how long is an estimated life span of an antminer s3?
how can anybody know this? it's a brand new product without almost any testing prior to shipment. they tested it for a a week....max |
|
|
|
Sorry, but I don't believe the final SP30 will be designed with two 1200W power supplies, if it was just that then 120v would be fully supported. I believe the wattage is going to be much higher, and that is why 120V is not recommend and probably not even worth looking at. What kind of magic happens at 208v on a 1200W power supply that a 120V on a 1200W supply cant give the same amount? It doesn't make ANY sense to me, and that is why I have been holding out, and will continue to hold out. I am thinking those power supplies would be 1800W-2000W, and now that would make MUCH more sense to me why you would want it on a 208V-240V circuit.
Perhaps I am wrong, and will have to eat my words. But please give me a reason why 120v would not work if I had two 20 amp circuits and the power supplies were only 1200W. Don't give me that stuff that a 120V circuit cannot supply enough voltage if its only 1200W. 120V*20amp=2400 - 25% = 1800W max on continuous load.
They said that they already sourced Emerson and Murata 1200W power supplies. There will be two per SP-30 (at least that was the plan) It should work even on two 15A circuits. 2X12(derated from 15)X110=2640W easily |
|
|
|
Deep night now in China,
I wanna make video tomorrow daytime.
We prepare 1 device for Dogie already.
If somebody in Shenzhen right now, can see and check working device.
i don't see anything like this (1.25TH) on the website-there is still pacific2000 rated at 2Th. |
|
|
|
Yes, I will get my July sp30 in the next days.
I am obviously not immune in my desire to speculate, as any naturally curious person isnīt.
OK, then. We can make a poll as to what TH Sp-30 will be at: In my opinion 4-4.5Th. If it is at 3-3.5 th-then it is a shocker. Thankfully opinions donīt change the hashrate a unit achieves (or my sp10 would already hash at 100TH). Letīs wait and see what spondoolies will tell soon. Ok, but then, let's not engage in speculating on who is responsible as well (TMSC or somebody at SPT)) |
|
|
|
Yes, I will get my July sp30 in the next days.
I am obviously not immune in my desire to speculate, as any naturally curious person isnīt.
OK, then. We can make a poll as to what TH Sp-30 will be at: In my opinion 4-4.5Th. If it is at 3-3.5 th-then it is a shocker. |
|
|
|
so ive got 4 units
2 hash at 437 and 441
2 hash at 418 and 415
emailed bitmain and they asked me for screenshots of the status page
I wonder, what they are going to do with this information? I'm surprised you got a response. I sent a message through their internal messaging system on the 26th and have not heard anything. Two units that won't hash stable over 415. Even offered to NAT two public ip addresses to the units so they could fiddle with them. so far my S3 units resisted any fiddling. One can be accelerated to 237.5, but then one chip drops out (X). I am going to check the unit today and see if, perhaps, there is no or too little thermal paste there. If not, i will just leave it alone at 218.75/441Gh. I just noticed my third unit hashing at 418 has an X, when i reboot it switches to " - " asic status.. resulting in my needed to do a hard power cycle from the PSU in my experience, a hard boot from PSU on S3 sometimes leaves them in a strange lock (pools are all mixed up, etc, etc) and it takes a software boot or two to even them out. However, if you ssh into it, you can issue the following: <enter/return> exit to get out of ssh and it will shutoff in 3sec (you can replace 3 with another number), with fans slowly decelerating over 20-30 sec psu OFF-ON boot after that is benign and does now screw up anything |
|
|
|
so ive got 4 units
2 hash at 437 and 441
2 hash at 418 and 415
emailed bitmain and they asked me for screenshots of the status page
I wonder, what they are going to do with this information? I'm surprised you got a response. I sent a message through their internal messaging system on the 26th and have not heard anything. Two units that won't hash stable over 415. Even offered to NAT two public ip addresses to the units so they could fiddle with them. so far my S3 units resisted any fiddling. One can be accelerated to 237.5, but then one chip drops out (X). I am going to check the unit today and see if, perhaps, there is no or too little thermal paste there. If not, i will just leave it alone at 218.75/441Gh. |
|
|
|
--snip
I can also share that we're zeroing on the issue and it's a TSMC production issue.
Future batches (October) will meet our desired spec.
Guy
This either means the chip is designed with a slight fault, or certain production methods at TSMC caused the chip to underperform. The comment really suggests the latter. Guy also seems to be fairly confident that it is not a packaging issue, and that it can be fixed for future chip batches (which would be great news.) If TSMC is at fault, spondoolies might get the new and improved mask at a discount. If it is more a question of implementing certain requirements posed by TSMC, it should really have been corrected during (initial) tapeout. I think that it is all speculation (Edit: I see that you added the note on the same after the initial post) until we know the facts. Were there some SP-30 that were sold as early (July) batch? If so, they would have to be shipped by July 30 and then we would know everything. |
|
|
|
so ive got 4 units
2 hash at 437 and 441
2 hash at 418 and 415
emailed bitmain and they asked me for screenshots of the status page
I wonder, what they are going to do with this information? |
|
|
|
|
everybody expects a bubble now, so since market is almost always acts in a contrarian fashion, there would be NO bubble.
