rates higher than consumers are used to. (It's super rare an off-the-shelf motherboard has capacitors that burn). The ambient temperature of 55C seems a bit strange to me, but it's quite possible a temperature sensor on the board is seeing this, or they're trying to operate the chips at higher than normal operation (e.g. 125C). I wouldn't want a firmware released that can permanently fry the chips, or rather, I'd like to know where the 'risk zone' is with respect to temperature, voltage, current, etc. As for the hash rate, one thing I noticed was that there are zero reported hardware errors in the screenshot. Granted, it's only 60s of hashing and it's most likely the software not reporting things; or, more optimistically, they're really underclocked. They know every week counts. They're in this for the long haul (they hope to be "the Intel of bitcoin" in Iyengar's words).
|
|
|
|
testerx
|
|
January 16, 2014, 06:49:10 AM |
|
So what is the consensus here, should we be concerned about late machines that also underperform?
Speculating about what the timing and numbers behind the latest news release mean by a bunch of yahoos like us isn't going to produce much. I'll bite, though. I'm guessing this was a guy taking a little time out of the normal stress testing electronics go through: http://en.wikipedia.org/wiki/High-temperature_operating_lifeKNC probably skimped on this a little, and thus had failure rates higher than consumers are used to. (It's super rare an off-the-shelf motherboard has capacitors that burn). The ambient temperature of 55C seems a bit strange to me, but it's quite possible a temperature sensor on the board is seeing this, or they're trying to operate the chips at higher than normal operation (e.g. 125C). I wouldn't want a firmware released that can permanently fry the chips, or rather, I'd like to know where the 'risk zone' is with respect to temperature, voltage, current, etc. As for the hash rate, one thing I noticed was that there are zero reported hardware errors in the screenshot. Granted, it's only 60s of hashing and it's most likely the software not reporting things; or, more optimistically, they're really underclocked. They know every week counts. They're in this for the long haul (they hope to be "the Intel of bitcoin" in Iyengar's words). Yeah I too noticed that there were no HW errors at all which is very impressive compared to the competition's ASICs. I'm sure there's still a lot of room for speed increases but they're probably trying to get the speed up without going over their power targets. Hopefully they succeed or I'm gonna have to pay for new power wiring before April, lol.
|
|
|
|
plato14
|
|
January 16, 2014, 03:20:42 PM |
|
Just breached over 2000MM on the next difficulty prediction, it will probably be much more for the next adjustment. There have been large spikes this month, several petahash...Hashfast is not shipping who is putting this hashing power on the network? Is it KNC?
|
|
|
|
|