The combined sidehack-novak usb stick review thread. AKA GekkoScience BM1384

[alh]

While I find the discussion of running a "full node" fascinating, can I ask what the motivation is? My simplistic view is that it's kinda like running Seti@Home in terms of "value", however you define value.

I don't want to hijack this thread from it's intended purpose, but thought I would ask. Feel free to delete this post if you wish.

[philipma1957]

While I find the discussion of running a "full node" fascinating, can I ask what the motivation is? My simplistic view is that it's kinda like running Seti@Home in terms of "value", however you define value.

I don't want to hijack this thread from it's intended purpose, but thought I would ask. Feel free to delete this post if you wish.

well I knew I had enough parts to rebuild 2 pc's both with low watt intel t cpu's

so if you are building a 'free' pc that idles at 23 watts and runs a node at 25 watts


vs a rasp pi the uses 10 watts.

 cost me 0 either way


  you have to ask yourself  for 15 watts  or 11 kwatts a month extra or 2 bucks a month 24 a year .

what is better to run sidehack's gear.

I would be willing to say the pc build will allow me 20-25 sticks using the 2 stud hubs I own.

I would be willing to say the pc build will allow me 8 to 10  18 chip boards.

 I think the rasp pi won't do that.   So the node  is really a  2-3 watt gift to bitcoin on my part.


I am not sure how many sticks and or boards will come to me  as a user and as a kit seller.

We need to see how it unfolds.

But picture a  small pc built to run cgminer / bitcoin node doing 25 sticks + 10 boards.

vs a rasp pi bogging down at 10 sticks and  or 4 boards .

Remember this is All conjecture since no one has run  more then 5 sticks on  1 hub.

So no one knows if that is an issue for rasp pi's or cubies or beagle bones or is that bugle bones.

So my 'free' pc build  may drop the node  if I determine the node lowers my sticks and or boards hashing.

This is the fun part of the gear testing limits figuring what's a good setup etc.

Time will tell.

[novak@gekkoscience]

That's pretty cool, I just ordered one. Cubie draws around 2-3 watts I believe. You can run a full node off of an rpi or cubie as long as you allocate enough swap space, but it will run dog slow. I think you would need at least 2gb of RAM to run a full node smoothly but I don't believe there aren't any SOC boards with that much RAM yet.

There are SoC boards with 2GB RAM, including the cubietruck (cubieboard 3).  It is not often terribly cheap to get a 2+GB RAM board, however.  I would vote against using much swap space on an SD card, it is slow at best.

I haven't personally played with it too much but I think it's likely that the required memory could be reduced, there are some flags for bitcoind which ought to help.  If 1GB is enough then a banana pi or raspberry pi 2 would work (both pretty cheap).  I have some experience putting together low-memory linux environments so it's possible we could get something to work.

Something to consider if you want to run a full node is how much hard disk space you'd need.  You might wish you had a spinning drive- which would make swap easy (and probably less slow).  However, you obviously add cost- and power.

--
novak

https://learn.adafruit.com/external-drive-as-raspberry-pi-root?view=all

This might help get the Pi running on a hard drive.

i would imagine running a hard drive off usb would be slower then running off a uhs class usb flash drive..
like this
http://www.newegg.com/Product/Product.aspx?Item=N82E16820226378
plus the drive with all those adapters would use more power.

Probably but why run a drive over USB when you have a SATA port (as in, how about using any of those SoCs besides the pi?)? 

Besides, using flash as swap space is kind of a no-no.  Not that it's likely you'll burn one out in less than several years, but you're using something in a way it wasn't really meant to be used.

--
novak

[philipma1957]

amazon meter arrived. some photos

full size links
http://imageshack.com/a/img901/5221/hytHGL.jpg
http://imageshack.com/a/img908/5674/pLpZKh.jpg
http://imageshack.com/a/img540/3377/HdwIvD.jpg

meter looks nice and it flips stick so tweaking stick is easy



close up


and we are pullin 9.5 gh or so

[philipma1957]

now I had a good  meter for 1 stick  so I dialed back power  the pot is  now  in this position  


+   in stead of ⅛ clock wise     the readings dropped from .66 amp to .6 amp

 so .66 amp x 4.91 v =  3.24 watts  and 9.5 gh


now .6 amp x 4.91 volt = 2.94 watts and 9.5gh hash

or 3.24/9.5 = .341 watts a gh

2.94/9.5 = .309 watts a gh


.309 watts a gh at freq 175 is really good and most likely accurate within 5%

[cavaliersrus]

so was the invalid ones due to power being to low in that btw very neat seeing these numbers from a usb drive .... next question how is the heat on the heatsink ?

[TheRealSteve]

Should be able to get slightly better performance still if the timeouts can be eradicated.  Theoretical at 175MHz is 9.625Gh/s, which you're already pretty close to in the average measure

[sidehack]

If it's not reporting HW errors during the delay, the "no valid hashes detected" timeout is still based on U3 timings, which the duration is frequency dependent. The U3 would have four times the hashrate at a given frequency (as it has four chips) so cgminer says "well heck, I haven't seen a share in this long it must be hung and needs a hard reset" but that's not really valid for the stick because it's a single-chip device. The "we have a problem" timeout should be four times longer. That'll be taken care of when we have our own driver.

