Bitmain's Released Antminer S9, World's First 16nm Miner Ready to Order

[kurbeks]

Problem is you are paying up front for 2 months to get regular miner like S9. It's not even somehting hot like GPU or L3's. L3 preorders are now aug/sept, While S9 is early august. Seems like bad buy vs L3+.

[ComputerGenie]

Problem is you are paying up front for 2 months to get regular miner like S9. It's not even somehting hot like GPU or L3's. L3 preorders are now aug/sept, While S9 is early august. Seems like bad buy vs L3+.

Yes, bananas are a bad purchase because walnuts stay fresh longer. 

[numnutz2009]

...ne way here r the pics:...

And here's to hoping that you save more in electric than the $35 you're giving away in reduced ROI.   gl2u


P.S. did you notice that you're pushing the middle board frequency higher than stock (aka overclocking).  

mkay let me correct u...

the middle board now has 8 hw errors in almost 24 hours. if you multiply that by 17 days and you get 136 hw errors which is far less than the almost 11k reported on the stock firmware. also now that the temps r lower the boards will have a better chance to last much much longer because they wont be soaking in their own temps lol. the repair of a single board is about $200 shipped to Denver so if your claim of me losing $35 over the roi times im completely fine with that considering i would lose much more if the board/boards die and i have to send one or more in for repair right? cant have any hashrate if i have no boards to mine with so that $200 turns into even more because of the 4th/s that will be missing from the miner the moment a board goes down.

you need the data sheet to determine the chips stock clock speed. without that you do not and cannot know what the default freq is for this model chip and since bitmain hasnt posted it anywhere that i (or ne one else based on posts made by others online) can see its safe to assume your guessing with your freq numbers lol. the miner has the same numbers as the stop firmware but the temps and freqs are now uniform. the miner runs 30 degrees (about) cooler than it did and the fans run faster now that i manually set them to run faster. 80% is still faster than they need to run at too. i can play around with the speeds for hours and hours until i find the right speed to keep them at for the maximum benefit (lower temps and uniform freqs).

also it might interest you to know that i used the firmware on another 11.85th/s miner but instead of 550M i used 600M and the fan speed is still the same as the first one so 80% and my hashrate after 16 hours is 12.875 th/s and the temps are holding steady at 78/71/74. even with the faster clock speed and increased hashrate the miner performs better than it did with the stock firmware keeping it at 100 degrees C constantly. i dont plan on running  the miners at 600M i just wanted to try it and see what your response would be knowing that more hashrate from the miner is possible and at temps well within the acceptable limits according to bitmain....at least i think 22 degrees lower than 100c is more than acceptable too dont you?? i wont be going to 650M though because thats just too much even for a test.

[ComputerGenie]

...P.S. did you notice that you're pushing the middle board frequency higher than stock (aka overclocking).  

mkay let me correct u...
...[blibityblablabla]...

I'm not sure how your think that error rate has to do with "correcting" me on pointing out that

Code:

550 > 477.47

But as I said, "I'm just to lazy to attempt any further explanations of electronics, mathematics, English, and/or thermodynamics to you", so I'll let you "win" in your own mind.  

[numnutz2009]

...P.S. did you notice that you're pushing the middle board frequency higher than stock (aka overclocking).  

mkay let me correct u...
...[blibityblablabla]...

I'm not sure how your think that error rate has to do with "correcting" me on pointing out that

Code:

550 > 477.47

But as I said, "I'm just to lazy to attempt any further explanations of electronics, mathematics, English, and/or thermodynamics to you", so I'll let you "win" in your own mind.  

477.47 isnt the default clock for that model chip.

[ComputerGenie]

...This pic shows the temps at their lowest using stock firmware...

...477.47 isnt the default clock for that model chip.

Make up my mind.

[PeaMine]

sooooo good news (at least i think it is)....i got the firmware with fan control . i downloaded the variations a few mins ago and will check my s9's sometime today to get the one with the most problematic temps to see how much of a difference the fan control makes.

i wonder if i will get that "sorry i was wrong" i was hoping to get?? pride is a great thing...but too much can ruin a man.

How has it been running?
I have about a 20% failure rate for the S9 boards, with summer coming up, I need all the cooling and downclocking I can get.

Also, now that the APW3 is sold out, what should we be using for the S9?
The 1300-2600 model is just short of powering 2 S9 properly on 220(It can, but at 100% usage almost), but a waste to only power 1.

[numnutz2009]

sooooo good news (at least i think it is)....i got the firmware with fan control . i downloaded the variations a few mins ago and will check my s9's sometime today to get the one with the most problematic temps to see how much of a difference the fan control makes.

i wonder if i will get that "sorry i was wrong" i was hoping to get?? pride is a great thing...but too much can ruin a man.

How has it been running?
I have about a 20% failure rate for the S9 boards, with summer coming up, I need all the cooling and downclocking I can get.

