Topic: Is the S5 string design safe and was it really tested properly? (Read 5868 times)

yeah I recall a few people tried doing the 24volt 2x series wiring 2 or 3 years ago with mixed results.

While I tinker quite a bit with electronics I am self taught and I am colorblind. (which is why I am self taught)

  Really fucking annoying to have colorblind issues when I like to do electrical projects.  I have reach the age of 58 without burning down my house or some friends homes (knock on a lot of wood)

 here is why  :
 If the wires are many and the colors are tough I call the wife  with 'honey'   purple or blue/ orange or red/ green or brown ?    So far so good.

When I said "isolated from neutral" what I meant to say was "isolated from ground" (chassis/earth ground). I've been doing more panel wiring than device-side wiring lately, where neutral is ALWAYS neutral and tied to ground. Thanks for the correction and clarification.

A lot of times the thing required to isolate a server PSU is insulated washers between the board and case standoffs. Depending on how ground planes are run, insulated (usually plastic) screws might be required as well. But yes, everything MrTeal just said is correct - if you do it wrong, the best case is the supply trips out; the worst case is it burns your house down and/or kills you so be careful.

*The following is not to be considered professional advice.

An modern ATX PSU has the mains lines (hot and neutral, or hot/hot) isolated from the output and from ground. Besides being a safety thing, it's a requirement of being able to run 240V into them in places like NA where 240V is two legs that are 120V w.r.t. ground. The chassis of the supply is of course connected to the ground pin on the incoming mains cable. That is a safety thing and should not be tampered with.

In and of itself, that would not cause an issue with stacking PSUs. However, ATX PSUs also have the 0V output connected to the chassis and therefore the earth ground. You would not be able to stack PSUs without shorting out the bottom PSU because of this. There are quite a few pages out there where people have converted two server or ATX PSUs into a large 24V supply for charging LiPo batteries or the like, but it involves opening up the case and removing or isolating the supply, so it is a risky "do it wrong and kill someone" kind of project.

I recall something on anandtech  someone ran 2  separate  12 volt rails from 1 psu and got 24 volts.  but that was a while back and I did not really check it out.

With PSUs the thing to worry about is isolation. Make sure the ground on the output is isolated from mains neutral or they'll trip out and/or smoke when you try to turn them on in series. DO NOT connect load-balancing pins on server supplies in series. It shouldn't be needed anyway.

atx psu 's

  2 or 3 can do series  to boost voltage  live and learn.  makes sense I just never tried it.  did it lots of times with batteries.   and when the batteries charge balancing the charge them is hard.

I'd really like to see individually fused chips and a backup parallel current path (like the FETs on Prismas) so if one chip starts to roast it takes itself out of the circuit before catching fire and the parallel path kicks on to keep the string's current flowing. As long as the parallel doesn't smoke out (as it'd be dropping the same power as one of the hashing chips) the thing should keep running at (N-1) hashrate.

The string design does open up a lot of efficiency headroom, but fluctuating series loads will always cause room for concern.

I'm not a big fan of the "chain" design on the S5's.  I'm down 3 S5's in under 2 weeks because of it.  When one chip goes, the whole hashboard is toast (not physically), 30X's.  At least with the S1/S3 you could blow a chip here and there and the rest of the board would still hash fine, albeit slightly slower.  Now, you lose the whole damn thing and are down half a miner in one shot.

One of them burnt up immediately after turning on, the fan didn't spin and we didn't catch it in time, the smell alerted us to something wrong.  I'm not sure if it would have physically burnt, but it was certainly too hot to touch right after powering it off.

The open air design itself shouldn't be a fire hazard, however, unless you've got metal on metal and an open wire hits it.

Most of them do reliably overclock to 387.5, resulting in roughly 1.25TH.  Some are decently stable at 412.5 @ 1.35TH.  Some are unstable at the stock speed.

I think he means stack two PSUs in series for 24V, and then put three S5 in series below it for 1/3 voltage each. It'd be about the same as people have suggested using 3x PSUs and 4x S5 for 9V each. That'd only work if the current draw per S5 is the same; likely you'll need to tweak clocking on a per-machine basis, possibly recursively, to get a series chain of miners to run stable.