Looking good, Phil. I won't get to play with my meters until probably Thursday. Thanks for getting some good numbers up.

[philipma1957]

so was the invalid ones due to power being to low in that btw very neat seeing these numbers from a usb drive .... next question how is the heat on the heatsink ?

with the fan blowing on it no worries.  I think I will turn the fan off for an hour. I don't have an infrared gun but I have a temp for food which if placed on the heatsink should give a good number.

27c with fan turned on

40c with fan turned off

full size
http://imageshack.com/a/img538/6752/j5BF12.jpg
http://imageshack.com/a/img537/6362/4cO2FR.jpg
http://imageshack.com/a/img537/415/R8jtxt.jpg
http://imageshack.com/a/img633/1166/35lmqw.jpg
http://imageshack.com/a/img540/2181/czebCV.jpg


went from 27 to 32 c with fan turned off


moved up to 38c



it is hard to read but 39.6c and I played with lowest amp setting here at 0.57 amps = constant hw's



fan drops it down fast to 28c this is in a 25c room temp   notice amps at 0.61 this is much better for HW's


lastly look at HW they are at 698

[TheRealSteve]

Maybe after the USB meter, the next thing to get is a thermal camera module?   ( FLIR reacted to the SEEK by dropping their prices a fair bit.. lower resolution but generally higher quality. )

[philipma1957]

Maybe after the USB meter, the next thing to get is a thermal camera module?  ( FLIR reacted to the SEEK by dropping their prices a fair bit.. lower resolution but generally higher quality. )

yeah but the oven temp meter is pretty good. cost = $0



I have also decided to  bump amps to 0.61-0.62   vs 0.6


reason is at 0.6 HW is 2%  at 0.61-0.62 HW is under 1%

so my stick seems very happy at freq 175 and 0.61-0.62 amps  still is that nice 0.31 watt per gh number.

[TheRealSteve]

yeah but the oven temp meter is pretty good. cost = $0

A very reasonable argument   If it piques your interest anyway, let me know

[philipma1957]

yeah but the oven temp meter is pretty good. cost = $0

A very reasonable argument   If it piques your interest anyway, let me know

of course it does and I want one. but so many parts with so much gear in house  and a very patient wife .

I don't want to add any straws at the moment.

Today the usb meter came an intel cpu came. some fan grills are due. some more parts tomorrow.

I have to build a pc in the next 2 days.

z97 asrock itx mobo
i5 4570t cpu
2 8gb sticks of ram
1 250gb msata
fsp plat 450watt psu
a silent zalman cpu cooler

s much as I would like the meter no more gear for a week or two. (or not and I will say fuck it and ask you for a link to buy the meter)

[wpgdeez]

Maybe after the USB meter, the next thing to get is a thermal camera module?  ( FLIR reacted to the SEEK by dropping their prices a fair bit.. lower resolution but generally higher quality. )

yeah but the oven temp meter is pretty good. cost = $0



I have also decided to  bump amps to 0.61-0.62   vs 0.6


reason is at 0.6 HW is 2%  at 0.61-0.62 HW is under 1%

so my stick seems very happy at freq 175 and 0.61-0.62 amps  still is that nice 0.31 watt per gh number.

What is the power draw when you figure in the host PC?

[philipma1957]

Maybe after the USB meter, the next thing to get is a thermal camera module?  ( FLIR reacted to the SEEK by dropping their prices a fair bit.. lower resolution but generally higher quality. )

yeah but the oven temp meter is pretty good. cost = $0



I have also decided to  bump amps to 0.61-0.62   vs 0.6


reason is at 0.6 HW is 2%  at 0.61-0.62 HW is under 1%

so my stick seems very happy at freq 175 and 0.61-0.62 amps  still is that nice 0.31 watt per gh number.

What is the power draw when you figure in the host PC?

read the thread all answers are here.  Does not that seem fair to write.

here is break down

  total of   36 watts for   the  entire setup  I don't count the monitor as it is turned off 23 of 24 hours

----------------------------watts

1)pc --------------------23-25
2)usb hub -------------  4-5
3)fan -------------------   1
4)usb meter -----------   1

5)1  stick --------------  3

The pc runs a full node and uses 23 watts idle 25 watts on a node

A rasp pi would do 10 watts idle and no node. So since I choose a real pc my penalty is 13-15 watts over a rasp pi, cubie, bugle bone etc

I am just waiting for a chance to pop in more sticks to see practical limits for sticks  + boards.

[philipma1957]

and lastly the .61 amp setting does give hw  errors but not to many.

 i may tweak it to .62 or .63 amps looking for a little less on hw's

[philipma1957]

I am now starting  freq 200 setting .


My meter reads 4.89 volts x .75 amps = 3.6675 watts  

 I seem to be running  11gh  this is 0.3334 watts per gh

Hw are low

I added the bridge  to give me lots of amps pictures coming


power used is good +  < ