ANTMINER S5: a freestyle review by Rabinovitch

[Rabinovitch]

Greetings, miners!

Finally I got my test unit, later than all others, so let's try to play with it. ))

All of you have seen the S5 many times, so here is only one S5 photo, from an unusual point of view...  


I've unpacked my new toy, and what do I see? Dust!



Oh, it's a test unit, so it doesn't matter. Besides... Never look the gift horse in the mouth.  

And that plastic side panels looks so... Unreliable...  

I found that the controller board is not neatly connected, but it looks that this is the only possible position:





Also I found a slot for a flash card, and it's empty.



I thought that "now I need to buy a card, flash it with some image..." But S5 started to work without any of that actions.  

Well, PSU is connected (Chieftec APS-850CB if any - 850 W, 80+ Bronze, two 12V rails, and it has 4 (!) separate PCI-E power cables). Let's go!

...After switching a PSU on I realized what does it mean when other forum members writing in their reviews that S5 is loud. YES, IT IS.   Especially right after powering on. It became a very little bit quieter after several minutes, but I still can hear it from another room even if the door to where S5 placed is closed. It's like a vacuum cleaner, someone already told this before... I have no a noise meter (and I doubt that I'll measure any new values, they all has been measured before me), but if anyone need, I'll take a noise meter for a while.

Right out of the box, at stock frequency 350 it gives about claimed 1.1 Th/s at Slush's pool:



Average speed in miner's GUI is a little less...

Working on BTCGuild pool the numbers are the same:



At BTCGuild's side:



Power consumption at the wall is 616 Wt with my PSU and at stock frequency.

Well, same value claimed at BITMAIN's website:



It's all on stock firmware:



Now let's try to update the firmware to newest version (in my case it's SD-S5-20150107.tar.gz) and see what will change. I've upgraded FW strictly according to S5's manual, and everything has been finished quickly and fine:



After a while I found that cgminer uses only half of controller's CPU time - before upgrade it was 90:



May be it's OK, I don't know, probably only Bitmain knows...  

Hashing speed seems to be the same with this new FW, but there is appearantly less discarded shares quantity (calculating to discarded shares per hour). Sure ckolivas knows why. )) May be it's because of newer cgminer version (4.8.0 in this firmware)?

After 47 minutes on BTCGuild:





With normal temperature in the living room the boards of S5 are 51 C and 55 C with this stock fan. And it looks like huge part of air from the fan is blowing away upward from this open semi-case (since my S5 is placed at the glass shelf for TV and audio equipment), so the idea of making a kind of cover is looking not so useless...

I have the Navipower HDB-1212UA fan (it's very medium-priced but... 2800 RPM... 105.81 CFM...), will try it later. Another fans are coming too but not yet in my hands...



If S5 be able to work with reasonable temperatures using that fan - it would be great.

Later I'll share another results of testing and experiments. ))

[1714900104]

1714900104

[Rabinovitch]

Now several words about overclocking.

At frequency 400 we can see average speed at miner's side about 1.3 Th, and 1.36 Th at BTCGuild side:





Comparable values we see at Slush's pool (they are a bit lower again - why?..):





Further increasing of frequency (to 412.5) gives us almost same hashrate while the temperature slightly rising up (about 1 degree on each blade; but it looks that it's just because of opening and closing the door):





If we can trust to M's Ant Monitor, there is no any huge HW error rate at 412.5 frequency. So why S5 can't increase the hashrate?..  



The same at Slush's, nothing new.







Further increasing of frequency (to 425) gives us completely nothing but more HW errors:





So we can conclude that 400 MHz is a limit for in some way effective overclocking in the hashrate rising sense.

I don't care about power efficiency, my opinion is that I should mine as many BTC as possible and as soon as possible. And come what might!  