Also, we should have a test PCB for a big ol' regulator on Thursday, and probably some efficiency and reliability data next week (I'm working on a good conversion efficiency measurement toolchain tomorrow). Just throwing that out there. I should really pick up an S5 to play with.


Also, raskul, don't lose hope. One of the prismas in my hosting threw a 31G share over the weekend, after throwing a 3.9G share the day before, after not working very well all week. Success is pretty random.

no.   biggest reason is  testing shows the s-5's don't want to go lower then 9.5 volts I believe MR Teal  was able to go no lower.


second reason is   I have no idea how 2x 12 volt psu's  would down volt to 8 volts.     what would happen is  each machine would get 12 volts   and a share of the available watts the 2 psu's can provide.  assuming you plugged the gear in correctly.


in order to down volt the power from a psu a big version of this is needed.

http://www.amazon.com/gp/product/B00OT58TL4/ref=oh_aui_detailpage_o02_s00?ie=UTF8&psc=1


plug in psu pcie 's and crank the volts down to 10 volts or so.

if the time comes, and S5's need to be undervolted, can one -instead of changing the PSU to 9V- run 3 Antminers on 2 PSU's so 24V/3=8V per S5?

solo.antpool.com should be antpool solo with 35-42ph. All my sp20 mining there. (switch from ghash)
p2p antpool not going well, many of us complain not getting share in few days.

meh. i reckon it'd take me about 3 months to get up to 40bill

On solo you get 25 BTC when you find a block. p2p antpool is a regular pool.

just mine 
i asked Yoshi if I could mine on antpool with my SP20's and he said yes.

what's the difference between solo. and p2p. on antpool?

there is not a lot of evidence of  an  s-5 burning up .   


clock it to freq 325   with a decent psu .  you should be safe.

I think the one that burned up for raskul

is the only documented one burning out.


In contrast the asic miner prisma had at least 5 or 6 photoed burnouts posted on this site.


In general Antminer gear has been pretty good for most of us.

So i just got 5 S5's

tbh i never had a bitcoin ASIC miner till now, i have been only mining with GPU's ( altcoins )

SO about the S5, after reading this thread i am kinda afraid to even start it

If i do not oc the miners, are they safe ?
And all here are talking from experience i guess ( they also have S5's ? )

sorry, my bad.. i meant it was the PSU which went up in smoke, when the S5 blew a hot-thingy-ma-jig. the SP20 is fine, no problems whatsoever. GUI's all returned no problem too, so we all good.
here is it now:



i think i recall responding to that one.



totally agree with that, i'm afraid.

I read here something else
https://bitcointalksearch.org/topic/powering-a-sp20-920699



No. Only if you have a good fire alarm system and own guardian angel 24/7

S5 is in my opinion the fire hazard in this open air case.

So is it safe to oc S5 ?

Did you burn the evga?

I did reach 1.7 too, with fan at 100%, max setings on voltage and wattage, and air at 0c.....

I had a 1600 EVGA which was only running one SP20... it went up in a puff of smoke... Sp20 is more about getting the settings right... I have three SP20 and each of them are able to go to 1.7TH/s

You can't really punch something above 288, because the webgui puts back 288 in that field that you tried to put higher setting!
I did somehow, and don't remember how, put 1350w in there. But we still got the 0.79v limit...

I mean someone with programming knowledge, maybe even SPtech...
They don't give it out because they don't want higher rates of returning machines; and they still have them on stock plus another 3month after that....

Or maybe the settings are hard written inside and can't be taken out!

What do you mean by "take out the 288watt cap" ? Blank out the field ? Set it to 0 ? Set it to a high number like 350 ?

+1

I have 11x EVGA G2 1300W

This is not about cables, but the software limit.

My second and third SP20 worked 1720GH/s and 1735GH/s  if they are accidentally have after update   SP10 software.

take out the 288watt cap or the voltage limit of 0.790 and we can hit 1.8th/s

get better cables  my SP20's are all perfectly capable of 1.7TH/s
i run them ~1.5.

I agree. All buy, but no one sees that SP20 is still underperformed 1.7 TH/s ± 10%

Based on my 11x SP20 I only see 11 x  1.7 TH/s - 5%


Only my very first SP20 was 1720TH/s of normal room temperature. Even that only have one day to the first software updates.
I look forward to  software that would allow me back 1.7 TH/s +

  let it sit in the home and warm a bit first.

