Topic: An Antminer S5 review by klondike_bar; the loud, affordable miner (Read 12673 times)

I have received mine yesterday and .......

It's more than LOUD. I could not believe.
I'm in a small room with two miners at work.
One is a asicminer Tube 800 which I was thinking very noisy after an S3+,
but S5 killed all the records.
My android cell phone shows 75 db when I go next to them. I'm standing up and measuring 1 meter away from.
I do not know how accurate my measurement is.

I'll try your solution when I find an appropriate spare fan frame...

Dog1965, I have a mechanical engineering diplom, bs. (Though I did not work as a Mechanical Engineer but started IT business with ZX Spectrum 48K computers, and I'm 48 years old now)
I congratulate your idea and appliance.
My S5 is on the way coming with UPS. I hope to get it in 10 more days.
I will first apply your solution and then run the machine.
I will tell the results here later.

I agree to your air direction modification thinking...I'm now looking for a spare appropriate fan. I think the ones who applied your solution hopefully successfull...
Good job.

1 silverstone 141 as a pull

and the stock fan at 9 volts as a push works well

Thank you for sharing the info! 

These are the specs on the S5 stock fan.


Dimension: 120*120*38mm
Voltage: 12V
Start voltage: ≧8V (Power ON/OFF)
Power consumption: 1600RPM 1.68W / 4000RPM 25.8W
Rated current: DutyCycle 100% @ 2.15A
Air Flow: 4000RPM @ 178.3CFM (Max be 239CFM)
Noise: 4000RPM @ 56.7 dB-A (Measured in a Non-EchoChamber)
Life Expectancy: 40,000 hrs @ 25℃  (May degrade faster at higher temperature)
Connector: 2510-4P

To replace this kind of fan to make the rig more quiet, you need to find a two fans that are either 25mm or 38mm thick, at least 3000 RPM's, have a CFM of 150 or more and a dB noise rating of less then 40 dB. And folks, let me tell you, it is impossible to find. The only thing that comes close is the expensive Noctua NF12-F12 IPPC 3000. 43dB, 109CFM and have to use both. I have been searching for two 120mm x 38mm fans that can push at least 150 CFM, spin up to 3000, have 4 pin PWM control and a dB rating of less than 40 dB. Impossible to find, since most of them have a high noisei dB rating.

So I stuck with the two high priced Noctua's priced at $24 each and each having a 6 year warranty. By all means they aren't silent but much better than anything I can find. Good luck.

Hello philipma,
for a 85F ambient temp, this looks very impressive. But I remember you have one of those 4000rpm delta and one or two others on that S5...
I would LOVE to get 55db at 1393gh .... I run at stock everything right now with very mixed results:
1.) about 1100gh, fan is at 3240rpm, temp is 48-50C, hw 0.0008%
2.) about 1150gh, fan is at 3600rpm, temp is 48-52C, hw 0.0291%
3.) about 1050gh, fan is at 3960rpm, temp is 55-47C, hw 0.0000%

Don't ask me what batch they are from, all have identical Corsair CM750M's, they are all next to each other. Ambient temp is about 73 in my basement.
(give or take) There are also 10 S1's, 9 S3's and one freaky loud S4)

The 3rd one is the grief, while trying the shroud approach I noticed that the fan blades had damage on the inside, not the grill side, two or three blades had smaller dents and some plastic strands sticking out. It might explain the high pitched whine and noise.

The first two do about 72db and the third one is 78-80db (android app)
I hope that I can get to your levels of overclocking, noise and temps with the vertical positioning and new fans.
(But come to find out that the stock fan is 140cfm, whereas the fans ordered reach 120)

Thanks for sharing your experiences... you have no idea how truly helpful they are.

I would be interested in that as well. I tried the mod with a 120x25mm (his is for sure a 38mm) and it did not work.
I took the grill completely off...

yeah he has wanted me to test it ..  I was happy with the delta's but I should test it with the shroud.  one thing I did was find longer screws.  so the grill is about ½  inch  off the delta fan.

I was getting 55 db and 393gh with my 1 s-5 and said fuck it . it is good enough.  If you find longer screws and space the grill off the delta it does help. somewhere i have a photo of that mod.


here is the s-5 in a 85f garage running with a lot of sp20e's . numbers are good 1290gh  with 99.9922% good and 0.0078% bad  


 and I don't mind the noise as it is in the garage. I suspect the temps are high since we got to 44f today in NJ.