Also, now that the APW3 is sold out, what should we be using for the S9?
The 1300-2600 model is just short of powering 2 S9 properly on 220(It can, but at 100% usage almost), but a waste to only power 1.

the s9 using the freq 550M and with 80% fan speed is running smoothly at 11.85th/s

the s9 using the freq 600M and with 80% fan speed is also running just fine at 12.88th/s

both are the same batch same shipment s9. both are the same speed miner as well (11.85th/s) and both are holding up perfectly so far.

the only issue i have had was installing the firmware....it was a bitch and a half. i didnt know all 3 sets of pool info (not including the password field) needed to be filled out. i entered only one set and the miner wouldnt start mining. i tried multiple pools and the miner would display the same message the s9's doing the auto tune display before they start hashing. i even waited an hour to see if it was part of the checker but it wasnt. after resetting the settings to default and having the stock antpool info added i walked away to get a drink of water and a min or 2 later the miner started hashing. once i found that bit out i just filled in all 3 pool slots with info and the miner was hashing away. in the other s9's i have this wasnt required but this time it was so thats good to know in case they make this firmware live.

...This pic shows the temps at their lowest using stock firmware...

...477.47 isnt the default clock for that model chip.

Make up my mind.

we already discussed this....bitmain shows the stock freq as being whatever freq the miner ships with. if the freq is raised higher your considered overclocking the chip and your warranty is voided out right away. "my mind" is already made up. bitmain didnt release papers on this chip so you have no list of freqs bitmain tested it with and with no information of the chip u cant tell ne one what the default freq is for that chip. u tried to and failed. if u cant provide ne benefit to the community can u please just hush ur face a bit?? im sure others r interested in this info especially the folks in the same situation as the guy above this reply. summer is right around the corner and better cooling is a must have.

ik ur salty and upset that i got done what no one else was able to get done. i know ur upset that i was right and you were wrong....look man i totally understand ur animosity towards me and eveything i say but dont take away from others trying to better their setups simply because you cant accept the facts. thats what it boils down to bud. agree...dont agree....we heard u the 50 other times u said it cant be done....well it can and has been done so please stop grasping at straws and just drop it. ok??

[ComputerGenie]

....bitmain shows the stock freq as being whatever freq the miner ships with...

Yes, and in the image you showed, the middle board shipped with an average frequency of 477.47.

[NotFuzzyWarm]

....bitmain shows the stock freq as being whatever freq the miner ships with...

Yes, and in the image you showed, the middle board shipped with an average frequency of 477.47.

Since when not bickering over # of Angels dancing on a pinhead you both do have good info from time to time -- Unignored and had to pop in...  

You are both right... To me the whole main issue stems from the 16/14nm node process variability and the resulting wide range of speed/voltages the chips will even run at. That in part is probably why unlike all their other chips, Bitmain has not released any data sheets on the chips. The yields per-wafer are too variable. Could be outstanding, often mostly usable with decent performance followed by usable but 'meh'.

As a result, Bitmain bins (grades) chips per-performance and builds boards with different grade chips. eg, they know w-board built with grade-x chips will be happiest running at y-speed/voltage for z-expected board hash rate. As Genie keeps saying, speed/voltage data for each board (I will assume it is after testing each completed board!) is burned into the PIC at the factory. To me those settings stored in the PIC are the OEM specs for that board.

However - that data read from the PIC can/will be overwritten if desired either by Auto-tune or by manually changing the volatile cgminer.conf file or whatever Bitmain calls it. Awesome Miner monitoring software used to do it all the time via the cgminer API but is of course temporary - a reboot puts you back to what the Firmware has recorded in NV memory. All Auto-tune does is just verify that the PIC settings work, if not, it tweaks the settings for .conf  a bit and re-tests until it can work or decides to give up on the board and Fail it.

Anywho, Bitmain bins completed hash boards by hash rate, then later selects different speed boards to assemble into miners that produce advertised total hash rate. Again, some boards can be great, other will be 'meh'. Either way the miner produces as-advertised hash rate and Bitmain gets to sell more chips-per wafer. Maybe not the best solution to the usable chip-yield problem but it works.

If Bitmain is giving back our ability to once again set speeds and have it stick between reboots then it better be per-board otherwise Genie's technical nit-picking definition of over clocking (which I agree with) certainly comes into play. In that light it must be the user manually-setting the speed faster than what the OEM PIC data shipped with.

When doing the IP Reporter button trick to reset Firmware to the embedded backup image the one problem with it is: Sure, we get to manually set speed to out hearts content BUT it is just 1 speed setting for all boards. Ergo the best board has to run at what the worst one likes... Sure, can speed it up but then are OC'ing the slower board

Along the lines of this, anyone ever cherry-pick 3 outstanding boards from their miners and put them into 1 miner? I have a few s9 boards that together should do over 17THs but have yet to try it. Wonder if Bitmain has code to prevent that? Each board should be running at or near the PIC data settings but if the miner in total is running faster than as-shipped, is that over clocking?

[ComputerGenie]

...Along the lines of this, anyone ever cherry-pick 3 outstanding boards from their miners and put them into 1 miner? I have a few s9 boards that together should do over 17THs but have yet to try it. Wonder if Bitmain has code to prevent that?...

We do it all the time. We have a "bad boy" rack that has rigs that run from 6.8TH to ~11TH (if memory serves me right on the top number) as a result of all of our good racks being 12+TH. The "bad boy" rack is all culls and the rest are picked (much in the same way BM does) to get at or above our chose threshold. If they have such code in the software as to disable doing so, I haven't seen or experienced it.

...Each board should be running at or near the PIC data settings but if the miner in total is running faster than as-shipped, is that over clocking?

Even by my "technical nit-picking definition", the answer would be "no", because the board is still set to what the board was set to. In the strictest of terms, this would be like asking if replacing a 1080 GTX with a 1080 GTX Ti is overclocking your computer.

[NotFuzzyWarm]

...Along the lines of this, anyone ever cherry-pick 3 outstanding boards from their miners and put them into 1 miner? I have a few s9 boards that together should do over 17THs but have yet to try it. Wonder if Bitmain has code to prevent that?...

We do it all the time. We have a "bad boy" rack that has rigs that run from 6.8TH to ~11TH (if memory serves me right on the top number) as a result of all of our good racks being 12+TH. The "bad boy" rack is all culls and the rest are picked (much in the same way BM does) to get at or above our chose threshold. If they have such code in the software as to disable doing so, I haven't seen or experienced it.

...Each board should be running at or near the PIC data settings but if the miner in total is running faster than as-shipped, is that over clocking?

Even by my "technical nit-picking definition", the answer would be "no", because the board is still set to what the board was set to. In the strictest of terms, this would be like asking if replacing a 1080 GTX with a 1080 GTX Ti is overclocking your computer.

And since the miner(s) has been modified anyway, OC does become a moot point as far as BM is concerned
Also did an edit to ^ post: ja, dinna clearly state it before but your def of OC is correct. It has to be based on PIC settings per-board as shipped by the OEM. Auto-tune may fiddle with it to go faster (or slower) but that is BM's concern to set bounds for how much it (Auto-tune) can deviate from OEM without risking damage to the chips.

I like the idea of putting low performance boards all in their own um, gimpy miners... Any power savings from the lower speeds? Could be a nice way to help balance power phase loads...

[ComputerGenie]

...Any power savings from the lower speeds? Could be a nice way to help balance power phase loads...

No real "savings", given that they are running at whatever they came with, so the overall usage is still what it would be if all boards were in their original containers. The rack mostly started because of a joke about 1 bad board that read at ~27,000GH/s (RT) due to a single bad chip and then went from there based on my OCD of not wanting to see a 3,500 board in the same box as 2 @ 4,700+ (it's a personal spare-time hobby of mine that gets done on "blow-out day").
As far as load balancing, I leave that to the electrician, that's what he gets paid for.  

[AriesIV10]

COME ON BITMAIN - We know you are a business, We know you need cashflow, We know you are the only source, BUT please do not take our money an hold the product!  We need the product.  OVER 2 Months before delivery??

[ComputerGenie]

COME ON BITMAIN - We know you are a business, We know you need cashflow, We know you are the only source, BUT please do not take our money an hold the product!  We need the product.  OVER 2 Months before delivery??

You don't need it, you want it. Setting that aside, pre-orders aren't a new thing in this world. 

[HagssFIN]

COME ON BITMAIN - We know you are a business, We know you need cashflow, We know you are the only source, BUT please do not take our money an hold the product!  We need the product.  OVER 2 Months before delivery??

You don't need it, you want it. Setting that aside, pre-orders aren't a new thing in this world. 

Pre-orders are a shitty, cancer thing in this world. Period.

[ComputerGenie]

Pre-orders are a shitty, cancer thing in this world. Period.

I don't deny that, but there is a precedent for Bitmain doing it (and an extensive list at that); it's not like this one batch of 13.5's is where they 1st started doing this and it justifies this kid throwing his toys all around the room.

[HagssFIN]

Yeah, Bitmain use to be cool with close to no pre-orders.
Maybe they calculate more profit with this new regular pre-order style.

[ComputerGenie]

...Maybe they calculate more profit with this new regular pre-order style.

TBH, I'm thinking it has more to do with not further flooding the market (which would reduce current product value). But that's just my opinion (for what that's worth).

[HagssFIN]

...Maybe they calculate more profit with this new regular pre-order style.

TBH, I'm thinking it has more to do with not further flooding the market (which would reduce current product value). But that's just my opinion (for what that's worth).

Yes, good point as well.