Bitmain's Released Antminer S9, World's First 16nm Miner Ready to Order - page 210.

numnutz2009

sr. member

Activity: 387

Merit: 254

Quote from: HashSpike on November 11, 2016, 11:05:57 AM

Hello all,

Can someone please explain to me how my S9 had a spike to 20.03 TH/s ( 5 Min Hash Rate ). I have had other spikes to around 18+/- TH/s also so this is not a one time occurrence. I am no hardware expert, so someone please clue me in...but if this thing can hash for 5 minutes at 20.03 TH/s, then is it not possible to figure out how I achieved that hash rate and then modify to retain that hash rate? I am running 650 freq. with the November 07 firmware.

http://imgur.com/a/Z2X8L

it could just mean it had a streak of shares being submitted really fast to the pool. it happens sometimes. sames goes on the pool end...you will sometimes see large spikes pool side of speeds almost double what the actual miner speed is but it evens back out shortly after...nothing that comes close to lasting for hours.

NotFuzzyWarm

legendary

Activity: 3822

Merit: 2703

Evil beware: We have waffles!

Quote from: gt_addict on November 11, 2016, 12:13:34 PM

Ive got three gecko usb sticks that consistently pop up with 18Ghs (set to 12ghs) but they don't reach it regularly, infact they often drop to 8-9Ghs (even though still set to 12Ghs). That's why you have an average hashrate (which is the only one you should be worried about). The next 5 min hashrate could only be 10Ths on your machine so it evens out.

Spikes like that are very very common. Hell, my s9b1 which averages around 13.5THs often spikes up to 20Ths but again only very briefly. Do a refresh and #'s are more reasonable.

As far as I know, it is from the particular data being crunched and not a physical device operation suddenly getting faster. Either for some reason the share diff is dropping as the pool adjusts share rate to the miner or something along that line.

Really need to find the software end of mining Guru's here such as CK or Luke Jr.

Biodom

legendary

Activity: 3892

Merit: 4331

Quote from: shield132 on November 11, 2016, 04:21:49 PM

Quote from: wolfen on November 10, 2016, 05:44:30 PM

Sent another sick s9 board to denver , repair was $175+$25 to ship back.
One day turn around in denver.
Will be here in two days.
Excellent.

Is your repaired antminer s9 working at it's best? I personallyw would say that their return warranty isn't bad but sometimes they try not to replace your damaged product until you post in this forum.
And shipping cost was only 25$, so cheap? it's realy cheap for it's weight to send and receive this item at this price, in my country it would be 50 if not higher.

You don't send the miner for repair, just a board, which is not very heavy.
In fact, you have to pay even less in US to send the board, it is just sending it back (within US) is a fixed price ($25).

HashSpike

newbie

Activity: 21

Merit: 0

Quote from: gt_addict on November 11, 2016, 12:13:34 PM

Ive got three gecko usb sticks that consistently pop up with 18Ghs (set to 12ghs) but they don't reach it regularly, infact they often drop to 8-9Ghs (even though still set to 12Ghs). That's why you have an average hashrate (which is the only one you should be worried about). The next 5 min hashrate could only be 10Ths on your machine so it evens out.

I understand about the average hash rate. What I am curious about is what is happening to the miner that raises the TH/s that high, even though its only 5 minutes and drops back to average. I have a no fluctuations in electrical current, so what internally is making this happen.

I was hoping to get more feedback from the master electricians and hardware hackers. What I am hoping is that if the machine can run at 20.03 TH/s for 5 minutes, then should I not be able to find/duplicate the action causing the spike? Even furthermore, once found should you not be able to keep that action stable?

I also see in my Kernel Log that is showing two odd errors-

"Test Patten on chain[3]: FAILED!
Try to add 0.1V on chain[3]..."

"Below ASIC's core didn't receive all the nonce, they should receive 8 nonce each!

asic[12]=908
core[058]=4

asic[38]=910
core[023]=6"

How would I add 0.1V to board 3 and why do asic 12 and 38 not receive all the nonce?

Thanks for the help in advance!

wolfen

hero member

Activity: 770

Merit: 523

Quote from: shield132 on November 11, 2016, 04:21:49 PM

Quote from: wolfen on November 10, 2016, 05:44:30 PM

Sent another sick s9 board to denver , repair was $175+$25 to ship back.
One day turn around in denver.
Will be here in two days.
Excellent.

Is your repaired antminer s9 working at it's best? I personallyw would say that their return warranty isn't bad but sometimes they try not to replace your damaged product until you post in this forum.
And shipping cost was only 25$, so cheap? it's realy cheap for it's weight to send and receive this item at this price, in my country it would be 50 if not higher.

Yes, this is my second repair. First one working well. $25 from Denver to Virginia for one board.

shield132

hero member

Activity: 2352

Merit: 905

Metawin.com - Truly the best casino ever

Quote from: wolfen on November 10, 2016, 05:44:30 PM

Sent another sick s9 board to denver , repair was $175+$25 to ship back.
One day turn around in denver.
Will be here in two days.
Excellent.

Is your repaired antminer s9 working at it's best? I personallyw would say that their return warranty isn't bad but sometimes they try not to replace your damaged product until you post in this forum.
And shipping cost was only 25$, so cheap? it's realy cheap for it's weight to send and receive this item at this price, in my country it would be 50 if not higher.

gt_addict

hero member

Activity: 2534

Merit: 623

Ive got three gecko usb sticks that consistently pop up with 18Ghs (set to 12ghs) but they don't reach it regularly, infact they often drop to 8-9Ghs (even though still set to 12Ghs). That's why you have an average hashrate (which is the only one you should be worried about). The next 5 min hashrate could only be 10Ths on your machine so it evens out.

HashSpike

newbie

Activity: 21

Merit: 0

Hello all,

Can someone please explain to me how my S9 had a spike to 20.03 TH/s ( 5 Min Hash Rate ). I have had other spikes to around 18+/- TH/s also so this is not a one time occurrence. I am no hardware expert, so someone please clue me in...but if this thing can hash for 5 minutes at 20.03 TH/s, then is it not possible to figure out how I achieved that hash rate and then modify to retain that hash rate? I am running 650 freq. with the November 07 firmware.

http://imgur.com/a/Z2X8L

yesrno

newbie

Activity: 16

Merit: 0

Quote from: Biffa on November 10, 2016, 09:42:13 AM

Quote from: yesrno on November 10, 2016, 07:55:18 AM

The S9 has been sold out for a while, any news on new batches?

Its only been sold out for 2 days. They released a new batch 2 days ago https://bitcointalksearch.org/topic/m.16810866 and they sold out in a couple of hours.

Oh wow. Missed that one, thanks anyway.

wolfen

hero member

Activity: 770

Merit: 523

Quote from: MyRig on November 10, 2016, 11:50:38 PM

USA, Russia, Ukraine, Iceland and other limited service location are primary repairing the original item send in as the priority, however, if the damages are extensive but repairble, to reduce the mining time loss, the repair cost estimate be issued based on the actual damages but the replacement PCB may be send out to you to save a few days.

Main goal is to return the same PCB the customer send in. Also, if we do ship out the replacement, BitmainWarranty.com will offer 30-day limited service guarantee (workmanship guarantee) on the item or retain the original warranty duration of the unit whichever the longest.

Usually, bad S9 hash PCB can be repaired and ship back to the customer the same day or the next business day. (common duration be 1-3 days but the main backlog be after the repair is confirmed, the time delay while the customer service rep wait for the payment from the customers. Item will not be send out until the payment is received in full.) Some occasion, notification email seem to end up in the Spam folder of the customers. Other few days delay from customer opening the new ticket (sometimes with a different email address inquiring the same repair on another ticket number)

Minimal down time is great. I am very pleased with the service from bitmainwarranty.com

MyRig

donator

Activity: 792

Merit: 510

[email protected]

USA, Russia, Ukraine, Iceland and other limited service location are primary repairing the original item send in as the priority, however, if the damages are extensive but repairble, to reduce the mining time loss, the repair cost estimate be issued based on the actual damages but the replacement PCB may be send out to you to save a few days.

Main goal is to return the same PCB the customer send in. Also, if we do ship out the replacement, BitmainWarranty.com will offer 30-day limited service guarantee (workmanship guarantee) on the item or retain the original warranty duration of the unit whichever the longest.

Usually, bad S9 hash PCB can be repaired and ship back to the customer the same day or the next business day. (common duration be 1-3 days but the main backlog be after the repair is confirmed, the time delay while the customer service rep wait for the payment from the customers. Item will not be send out until the payment is received in full.) Some occasion, notification email seem to end up in the Spam folder of the customers. Other few days delay from customer opening the new ticket (sometimes with a different email address inquiring the same repair on another ticket number)

NotFuzzyWarm

legendary

Activity: 3822

Merit: 2703

Evil beware: We have waffles!

Quote from: wolfen on November 10, 2016, 08:28:38 PM

I think they check to see if it has any burns, if not you get a replacement from their stock.
Then yours gets sent to China for repair. Just my thoughts.
One day turn around? Hmmmm

I have picture of all boards I sent out and yes most maybe all of the times got different boards in return from both Bitmain with boards sent to them under RMA along with boards sent to BitmainWarranty for paid service.

Works for me. If the returned boards are reworks that were sent to China, also fine also long at it works and in fact, unless explicitly stated otherwise it is normal at least for industrial equipment industry-wide practice for Factory-Certified rebuilt replacements to be provided - even under Warranty Terms - when billable repair costs exceed value provided the mfgr/repair depot keeps the dead item (to be looked at elsewhere). If the final repair location can more economically fix the boards and put them back into the for Warranty repairs pool so be it.

wolfen

hero member

Activity: 770

Merit: 523

Quote from: NotFuzzyWarm on November 10, 2016, 06:51:02 PM

Quote from: wolfen on November 10, 2016, 05:44:30 PM

Sent another sick s9 board to denver , repair was $175+$25 to ship back.
One day turn around in denver.
Will be here in two days.
Excellent.

Ja. gotta love the ability to insure at full value + no concerns with Customs.

Now -- if we can only get Bitmain Warranty to tell us WHAT is failing on these boards! I asked them about it a couple time in email and is usually the same answer: "Well it means having the tech(s) record what they find and what is done to resolve it. That means someone (in Bitmain Warranty) has to enter the info in to logs and update the tickets with the information...."

In short - they won't do it because "it takes time and people.."
Riiiiight. Bet each tech has a daily work log of what they do eg. what Ticket#'s they work on, how many parts, when started, when done.. Bet *somewhere* in that work Ticket information is the diagnosis and resolution of problem. In other words: Fault(s) found and how repaired/why replaced. So why not lets us know?

Given what the repair cost runs, mine in the past have been around $140+ return shipping, Whatever is failing cannot be taking long to find/fix nor cost much if it a physical part that needs to be attended to. My company bills board/module (laser systems) repairs @ $115/hr + parts and based on that -- you get what I mean.

I think they check to see if it has any burns, if not you get a replacement from their stock.
Then yours gets sent to China for repair. Just my thoughts.
One day turn around? Hmmmm

NotFuzzyWarm

legendary

Activity: 3822

Merit: 2703

Evil beware: We have waffles!

Quote from: wolfen on November 10, 2016, 05:44:30 PM

Sent another sick s9 board to denver , repair was $175+$25 to ship back.
One day turn around in denver.
Will be here in two days.
Excellent.

Ja. gotta love the ability to insure at full value + no concerns with Customs.

Now -- if we can only get Bitmain Warranty to tell us WHAT is failing on these boards! I asked them about it a couple time in email and is usually the same answer: "Well it means having the tech(s) record what they find and what is done to resolve it. That means someone (in Bitmain Warranty) has to enter the info in to logs and update the tickets with the information...."

