ANTMINER S3+ Discussion and Support Thread

[Biodom]

I'm guessing this is not normal?

[IM]https://i.imgur.com/fymGVXjl.jpg[/img]
[IM]https://i.imgur.com/Mr400HGl.jpg[/img]

Seems like you overclocked the miner. right? or you applied a non standard DC input?
A change of that capacitor will fix your issue. Those resistors are damaged of not?

If overclocking did that, I'm not sure I want to try and OC mine now. I am one of a lucky few that got four miners that work stable at 440Gh/s on sock freq 218.75.

With this hardware, overclocking id risky.

But the question at hand is did you OC and get that damage from it and if so, what freq was it on and what did you jump to from it ?

There's been plenty of people OC here and none I have seen reported these kinds of issues.

To booradlly
No I didn't do an overclock. But your symptoms as per the images force me to think like that.
I used to overclock my gpus long back and got into this kind of isues.
I usually get this kind of a thing when voltage of the chip(s) increased.

Looks a lot like a capacitor blew after a power surge from something like nearby lightning, its almost identical to what the PCB on my LG monitor looked like after lightning struck the parking lot across the highway from my office.

Are you guys using high joules surge protectors for your miners?
If not, you are risking the same as shown (boards blown). There are plenty of thunderstorms in the summer in my area, so I stocked up at the Microcenter on a few of surge protectors.

[1715077730]

1715077730

[1715077730]

1715077730

[1715077730]

1715077730

[edonkey]

If anyone is interested, I'm posting the stats on my two B3 units where one is slightly over clocked and the other is stock. I picked which unit to over clock based on the fact that it had the lowest hardware errors after running for several days.

Note that both units are running at a colo facility with 208 volt power and temperature control. The stock unit runs about 38 degrees and the over clocked one between 38 and 39.

I'm using 2 CX750M PSUs, one per S3. I know it's overkill, but I bought them before we all found out that the S3s are not over clock friendly. I used all 4 power connectors because I wanted to experiment with over clocking from a distance without having to go to colo and move plugs around. FYI, none of the plugs or cables have ever felt warm to the touch.

Here are the stats after running for almost 6 days with the above configuration:

Frequency	GH/s	HW	DiffA	DiffR	Error Rate %
218.75	441.44	1,732	50,942,208	63,488	0.00339558387456823
225	453.36	8,056	52,252,928	71,680	0.0153938274573601

The delta is that the over clocked unit is providing about 2.7% more accepts with about 3.5 times more hardware errors. Prior to over clocking the hardware errors were quite a bit less than the current stock unit. But the nearly 3% increase in accepts seems worth it to me.

Note that I also tried a frequency of 237.5 on the over clocked unit, but I got 824 HW errors after 20 minutes. This seemed like an insane error rate, so I backed it off to 225.

[PlanetCrypto]

I'm guessing this is not normal?

There is Magic Smoke that is trapped inside all electronics devices.
The Magic Smoke has escaped those components.
You must replace those components with new components that have the Magic Smoke still imprisoned in them.
Then your S3 will function again.

Seriously, from your pics, looks like a couple of SMD (Surface Mount Device) resistors/capacitors experienced an excessive over current condition.
Can't read the litho on the card next to the components in your pics. RXXX = a resistor, CXXX = a capacitor where XXX is the component number.
Take a que-tip dipped in iso-propol alcohol and clean off the charring next to those components to be able to read the litho on the circuit card. Make sure you're well grounded when doing this, else static electricity might toast something else.

To read the markings on the components that identify their values you'll likely have to look at another undamaged machine.
This type of component is dirt cheap, shipping on new parts will likely exceed the cost of all those components combined. We're talking a total order $$$ amount under $25.
They can be ordered from a plethora of online sources. I order a lot from Digikey cuz' they're nearby and shipping times are brief.
Once obtained and still being well grounded, de-solder the burnt parts and solder in the new.

Identifying why the Magic Smoke escaped those components is of particular interest to me and likely to others.

[Biodom]

If anyone is interested, I'm posting the stats on my two B3 units where one is slightly over clocked and the other is stock. I picked which unit to over clock based on the fact that it had the lowest hardware errors after running for several days.

Note that both units are running at a colo facility with 208 volt power and temperature control. The stock unit runs about 38 degrees and the over clocked one between 38 and 39.

I'm using 2 CX750M PSUs, one per S3. I know it's overkill, but I bought them before we all found out that the S3s are not over clock friendly. I used all 4 power connectors because I wanted to experiment with over clocking from a distance without having to go to colo and move plugs around. FYI, none of the plugs or cables have ever felt warm to the touch.

Here are the stats after running for almost 6 days with the above configuration:

Frequency	GH/s	HW	DiffA	DiffR	Error Rate %
218.75	441.44	1,732	50,942,208	63,488	0.00339558387456823
225	453.36	8,056	52,252,928	71,680	0.0153938274573601

The delta is that the over clocked unit is providing about 2.7% more accepts with about 3.5 times more hardware errors. Prior to over clocking the hardware errors were quite a bit less than the current stock unit. But the nearly 3% increase in accepts seems worth it to me.

Note that I also tried a frequency of 237.5 on the over clocked unit, but I got 824 HW errors after 20 minutes. This seemed like an insane error rate, so I backed it off to 225.

That's good stats, but why do you care so much about the HW rate? All what is important is production per time period. 824 errors if not possible to calculate without diff A and diff R numbers. maybe it is still low. I would say that if HW error (calculated properly) is less than 1%, then it is still worth it to go to 237.5. I have one S3 that has one asic with an X at 237.5, but total speed is still 464-465 GH, which is better than 441, so I stay with 237.5 for it.

[PlanetCrypto]

If anyone is interested, I'm posting the stats on my two B3 units where one is slightly over clocked and the other is stock. I picked which unit to over clock based on the fact that it had the lowest hardware errors after running for several days.

Note that both units are running at a colo facility with 208 volt power and temperature control. The stock unit runs about 38 degrees and the over clocked one between 38 and 39.

I'm using 2 CX750M PSUs, one per S3. I know it's overkill, but I bought them before we all found out that the S3s are not over clock friendly. I used all 4 power connectors because I wanted to experiment with over clocking from a distance without having to go to colo and move plugs around. FYI, none of the plugs or cables have ever felt warm to the touch.

Here are the stats after running for almost 6 days with the above configuration:

Frequency	GH/s	HW	DiffA	DiffR	Error Rate %
218.75	441.44	1,732	50,942,208	63,488	0.00339558387456823
225	453.36	8,056	52,252,928	71,680	0.0153938274573601

The delta is that the over clocked unit is providing about 2.7% more accepts with about 3.5 times more hardware errors. Prior to over clocking the hardware errors were quite a bit less than the current stock unit. But the nearly 3% increase in accepts seems worth it to me.

Note that I also tried a frequency of 237.5 on the over clocked unit, but I got 824 HW errors after 20 minutes. This seemed like an insane error rate, so I backed it off to 225.

Yes am interested, thanks.
Especially for the stat analysis and logic behind it.

On a different note, eMule has be a close friend of mine for years.
Are you related?

[edonkey]

Yes am interested, thanks.
Especially for the stat analysis and logic behind it.

On a different note, eMule has be a close friend of mine for years.
Are you related?

You're welcome.

Sorry, I don't know eMule   I used eDonkey years ago for online gaming, and it just stuck.

[edonkey]

That's good stats, but why do you care so much about the HW rate? All what is important is production per time period. 824 errors if not possible to calculate without diff A and diff R numbers. maybe it is still low. I would say that if HW error (calculated properly) is less than 1%, then it is still worth it to go to 237.5. I have one S3 that has one asic with an X at 237.5, but total speed is still 464-465 GH, which is better than 441, so I stay with 237.5 for it.

Good point. I guess I just panicked because the hardware error increase was so dramatic so quickly.

I might try bumping it back up to 237.5 and leaving it for a day, then recompute the stats. I might also try bumping up the stock unit. It's worth seeing what it can do.

[jpchrist]

I'm dirty blonde if you want to call it that....

Clearly NOT a dorm room

[JennaK]

I'm home for the summer??? I never said I took the pic at my university.. Lol
I love how you all try to play detective.

[jpchrist]

I'm home for the summer??? I never said I took the pic at my university.. Lol
I love how you all try to play detective.

Dude, give it up man. No one believes you and now youre just embarrassing yourself.

[JennaK]

I'm home for the summer??? I never said I took the pic at my university.. Lol
I love how you all try to play detective.