So you modded a fan shroud and it reduces the noise? Interesting. Do you have DB measurements?

Hello philipma,
thank you for your expert opinion. I feel much better now about the impulse purchase.

I read all about your experiments and downloaded all pictures you posted.That mod with the 4 pin floppy and molex connector is a keeper.

I found that other thread / guide about the S5's in vertical position and the improvement in cooling.
I will post there and ask you maybe another question or two....

These fans are different that what you have and everybody else it talking about though.
At least I will be able to rest easier tonight knowing that the deltas's might work to quiet down my little private basement airport. Thanks!!!

those deltas should work for you

Your ambient is more than warm, you have summer in your house...
I finally found your pics with the Noc/Scyth on your miners and see that you want to sell them, did you sell them already?

Whilst shopping for fans (some other thread had a user that was adamant that the Noc is not sufficient) I stumbled across a deal I just could not pass
up at 45 bucks. I am getting sometime this week a box of 18 used Delta AFB1212HHE-F00 with 4 pin connectors. Rated at 120 cfm, 2900 rpm it should be sufficient to quiet my basement down... And with 18, I can try push alone or push/pull on all my S5's with hopefully acceptable results....
Thank you for your replies, they gave me a good idea how to approach these 'vacuums'!!
Now I have to figure out if they would work on the S4 also. Right now with the guards off and the case open it is manageable (not really actually). I don't think I will ever go for another S4, but if this noise factor on the S5's is working out, I will get some more.
cheers

btctalk ate my long post.
In short: somehow Noctua is better than Scythe in push-maybe because of the fan geometry (with larger central part in scythe).
I tried scythe for push, but temp was rising too fast to 65c, and scythe is not PWM, but Noctua is.
However, Scythe does help to lower temp in pull position.
Noctua alone worked in push, but only at 325 or 337.5mhz setting. I did not want temp to exceed 60-61C. I have ambient at 29-30, though. lower ambient might help.

I searched the threads with 'all' your posts ... even 55db would be great. One of these S5's is 78db the others about 72ish. Why did the Scythe not work for you in the push setup? At 133 cfm it should do the same/better than stock. Rabinovich in another thread claims that the stock fan is 107 cfm?
I can't see any benefit of having push/pull with an open case. Why did you decide to go with that?
Thanks for your many informative posts though.... I did try the 'double' fan thing and it does not work for me at all....it's still 75db

not to plug it, but i have a post in marketplace (computers) with the description and pics that describes what happen if you use the Noctua NF-12 3000rpm and Scythe ultra kaze combo. Much less noise, but NOT 40 dB, that's just bs from manufacturers.
Real life noise of that config would be ~55dB, but it is day and night in comparison with 72dB howling from the default fans- I used those for less than 24hr and by the end of the day my ears were ringing nonstop.

I found this thread whilst searching to DE-NOISE the 3 Antminer S5's I have in my basement. Howling banshee's would be a better term for them... ;-)
So I looked at the interesting  pics to 'double' fan (shroud), but looking at the pics and the 120mm I have, it doesn't match. The fan I have is all around open, so there is no shrouding effect.
Spending hours upon hours searching forums and what not, I would just throw the towel and find me some good replacement fans, suitable for the job of cooling (summer coming soon) the S5 and cutting the noise from 65db to about 40.
Do any of you fellow miners have any suggestions?
I am seriously looking at this fan: Noctua NF-F12 industrialPPC 3000 PWM
but there are so many other brands/ choices like Corsair SP120, or cooler masters etc.
Does anybody have any idea what the stock fans value for CFM is?
I vaguely remember seeing something at 4:30 a.m that looked like 286 CFM...
That is a lot of cfm for an 'open' system.... I don't think a push/pull as in the S3 does any good.
I don't overclock at this time, and rather have under-clocking on my mind.
Recommendations would really be very welcomed, as well as opinions. (this water cooling thing is just not for me)

I also would like to know if the same fan solution would work in this bloody S4 that I got myself into. (These are definitely not 'home' mining friendly)
thanks

This started me out with the idea and the fact that I could not take the noise anymore.

http://www.silverstonetek.com/techtalk_cont.php?area=&tid=wh10_0061


then I thought how can I make the air more directional so I was looking at shrouds like this. This is what you use by the way not to fans.

http://pc.mmgn.com/Gallery/120mm-Fan-Shroud

Then I said wait a minute. So I ripped out a 120mm fan from a burnt out one that I found in my closet and just kept the hollow frame and then mounted the original S5 fan back on and WOLLA! noise is gone.