P.S. after all that I made some additional measurements at miner's side (I started to measure HWE with M's Ant Monitor a bit later, after measurements for frequency values 356.25 and 375), about 1 hour and more for each frequency:

Frequency 350 (default) - 1157+ Gh/s, 0.0001% HWE (1 HW per hour), 616 Wt
Frequency 356.25 - 1160+ Gh/s, 616 Wt
Frequency 362.5 - 1200+ Gh/s, 0.0004% HWE, 626 Wt
Frequency 368.75 - 1200+ Gh/s, 0.0004% HWE, 639 Wt
Frequency 375 - 1220+ Gh/s, 647 Wt
Frequency 381.25 - 1250+ Gh/s, 0.0045 HWE, 657 Wt
Frequency 387.5 - 1270 Gh/s, 0.0093 HWE, 665 Wt
Frequency 393.75 - 1270+ Gh/s, 0.0156 HWE, 677 Wt (upd: 1288 Gh at PMpool.net, diff 4096)
Frequency 400 - 1300 Gh/s, 0.0258 HWE, 687 Wt

At frequency 400 and with PSU Corsair RM1000 80+ Gold power consumption at the wall is about 666 Wt (peak value), hashing speed is 1301 Gh/s (after 11+ hours of running), and HWE is 0.013%.

Interesting thing is that temperature of blades remains almost the same at each frequency while the fan's speed is constant (4080 in my case)... For last value it's equal to 53 and 57 C.




p.s. I found one strange fact. At frequency 400, right after power up power consumption is about 100 Wt (at the wall with 80Plus Bronze PSU I mentioned before):



After a while in idle mode (here I mean that ethernet cable is not connected to S5) power consumption is... Guess how much?..



It's an upsetting fact... After ethernet cable connection power consumption increases a little, and finally at 400 MHz frequency it's about:



Why S5 doesn't became REALLY idle of it has no connection to Internet?..  

[Rabinovitch]

Well, I have replaced the fan with Navipower HDB-1212UA as I promised before. It's 3-pin fan and I have connected it to controller board instead of stock fan (there it is, laying on the floor).





There is no need to cut something or to find any special ethernet cables:  



1 hour and 30 minutes of silence:





Sure the temperature is a bit higher than we want to see, but the only noise S5 generates now is like a noise of a PC under some load.   If I place it to more cool environment the values will be 1-2 degrees lower.

Later I'll try some another fans as they arrive.



Update: here are the results of another fan change, this time to another cheap fan Gembird DC120SF-12AS3 (120x120x38 mm, 3000 rpm, 124.74 CFM) for air intake:



S5 became much louder than with Navipower HDB-1212UA, but still a little quieter than S3. Now let's take a look at temperature:



From 6 C to 7 C less! Good, but still not enough.

Later we will try two fans configuration (2 Gembird DC120SF-12AS3 and 2 another inexpensive fans NaviPower HDS-1212MA). I remind you that goal is to strike the happy medium between noise and temperature of the blades.  

[Rabinovitch]

Experiment 1.

I wanted to direct more dense airflow between the radiator fins, so I pasted an insulation tape stripes on side gaps under the fan, in this way:



Well, it looks that there is nothing that needs cooling (because there are no any heatsinks ore something like that, right?  ), so more dense airflow should cool down the chips under the big main heatsinks more effectively, right? That's how I was thinking then...

The fan remains the same as before - Navipower HDB-1212UA. With that blue stripes on side gaps S5 became a bit more noisy and started to mine. BUT:



Investigating this I found that air blowing through that side gaps cool down that little elements (I am not familiar with electronic components) which seems to be a power-operated element.



They are very hot, something about 80-90C. It means that I should find some little (copper, I believe)  radiators like those being sold im many places:

http://www.newegg.com/Product/Product.aspx?Item=9SIA2RP14G1571&cm_re=chips_radiator-_-0Z5-000E-00003-_-Product

and stick it with thermal glue. Anyway it is a good mod to make.   And I recommend it to all who buy the S5. The more heat removal - the less failures of your miners...

p.s. now my S5 is OK and waiting for another experiments.  


Experiment 2.

Still have no time to describe it, but I replaced the thermal compound, twice, and with AlSil-3 my results are not so good as I expected (too low thermal conductivity...). And with Arctic MX-4 I just achieved the same temperature as with stock thermal grease... Seems that Bitmain use GOOD thermal grease. )

 More details will be here later.

[aurel57]

Your fan speed it abnormally high as even OC at 375 freq mine does not go about 3600 but then my temps are 49/48. So I take it you have it in a warmer place than mine?  

EDIT: my fan is staying at 3480 after the new firmware but hashing a little less now at 1230 GH at 375 Freq

[Rabinovitch]

Yes, my S5 is placed in in the living room temporarily, and I am closing the door to that room from time to time to avoid listening that hoover...  

And such warm ambient is good for my gouldian finches.  

[BitLeafClover]

Awesome review. Thanks. Here's a question that will leave me feeling either very dumb or very bummed:
Do you need to do anything else after connecting the Ethernet cable and 4 PCI-E 6pin cables, plugging in the PSU and flipping the PSU switch on?
Because I've done that with both my S5s, using 3 different PSUs (2 of which are brand new), and neither are turning on. So my hunch is my S5s came DOA, unless I'm missing something.
Anything stand out to you that I could be missing?

[aurel57]

Awesome review. Thanks. Here's a question that will leave me feeling either very dumb or very bummed:
Do you need to do anything else after connecting the Ethernet cable and 4 PCI-E 6pin cables, plugging in the PSU and flipping the PSU switch on?
Because I've done that with both my S5s, using 3 different PSUs (2 of which are brand new), and neither are turning on. So my hunch is my S5s came DOA, unless I'm missing something.
Anything stand out to you that I could be missing?

Did you paperclip the PSU's?

EDIT: need to paperclip the green and a black wire (either side of the green) this acts a the switch on a computer since there is not one. Then you can use the switch on the back of the PSU to turn it off/on.

[visdude]

... I have no a noise meter (and I doubt that I'll measure any new values, they all has been measured before me), but if anyone need, I'll take a noise meter for a while.

The free Android Sound Meter app works pretty good.

[BitLeafClover]

Awesome review. Thanks. Here's a question that will leave me feeling either very dumb or very bummed:
Do you need to do anything else after connecting the Ethernet cable and 4 PCI-E 6pin cables, plugging in the PSU and flipping the PSU switch on?
Because I've done that with both my S5s, using 3 different PSUs (2 of which are brand new), and neither are turning on. So my hunch is my S5s came DOA, unless I'm missing something.
Anything stand out to you that I could be missing?

Did you paperclip the PSU's?

EDIT: need to paperclip the green and a black wire (either side of the green) this acts a the switch on a computer since there is not one. Then you can use the switch on the back of the PSU to turn it off/on.

Never been so happy to feel dumb. Thanks buddy! It really does sound like a vacuum, its so loud

[Rabinovitch]

... I have no a noise meter (and I doubt that I'll measure any new values, they all has been measured before me), but if anyone need, I'll take a noise meter for a while.

The free Android Sound Meter app works pretty good.

You wouldn't believe, but my first Android smartphone that I have ordered before New Year holidays hasn't been delivered yet. 

[coinits]

Well, I have replaced the fan with Navipower HDB-1212UA as I promised before. It's 3-pin fan and I have connected it to controller board instead of stock fan.

I can not find that fan in North America. A Google Search just takes me to Russian sites. An English-only search just leads me back here. Is there an equivalent on this side of the ocean?

EDIT: How many RPMs does it give? I have found equivalents.

[Rabinovitch]

On this photo you see the stock S5 fan.

You can replace my Navipower HDB-1212UA with any 120 mm fan which:

0) has 3 or 4 pin connector and better if it is PWM fan (you should cut one leading plastic sliding rail in case of 3-pin connector);
1) can spin at 2500-2800 rpm (Schytche Ultra Case maybe?);
2) can provide 100-130 CFM airflow (many of forum members are using Delta fans);
3) as quiet as possible amongst meeting the above conditions fans.