$500-650 for two years. by the end of the first year, 70% miners or more would drop out.
|
|
|
|
Can anyone get a db meter and tell me how loud these things are. I currently have knc nov jups that are whisper quiet. I'm planning on unloading the kncs and replacing with more efficient miners (S3 prolly), but have to keep the noise to a minimum.
It depends on mhz, naturally at 212.5 it is VERY quiet and CONSISTENT (not much acceleration-deceleration)-57db at 1m, 63-64db at fans (my A/C is louder) at 218.75-it is 62 at one meter (when fans are at ~2-2.2K), dropping to 57db at one meter when fans are at 1.8-1.9K and 64-68dB at fans One is in my small office, it is tolerable, but mildly annoying at 218.75, but totally OK at 212.5 At 5-10 m distance you don't hear much. measurement with dB Meter Pro I wouldn't call them whisper quiet, though |
|
|
|
How can balances not yet rewarded/earned show up in the payout queue?
When work is given, CPPSRB has precalculated everyone's exact reward for if that work results in a block. So if it does, everyone is paid properly. The payout queue displayed is simply a periodic snapshot of this data which includes the rewards for the block not yet found... which can't be included in a manual payout. Related, what if the payout queue is completely emptied out by a manual payout, and (due to vagaries of the queue and people not yet hitting their threshold) there are not yet enough users in the queue for payouts to total 25 BTC in the currently-being-mined block? Is the remainder just sent to the cold wallet for future manual payouts? Or will the pool just pay out accounts before they hit their thresholds?
It goes to the cold wallet, yes. The only time the pool pays a balance before it reaches the minimum payout, either default or custom, is due to inactivity. Balances < ~0.001 BTC are never paid out. 1) So manual payouts are just normal transactions pushed to the Bitcoin network?
Yes. I.e., they often end up in blocks that aren't mined by Eligius? If so, it seems that it would be nice to just have that transaction included in the next Eligius-mined block so that the pool can keep the fees (equivalent to not paying fees). Obviously this could lead to a delay in payments if there is a long round, but I think that most people would understand. I usually just pay the fees, or send it fee-less and prioritize it on Eligius. 2) How often does Eligius push new block contents to miners? If the exact payout amount is fully paid out and kept current (not just the "as of last block" amount) in every block, this would have to be pretty frequent..... several times a minute? So if a miner successfully finds an "old" block (with "old" version of the coinbase transaction that has "old" balances), I suppose the new balance differentials are just kept in the user's account to be paid out the next time they enter the queue?
Each miner individually gets updates every 55 seconds or when a new network block is found. But work is constantly being generated and given to other miners on offset schedules. The result is up-to-the-second work being out there somewhere at all times. In reality, the share log is paid based on when the work was issued, not when the block is actually found/submitted. This prevents a few unlikely attack vectors like a user finding a block, holding on to it, then adding a bunch of hash power to the pool before submitting the block thinking those shares will be paid by the current round. In reality, they would be put towards the next round since shares submitted after the block-finding-work was issued isn't paid/rewarded by the block. (Edited a few times for clarity) Is eligius in OK status? Statistics on my miners is frozen for the last two-three hours-waiting for payout that just does not seem coming. it seems to occur on sundays for some reason. Comment? Edit: in the last two hours the amount of earned BTC actually went down with miner reporting normal speed. How is this possible? |
|
|
|
|
i have a strangest unit on most recent firmare.
At 212.5 speed is ~429 GH (after 12 hr), almost no HW errors, temp ~40
At 218.75, speed is ~420GH (after same time), almost no HW errors, same temp
At 225, speed is ~420GH after 12 hours (initially bumps to 455, then slowly drifts down), no errors, same temp
What exactly is the limiting factor here?
I am thinking that is is that darned Dc-Dc converter, so would be very eager to learn if someone got an improvement by attaching heat sink, etc.