In short - they won't do it because "it takes time and people.."
Riiiiight. Bet each tech has a daily work log of what they do eg. what Ticket#'s they work on, how many parts, when started, when done.. Bet *somewhere* in that work Ticket information is the diagnosis and resolution of problem. In other words: Fault(s) found and how repaired/why replaced. So why not lets us know?

Given what the repair cost runs, mine in the past have been around $140+ return shipping, Whatever is failing cannot be taking long to find/fix nor cost much if it a physical part that needs to be attended to. My company bills board/module (laser systems) repairs @ $115/hr + parts and based on that -- you get what I mean.

wolfen

hero member

Activity: 770

Merit: 523

Sent another sick s9 board to denver , repair was $175+$25 to ship back.
One day turn around in denver.
Will be here in two days.
Excellent.

Biffa

legendary

Activity: 3234

Merit: 1220

Quote from: yesrno on November 10, 2016, 07:55:18 AM

The S9 has been sold out for a while, any news on new batches?

Its only been sold out for 2 days. They released a new batch 2 days ago https://bitcointalksearch.org/topic/m.16810866 and they sold out in a couple of hours.

yesrno

newbie

Activity: 16

Merit: 0

The S9 has been sold out for a while, any news on new batches?

Biffa

legendary

Activity: 3234

Merit: 1220

Quote from: Biffa on November 09, 2016, 04:47:44 PM

Quote from: wolfen on November 09, 2016, 04:39:35 PM

Quote from: Biffa on November 09, 2016, 04:37:53 PM

Quote from: wolfen on November 09, 2016, 01:04:28 PM

Quote from: Biffa on November 09, 2016, 09:58:05 AM

Both my batch 18's look like they might have a dead board in them after just 8 days use. Sad

Need to check when I get home

What pool?

Kano

Kano is nice and smooth, should be good.
You may be able to send the dead board to Denver and they may fix it for $175+$25 shipping.
One of mine just arrived there now. You need a ticket at bitmainwarrany.com

Lol yes I know I've been mining there since a few months after it started.

They aren't dead just acting wierd. bitmainwarranty is an option, depends what costs more, the last one I sent back to China worked out ok.

Reboot seemed to sort them out. Back to 11TH

kodokbuduk

full member

Activity: 294

Merit: 100

Quote from: mamiakimo on November 09, 2016, 06:36:39 PM

Quote from: wolfen on November 09, 2016, 06:25:52 PM

Quote from: mamiakimo on November 09, 2016, 05:51:01 PM

Quote from: wolfen on November 09, 2016, 04:38:59 PM

Quote from: mamiakimo on November 09, 2016, 04:11:45 PM

should i return to old firmware or what ?
they told me don't worry it's fine

Different freq boards means it is autotune.
New stuff.
New firmware, new hardware controller.
I don't think the autotune s9 will work with old firmware, (pre autotune) but I have not tried it.

this version Antminer-S9-all-201611020159-autofreq-user.tar doesn't have this issue

1102 seems to run too hot. There is an 1108 version.

i returned to Antminer-S9-all-201610260255-autofreq-user.tar
best one

Previous early batch s9 default freq 550/600, the hashboards have missing heatsinks (bitmain took out intentionally). If we update the new auto tune firmware s9, i think the boards will be too hot..so with the new auto tune s9 firmware, is it OK to update?

not.you

legendary

Activity: 1726

Merit: 1018

I would like to see one that can keep the temps under 100 without ramping up and down over and over again. How hard could it be to set it up to make smaller speed adjustments when slowing and only do large adjustments when speeding up? If it's hot and you need it cool you need it to speed up quick but if it is cool you don't need it to drop from 100% all the way to 10% the way some of them do. That's dumb.

EDIT: Or just give us manual control back...

Topic: Bitmain's Released Antminer S9, World's First 16nm Miner Ready to Order - page 210. (Read 531173 times)