The trick is to add some space between the case and the fan especially with delta fans you cant have nothing to close to the blades not even a fan guard. only fan guard you can use is the one that came with it.

Take your hand and put it over the fan guard you see how much more noise there is.

It cost me nothing but time and now it sound like a s3 a tiny bit louder with the fan guard on but I am going to ad another shroud just to space the guard with the fan.

Brings down temperature as well.

 

So the net effect of this is to use a 120mm fan frame as a spacer between the original S5 fan, and the case itself? How did you stumble on this idea? If true, it's about the cheapest change I have seen mentioned for S5 fan noise.

Wait, you mounted a fan on top of another fan? Are they both blowing in the same direction? I'd love to buy 1 or 2 of these rigs, but I just can't have the noise. That's what I love about the C1's. Have to deal with a pump and coolant, but you can't even tell it is in the room. Runs Silent.

want to bring down the noise and the s5 fan do this it helps a lot.




Take a old 120mm fan rip the fan out just leave the frame remount the fan back on and wolla!! your S5 will sound like A s3 even runs cooler to.



I can sleep at night my wife is not beating me over the head lol lol. it worked for me and the best part about it its still the original S5 fan!!!

it could if you had a 300W 10.5V (approx) power supply

How good S5 with underclock ?
Ex can it work 800GH with only 300 watt ?

Thanks...

Thank you yes the voltage settings is what i am looking at they run pretty decent
My psu's are 1 antec 1200 quattro which does not even get warm at all on cables or casing even at full speed of the sp20's
The other psu's i own are a recom 1600 watt which does get handwarm at cables and casing
I am wondering which miner is more efficient the s5 or the sp20, noise wise i think the sp20 is better after i heard the s5 run over a skype call
 

I sold the SP20 units, and dont have a killawatt. My estimates are based on wire/PSU temp (the CS650M gets quite warm over 580W, and very hot in spots at 600W+, same for the PCIe cables which get barely warm at a ~300W draw each), and the DC12V draw readings of the SP20.

for the SP20 my settings were similar to most people getting 1.15-1.3TH from the device at low fan speeds. As i recall;
start: 0.635 max: 0.655 wattage: 155 fan: 10% (cold air at inlet)
running like this said 150W draw per PCIe (so presumably 600W 12VDC)

Can you post the settings you use for the sp20 so i can compare and measure power on the wall.
I am in the eu, with 220-240 volt. I think we get a slightly higher effiency on the psu's
Was hoping bitmaintech start selling a 4 boards type like the c1, that would be very good for home miners
Even without a slight overclock they would be a great solution to avoid noise and would give me the ability to place it inside the house.
Luckily i have now 3 phase 35 amps so can feed more machines than i can afford

at the wall? Im guessing that would be 1.22-1.32Th on the SP20 and 1.25-1.35TH on the S5, thouh the S5 seems to max out on 12V at about 412.5MHz (~1.32TH)

I'm going to tune both S5 and SP20 to 700 watts each, and see what performance they get.

Im measuring at the PSU, not at the wall. and by measure, I only know what the SP20 tells me (i set it to 150W/cable) and whether the PSU trips (612.5W 12V rail, seems to trip at 625W+). Right now the SP20 is at 10% fan, 18/52/47 temps, and 1.195TH. noise is about the same as my S5 at 5V fan power, 56/62 temps, and 1.23TH

update: pushed the Sp20 to 8% fan and 155W per connector. PSU seems steady so far and volume is about the same as the S5. 50/55C exhaust on it with 1.21TH. Both peices of hardware are excellent and capable of similar specs when running at 1.2TH

I haven't seen anyone get the SP20 that good. It's more like 650-700 watts for 1.2 TH/s.

The S5 can be very managable if the fan is driven at 5V, and can still push 1.2TH at about 600W. This causes temperatures around 56-62C, with 5C intake air and a makeshift top shroud of packing tape.

The SP20 is about the same volume at ~10% fan speed and can drive about 1.25TH/600W in the same ambient settings.