It was hard to find an airflow value for this fan, but I did it, I found it in some chinese catalogue. It says that this fan gives 105.81 CFM and spins at 2800 rpm, but miner's GUI displays 2400 rpm.

PM me if you need help searching such fan(s).  

[coinits]

On this photo you see the stock S5 fan.

You can replace my Navipower HDB-1212UA with any 120 mm fan which:

0) has 3 or 4 pin connector and better if it is PWM fan (you should cut one leading plastic sliding rail in case of 3-pin connector);
1) can spin at 2500-2800 rpm (Schytche Ultra Case maybe?);
2) can provide 100-130 CFM airflow (many of forum members are using Delta fans);
3) as quiet as possible amongst meeting the above conditions fans.

It was hard to find an airflow value for this fan, but I did it, I found it in some chinese catalogue. It says that this fan gives 105.81 CFM and spins at 2800 rpm, but miner's GUI displays 2400 rpm.

PM me if you need help searching such fan(s).  

I should be able to find one now. I have one S5 on order as a test and will go from there.

Thanks for the help!

[Biodom]

On this photo you see the stock S5 fan.

You can replace my Navipower HDB-1212UA with any 120 mm fan which:

0) has 3 or 4 pin connector and better if it is PWM fan (you should cut one leading plastic sliding rail in case of 3-pin connector);
1) can spin at 2500-2800 rpm (Schytche Ultra Case maybe?);
2) can provide 100-130 CFM airflow (many of forum members are using Delta fans);
3) as quiet as possible amongst meeting the above conditions fans.

It was hard to find an airflow value for this fan, but I did it, I found it in some chinese catalogue. It says that this fan gives 105.81 CFM and spins at 2800 rpm, but miner's GUI displays 2400 rpm.

PM me if you need help searching such fan(s).  

Scythe Ultra Kaze did not work well in push orientation-see my post on main S5 board.

[nexus99]

Great review! Thanks for taking the time to write and post. Are you able to measure power draw at the wall?

[Rabinovitch]

Unfortunately no... If I decide to buy a power meter, it will be shipped from Moscow and it takes about 4-7 days!

Now I got this marvel of Chinese craftsmanship:





I don't know how accurate it is, but all my posts in this review will be updated with it's measurements soon.

I have also ordered a little radiators for my next cooling mod. They will arrive in 10 days. 

[KeezAZA]

Any updates on your cooling methods?

[Rabinovitch]

I am waiting for Arctic Silver thermal adhesive to arrive. Little radiators are already in hands so on 26.02 or 27.02 there will be some update. 

[gemstone]

On this photo you see the stock S5 fan.

You can replace my Navipower HDB-1212UA with any 120 mm fan which:

0) has 3 or 4 pin connector and better if it is PWM fan (you should cut one leading plastic sliding rail in case of 3-pin connector);
1) can spin at 2500-2800 rpm (Schytche Ultra Case maybe?);
2) can provide 100-130 CFM airflow (many of forum members are using Delta fans);
3) as quiet as possible amongst meeting the above conditions fans.

It was hard to find an airflow value for this fan, but I did it, I found it in some chinese catalogue. It says that this fan gives 105.81 CFM and spins at 2800 rpm, but miner's GUI displays 2400 rpm.

Danke vielmals / spasibo for your most excellent thread.
Are you certain that these stock fans are 105 CFM?
The 3 howling banshees in my basement are between 72 - 78db with the stock fans and NEVER EVER seem to go at less than 3800 rpm (not overclocked either, all is stock) at a temp of about 48 - 52C.

I don't understand why a scythe rated at 133cfm is not sufficient for cooling as some users have claimed in other threads.
I expect some used Apple 4 pin Delta AFB1212HHE-F00 sometime this week, that are rated
2900 RPM 120.07 CFM  44 dBa and hope they will do a better and MUCH, MUCH less noisier job.

I'm also not as brave as you to run the miners at anything over 70C.

Thank you for your courage, and I will be looking forward to your further documentations of your experiments. (are you giving the vertical orientation of the miners any consideration?)

mit Deutscher Liebe von Kansas USA :-)