I'm reserving judgement until i test my own unit.
I intend to draw a circle on the floor and surround it with candles to keep away evil spirits and unwanted outside influences.

Teehee 😜

First S5 in the wild: https://bitcointalksearch.org/topic/m.9977426

raskul you missed your 15 minutes of glory!

They were out of stock of C1's for a bit that makes me think they did run out to stop selling.  But again I could be wrong I have no proof either way.

Thanks!
But shitty PSU that you said is come from CHINA surely, not from TAIWAN, we were all called as chinese but  there is a BIG difference between us.
I use super flower's PSU since I started mining, high C/P and realible!
90% load and no anyone fault so far
https://i.imgur.com/QhI8Nih.jpg

Batches are artificiality created as a marketing tactic.
For all we know the first batch might be few tens of S5 to test the product on live customers.

I made really good experiences with Antminers so far (except the big ones S2 and S4)

next time use seasonic cable.
http://www.ebay.com/itm/Single-Seasonic-power-supply-modular-cable-/321608923301?pt=US_Power_Cables_Connectors&hash=item4ae162b8a5
after the experience with mining ati r9 290 all this noname chinese shit PSU is simple shit. only premium PSU have adequate cable. I burn a lot of cable until I realized Seasonic PSU is, other PSU is Simpley lottery
When you baying seasonic cable check is seller official seasonic dealer,because lot  bad chinese copy is available.I do not understand how the noname Chinese think that if Avg 17 cable can haul 25A.

Yes, that is where the market is at the moment ... interesting times ... Bitmain celebrating 12 months ... Sponds yet to do that.

I think them doing a batch 2 proves it.   But I don't have the info you do so I could be wrong.  I think they would have stopped after batch one, or stopped it completely like L1's.  

For what it is worth I think we will see SP and Bitmain continue to battle it out.

here is my take on this.

we don't know if gear works well until the first adapters use it test it and post on it.

Asic Miner

1)Usb sticks pretty solid worked pretty well
2)Blades pretty solid worked pretty well
3)49 port usb hub issues with the stock fuse many melted.   most of us used a lessor fuse problem solved
4) cubes many batches had issues and burned or melted fuses
5)little tubes  had issues some melted some popped caps
6) prismas worse then the tubes

I owned all of the above.  I only felt safe with the usb sticks and the blades

Bitmaintech

S-1 pretty safe a few but not many issues
S-3 very safe a few but not many issues
S-2  shipped poorly lots of psu issues
S-4 better then the  s-2 by far
C-1 some water pump issues

Spondoolies

sp10   worked well not many burnups
Sp20  I have 8  they work great at 1500gh or under  If I owned the company I would have called them 1500gh miners
sp30  Some psu issues underpreformed a bit.  spondoolies paid comp money
sp31 a few psu issues
sp35 a few psu issues.

For the sake of argument  I am using the 3 companies above.
I would fear Asic Miners new build based on the last 4 things they sold.  but I will buy one and test it.
I am getting an S-5  as a test demo... I will test the shit out of it would I fear to buy it no.  Would I buy 100 from the first batch no.
Sp20E  do I fear it no would I buy more yes. 

 BUT  I drive my Kia Forte at 70mph max or about 115KPH max.
I clock my Sp20E at 1400gh or 1430gh max   I do not even attempt to think 1600 or 1700gh.

So when I get the S-5 I will test the shit out of it and report back.
Until then I won't fear the design.  PS I will power it with my 760 watt seasonic platinum for the first set of tests.  why is that the seasonic has short and surge protection.

Who said they aren't losing $100s on the C1 at that price? You don't know.

My main concern with the S5 is why does it cost so much?

The C1 is $325 and has FOUR hashing boards. The C1 boards also have a DC/DC converter for every two chips IIRC. The buck controller ICs probably cost more per unit than their custom silicon.

There is a major BOM cost reduction going to the chained design. The controller looks the same (cost reduced beaglebone black), the new chips of course have an NRE that needs to be recouped so they're probably just milking the first batch of people who are gonna buy it just because it's new.

Not even considering the possibility of instability or catastrophic failure due to the chain design, I'm gonna sit on my hands for a while. I'm willing to bet the first few price drops will be pretty good.

I have no evidence as units are yet to ship, simply from what I see on the announcement thread, makes me irk towards something not quite right about the design. 