|
|
|
|
 Overnight run  so far so good... ~450watts for ~500gh/s I think the massive HW spike is from internet outtage, it was at 40HW very early in the morning. Who knows. Either way it's minimal compared to 22,000 shares at 256 diff ( ~5mil shares ) Did you mod this machine? Looks awesome I see that your discards are very low. Did you change any other settings besides the frequency when you OCed? I am sure he did the cgminer parameter changes as many of us did, go back a couple of pages and read through the previous posts. I have read this entire thread over the last couple of weeks and I have been trying. What do I have to type into putty to access the parameter that need changing? We've covered this in depth today. The thread is messy, but this one liner will change what you need once you've SSHed onto your miner: sed -i 's/--queue 4096"/--queue 0 --scan-time 1 --expiry 1"/g' /etc/init.d/cgminer Go restart cgminer in the web interface by going to miner configuration and hitting save and apply. That will restart cgminer with the new config above. Ok, I ssh into miner, enter the code above, hit return/enter then (without exiting ssh?) go to web gui and just restart the miner by "save and apply" in configuration (although change was made by sshing) right? you cannot just use :wq or other command while still in ssh mode (after you entered the code string)? sed is basically doing what vi does. 1. SSH to miner 2. enter this sed -i 's/--queue 4096"/--queue 0 --scan-time 1 --expiry 1"/g' /etc/init.d/cgminer and press enter 3. exit SSH (CTRL-D) or exit 4. go do the web admin thing Duce came up with the idea on making the changes. His guide here: https://bitcointalk.org/index.php?topic=671189.msg8029279#msg8029279 goes the vi route. sed/vi accomplish the same task. We're just showing you guys different ways of doing it. The reason I used sed in my example is that I know that there is only one value `--queue 4096"` that will get changed and it's less daunting than vi. Great, thanks...not a coder, but I think I got it... big thanks to Duce and you. |
|
|
|
Overnight run so far so good... ~450watts for ~500gh/s I think the massive HW spike is from internet outtage, it was at 40HW very early in the morning. Who knows. Either way it's minimal compared to 22,000 shares at 256 diff ( ~5mil shares ) Did you mod this machine? Looks awesome I see that your discards are very low. Did you change any other settings besides the frequency when you OCed? I am sure he did the cgminer parameter changes as many of us did, go back a couple of pages and read through the previous posts. I have read this entire thread over the last couple of weeks and I have been trying. What do I have to type into putty to access the parameter that need changing? We've covered this in depth today. The thread is messy, but this one liner will change what you need once you've SSHed onto your miner: sed -i 's/--queue 4096"/--queue 0 --scan-time 1 --expiry 1"/g' /etc/init.d/cgminer Go restart cgminer in the web interface by going to miner configuration and hitting save and apply. That will restart cgminer with the new config above. Ok, I ssh into miner, enter the code above, hit return/enter then (without exiting ssh?) go to web gui and just restart the miner by "save and apply" in configuration (although change was made by sshing) right? you cannot just use :wq or other command while still in ssh mode (after you entered the code string)? |
|
|
|
Now lets talk about 100K. 100K would be a dream. Is this doable in the next 20 years? Otherwise I will be too old to enjoy my new found wealth. God damn I could be living the life right now, but nooo! I have to be stuck in his fucking shithole of a country. Lucky is literally everything.
If bitcoin is less than 4-5K in 20 years, then we would wasted our time with it. ~4K in 20 years corresponds to 10% compounded yearly appreciation. |
|
|
|
same thing happened i powered down the psu, plugged in the second s3 and then powered it back on, all came up, then as soon as they beep it shuts down the psu. could this be a problem with my jumper cable not being adequate enough?
is this single rail psu? most current in use by miners are single rail, like CX750M, CX500 or EVGA1300 |
|
|
|
Observations re S3 (vs S1)
S3 is quite a different beast. S1 was like a soldier. You give it an order to march at higher speed-and it does or "dies" trying-gives you XX.
One of my S3-I give it more juice (225 vs 218.75)-it goes up to 455, then back to the same 420 GH (while keeping all 0000).
Now trying 212.5, but it can do 430GH max on this-we will see.
There must be a controller (probably R47) that limits the performance no matter of what mhz you give it.
Cooling the case with an additional airflow from two fans directed at opposite sides did not help at all-I tried it.
However, i think that experimenting with heat sinks on R47 is a worthwhile endeavor.
I am wondering how the load average impact these, I ask because seeing the system at an average load average of 2.3 and 0% CPU cycles all of the time I wonder if cgminer is getting resource starved at all. On the surface it sure seems like it.. Very well may be. They should hire some cgminer developer as a consultant. This miner (S3) feels like it might need some software optimization. |
|
|
|
|
Observations re S3 (vs S1)
S3 is quite a different beast. S1 was like a soldier. You give it an order to march at higher speed-and it does or "dies" trying-gives you XX.
One of my S3-I give it more juice (225 vs 218.75)-it goes up to 455, then back to the same 420 GH (while keeping all 0000).
Now trying 212.5, but it can do 430GH max on this-we will see.
There must be a controller (probably R47) that limits the performance no matter of what mhz you give it.
Cooling the case with an additional airflow from two fans directed at opposite sides did not help at all-I tried it.
However, i think that experimenting with heat sinks on R47 is a worthwhile endeavor.
|
|
|
|
Overnight run so far so good... ~450watts for ~500gh/s I think the massive HW spike is from internet outtage, it was at 40HW very early in the morning. Who knows. Either way it's minimal compared to 22,000 shares at 256 diff ( ~5mil shares ) Did you mod this machine? Looks awesome |
|
|
|
|
Show Posts