I dont have a killawatt, but I know neither is tripping my CS650M GOLD supplies (they trip at about 625W+ on 12V) (The S5 tripped it at 400MHz, or about 640W)

If I had to choose a winner, it would be the SP20. But with pricing considered they are about equal - I think bitmain could revisit the drawing board and come back with a quieter model that has a top shroud and is a cheaper than the SP20

swapped in an S1 fan, driven at 7V only. At this power the unit is a reasonable volume for home mining, but at 350MHz or 363.5Mhz requires some cool air intake to keep below 65C. at room temperature it would probably be at about 325MHz to maintain less than 65C

it really wouldnt change much. The SP20 changes voltage to change efficiency, whereas the S5 is fixed based on the input (12.05V). even at 412.5MHz its likely 0.54w/gh or better

I was wondering when it got to 400-412.5 what it was pulling at the wall where it would probably be not so linear at the peak.

This is far more linear then the sp20 is..    about .49 to .51 from  freq 225 to 356.

I don't, but it should be pretty linear as the chip voltage is no different, only frequency. (ie: if stock is 350MHz/560W, 400MHZ would use about 640W) One of the other reviews found that efficiency at stock was about 0.52w/GH whereas at 250MHz it was around 0.50w/gh.

Do you have a kill-a-watt meter?  If so, what is the power draw at the wall at these different settings?

This unit can overclock, but needs some cool air.

-5C intake air, 412.5MHz, 53-56C temps, 1.32TH (1.35Th expected). Any higher than 412.5MHz seems to result in losing hashrate, even if temperatures are kept under 56C
-5C intake air, 400MHz, 51-54C temps, 1.32TH (1.35Th expected).
17C intake air, 400MHz, 61-65C temps, 1.32TH (1.33Th expected). almost no HW errors at this temperature.

This is with the stock fan connected to a 7V power, no PWM. Its about 3000RPM approx.

My next step will be to swap in an S1 fan at 7V (about 2/3 the airflow) and see what happens. Should be very manageable volume, but to keep temeratures under 65C it will need either a cold intake at 400MHz or warm intake at stock 350MHz

yeah, assuming theres no random limitations, you could upgrade a C1 with S5 boards as long as the heatsinks/liquid can keep pace

Like a C2 would probably need?

It kind of looks like the controller board can run 4 blades, which would make a <$700 4blade/controller/BBB package very attractive for existing S1 owners.

If the heatsink is the same (looks like), just have to get some hash boards and controllers... 

It looks to be the exact same heatsink with holes in all the same spots, but I dont want to dissassemble everything and much up the thermal paste to confirm if its millimeter-for-millimeter identical. Im pretty sure it is though.

I added an S1 fan as a secondary, and it helped a bit, but barely. Now i have two loud fans. Also added a bit of clear tape to the top as shrouding and i think that helps, perhaps even moreso than the pull fan did. I think Bitmain needs to include a top shroud made of the same flexible thin plastic as the sides - a lot of airflow is lost out the top, just like it was in the S1

Hey Klondike, can you confirm that the S5 hole pattern matches the hole pattern on the S1 heatsink?

I think it is great how the manufactures are now sending out review units to members! I am looking forward to some S5/SP20 comparisons.

Looks great might buy one myself

To preface this review, I have owned, operated, and over/underclocked S1, S2, S3, and S3+ hardware from Bitmain, as well as Bitfury, BTCGarden, SP10/SP31, and rockminer hardware. Bitmain asked me if I would like a demo unit of the S5 and I was happy to say yes, as it was an interesting piece of hardware with pre-conceived opinions.

I want to make this review focus on some of the details that others might not. I want to use the miner with a smaller PSU and focus on specific issues rather than waste time and text talking about things like efficiency and overclocking, as this information is already available from several other reviews and threads.