Dude, give it up man. No one believes you and now youre just embarrassing yourself.

I don't care if you think I'm female or male.. Lol

[philipma1957]

I'm home for the summer??? I never said I took the pic at my university.. Lol
I love how you all try to play detective.

Dude, give it up man. No one believes you and now youre just embarrassing yourself.

I don't care if you think I'm female or male.. Lol

 I don't care either , but it is a very good way to do referrals.  So if you are the young lady in the photo or if you took the photo of your girl friend I don't mind since I won't be using your referral link either way.

 Still a nice  way to market   pbmining

[jpchrist]

I'm home for the summer??? I never said I took the pic at my university.. Lol
I love how you all try to play detective.

Dude, give it up man. No one believes you and now youre just embarrassing yourself.

I don't care if you think I'm female or male.. Lol

Exactly what a dude would say....

[JennaK]

I've yet to get one PBMining referral lol

[PlanetCrypto]

I'm home for the summer??? I never said I took the pic at my university.. Lol
I love how you all try to play detective.

Dude, give it up man. No one believes you and now youre just embarrassing yourself.

I don't care if you think I'm female or male.. Lol

 I don't care either , but it is a very good way to do referrals.  So if you are the young lady in the photo or if you took the photo of your girl friend I don't mind since I won't be using your referral link either way.

 Still a nice  way to market   pbmining

Or posting that referral link might just solicit a DDOS or SSH flood attack with JennaK as the referrer.

Just sayin'

[PlanetCrypto]

I've yet to get one PBMining referral lol

And I guess that speaks volumes for the worth of having that in your signature.

[JennaK]

I'm home for the summer??? I never said I took the pic at my university.. Lol
I love how you all try to play detective.

Dude, give it up man. No one believes you and now youre just embarrassing yourself.

I don't care if you think I'm female or male.. Lol

 I don't care either , but it is a very good way to do referrals.  So if you are the young lady in the photo or if you took the photo of your girl friend I don't mind since I won't be using your referral link either way.

 Still a nice  way to market   pbmining

Or posting that referral link might just solicit a DDOS or SSH flood attack with JennaK as the referrer.

Just sayin'

That would be awesome lol

[maxood]

Reply from bitmaintech can be instant or sometimes it will take some time.
No no invoice is provided with the shipment and you can use the one available in the website.
In the shipping declaration unlike S1 and S2 they declared the full correct value in US dollar.
In declaration the product is described as "Bitcoin Miner". No customs issues.
They may call if some officers don't know what Bitcoin Miner is. Bitcoin is famous but mining thing is not so familiar to all.

allcoinminer, thank you for the reply!
It means that I have to pay 21% more (VAT) than US miners.
Guys, is there any EU state where you can import, say 2*S3, without VAT payment?
Is there any distributor/reseller of Bitmain in EU? I can't find any reasonable offer on EU Ebay segment.

[philipma1957]

Reply from bitmaintech can be instant or sometimes it will take some time.
No no invoice is provided with the shipment and you can use the one available in the website.
In the shipping declaration unlike S1 and S2 they declared the full correct value in US dollar.
In declaration the product is described as "Bitcoin Miner". No customs issues.
They may call if some officers don't know what Bitcoin Miner is. Bitcoin is famous but mining thing is not so familiar to all.

allcoinminer, thank you for the reply!
It means that I have to pay 21% more (VAT) than US miners.
Guys, is there any EU state where you can import, say 2*S3, without VAT payment?
Is there any distributor/reseller of Bitmain in EU? I can't find any reasonable offer on EU Ebay segment.

no a buyer like you is better off ordering from a hosting service like gawminers   .  you own the gear they host it.   I am pretty sure you avoid vat.

they allow a cc to purchase it.   I could link you to the site but it is kind of a thread hijack.  I know of 3 that will host the gear you buy one allows paypal one allows cc's. 

since the gear does not enter the EU no vat.  not too sure as I am in the USA .

[justin.ramos]

I have 2 S3s that are now over-clocked. I'm using a Corsair RM850 to power both, with 4 dual PCI-E cables. I didn't have any luck getting them to run stable at 250Mhz, so I backed off to 237.5Mhz. After about 30 minutes....

http://www.ramosresearch.com/images/bitcointalk/overclock.results.2014-08-10.png

Stats collected using my CgminerManager (https://github.com/jramos/cgminer_manager).