I never buy batch 1 miners for the specific reason that you just can't be sure and I only have experience in these fields from my home mining endeavours. It's a serious gamble to lay out vast amounts of money on something which has not gone through proper independent testing. I'd preferred to see a few independent reviews before any miner goes on sale, not just BMT.


I cannot make decisions for you, you should weigh up the facts for yourself and decide accordingly. I'd just like to see any initial bumps evened out before anyone puts their money into it.


all the best to you and yours this festive season.

Seems serious old boy....

Is there any evidence of localized heating or hot spots on the boards?
How about in the Bui, any sign of uneven cooling?
Are you running at 12v or 9?
What guage cables are you running from the PSU and is there any sign of excess heating?

One wouldn't want to invest in something which has design problems. Do you have much experience in asic or electronic design yourself?

I'll put my bulk order of 52 on hold till I hear back from you old chap.
Have a tidy profit from those nice chaps at GAW I'm looking to re invest and these look like just the ticket.

Toodle pip for now....

absolutely

But please use good cables this time...

...and so endeth the lesson in customer service...

 

THIS IS MY BAD
spondoolies is so kind that give me a new one just for free!!

Thank you for the double post.
https://bitcointalksearch.org/topic/m.9924332

you can see the problem. the user put the wrong cables.

Today I bought 12 pieces Antminer S5

I trust them today more than SP20 , especially after seeing this picture. Because they have  4  PCI-E slot, For 600W.
SP20 has 4  PCI-E slot for 1200W (Max)


I have 3 x SP20 and
6pc SP20 purchased by December 17, to get my hands probably 29.12

I also have a 5pc KNC Neptune, with no burnt PCI-E slot.


The rest are all Bitmaintech production. (37x S3, 6x C1, 2x S4 )

It would be seen that picture before, then I would not have ordered more SP20.

this.

to be fair it's not just me with apprehension... cross-quoting these here as I've noted some of my own posts being deleted from the S5 thread:

i never buy batch1 of any miner anyway so i'm reserving my full judgement until we can read some proper unbiased reviews. from reading your post in the S3 thread it does indeed seem like Bitmain had actually sent you refurbished units from prior batches, rather than new stock... I can only presume that the S5's which don't go up in smoke will be subject to the same.
And initially, I thought the open top idea was quite a novel way of dissipating the heat from the chips, but I was a bit drunk at the time. quack-quack.

I have to say it looks a lot like the S3 as Raskul says it looks like an S3++ and frankly after the performance of my 18 x s3's - I have to say.

If it walks like a duck, looks like a duck and smells like a duck..........................its probably a duck.

I wouldnt invest my money in these.

It's a nice idea in that you save a hell of a lot of copper, I just don't know if anybody is actually able to pull it off. There's a lot of hints here that the Antminer chips use the same masks as the Bitfury ones, the specs of the Bitfury 1 and Antminer's first chips are almost identical other than the choice of package. It's suggested elsewhere that Bitfury would be making a new chip, what if this is it re-badged? I'd have a lot more faith in it working if the design is a third generation Bitfury rather than a third generation Antminer.

I have some serious reservations about how Bitmain have gone about creating their newest miner, the S5.

After seeing the magazine adverts and watching the speculation build, it would seem, to me at least, that they are designing a miner which has a good chance of failure due to the risky design.

Bitmain have done a private batch of S3++ using the chain design on an entirely new chip, in preparation to sell the S5 en-mass.
S3++ hasn't been released to the public, so we can only presume that there are a few large farms, with vested interest, mining large scale, across the world on a private batch of S3++ miners.

The string design is a very bad method to use, after so many experienced the troubles with the Prisma from AM... and... AM had to recall an entire batch.

Sure the S5 is BMT's most efficient miner to date but it's quite late to the party, the SP20 can reach this efficiency and below with proven dc2dc stable design and... without what seems to be, by Bitmain, a risky cost cutting exercise.

I can't help but think something is not quite right here... presuming that there were problems with the chips, or the boards on the S5's for them to appear this late after advertising... having an open-top design of the casing, surely won't allow enough airflow, either?

I can't get my head around what Bitmain are trying to accomplish here, so I'm putting it out there... it's a product which has high chance of failure and I can foresee many RMA'd units.