UNBOXING: This is a breeze. pretty much the exact same as an S3 unit; bubblewrap, 1 peice of tape, foam clamshell, beat-up cardboard box with staples keeping it securely sealed
SETUP: Again, a breeze. Plug in power, plug in ethernet, turn on. It uses DHCP this time and findes itself an address (in my case .135). Use an ip scanner to locate whaich ip it chose. in comparison the SP-tech hardware is very easy to find via http://start.spondoolies-tech.com/.
POWER: I am using a Corsair CS650M GOLD PSU. It has 2 paired PCIe cables (4 connectors total). I used one paired cable per blade of the miner.
MINING: Once connected the unit began mining. accessing the web interface with root/root as before presents the well-known interface and after entering my eligius credentials I have 1.17TH speed average over 20 minutes. This miner is easily capable of meeting spec right from the box, compared to the SP20 which only achieves spec in very cold environments.
NOISE: This is not an S3 - its much louder. The fans run at 100% for the first 10 seconds at boot, then tame down until the unit starts hashing and heats up. at that point, fans are up at 3600-3900 RPM (according to GUI) and loud. Sound is almost identical to a vaccuum cleaner, and sounds similar to a 1.3A rockminer fan or a wall of s1 units. noise quality isnt bad, its loud but not shrieking. A relatively loud vaccuum cleaner is the most accurate description.
OVER AND UNDER CLOCKING: Underclocking down as low as 225MHz was tried, with a linear adjustment in hashrate and temperatures reducing from 54C stock to 46C. Fan speed was reduced, but only by 600rpm, and is still too loud to conceive running in the same room as you work. Overclocking was tried to 362.5MHz and seemed to handle just fine. hashrates and temperature up slighty, fan noise stayed about the same as at 350MHz
BOARD TEMPERATURES: at stock speed and 20C ambient the readings are about 53C. With 0C intake air and 225MHz the unit reads 39C with the fan at 2880RPM. The exhaust air is far less, possibly only 10C. I think this difference will become an issue because right now the stock fans run far too fast most of the time, and I think setting higher temperature targets is possible. 2880RPM at 39C is silly, thats like 75% fan speed for about half the power usage and a fresh air intake.

so, the next step is pictures!
first, lets compare this to an antminer S1: Apart from a different controller, the unit is almost identical. The frame and heatsinks are the same (frame ends are flipped though), and the real difference is the presence of plastic side panels, which is something that was missing in the S1 but overkill in the S3. Assuming screw locations are the same you could easily have an S1->S5 upgrade kit. An interesting difference is that the screws are all spring-loaded for whatever reason. (i found that at least half the screws could be tightened about a 1/4 rotation more though)


Finally, I think the most important aspect of this review is to bring up the pros and cons of the hardware as i see it
CONS:
1) the final picture seen above shows that the reset button is not aligned with the hole in the frame. This is minor, but worth pointing out I feel. could make resetting a tad more difficult.
2) THE FAN IS LOUD!!! Seriously Bitmain, install a second fan like on the S3. I think some extra shrouding at the top could help also, just another simple piece of plastic with 4 tie-down points. I plan to install an S1 fan as a secondary soon and see how that helps.
3) undervolting the hardware is not as easy as a pencil mod or software tuning - it requires a strong 9V or adjustable power supply that few people own, and isnt necessary with the SP20
4) ITS LOUD. Im saying this again because you will not be able to run this in your living room like you could with the S3. You can hear the fan through the closed doors.

NEUTRALS:
1) It isnt fully enclosed at the top or bottom. This wont affect the majority of people and makes the unit well suited as an inexpensive miner for farms. For the home user though this might frighten off people who dont want to have things falling onto/into the miner. Theres no real risk of running this on a metal surface that i can tell, as the boards are elevated from the bottom slightly
2) DHCP - its nice to see this choose its own IP, but can make it difficult to find without an IP scanner. For the bulk user though, this saves the tedious process of setting up only one machine at a time at 192.168.1.99
3) Stock settings are enough to make the cables on my CS650M get warm. not hot by any means, but warm. A lower quality power supply might get concerningly hot cables if they are not made with thick wire.

PROS:
1) at ~0.51w/GH stock, its very power efficient.
2) Simple design. likely much cheaper to produce than the SP20, and easy to setup.
3) comes with a stripped-down beaglebone that could be repurposed or used again on future hardware versions when the S5 is no longer in use.
4) a secondary fan can easily be installed and there are holes available to screw one onto.
5) The Minerlink 'findyourminer' application has the ability to tell any specific S5 units to beep loudly and flash the LEDs - This is a fantastic idea for anyone who wants to distinguish miners from one another when dealing with DHCP changes for dozens or more of the S5

Conclusion: Its a decent miner, but in its current revision it is too loud for the home user. Underclocking could improve the volume, but it does not improve efficiency and in contrast the SP20 is a better miner if you want to underclock/undervolt for efficiency. I think Bitmain needs to revisit the drawing board for this miner, and come back with a version that has 16 chips per chain to achieve ~0.44W/GH at a similar hashrate but lower power consumption, and/or a secondary fan to keep the noise down


edit: I have a trio of SP20 units delivering tomorrow, and will follow-up this review with a better comparison of the two units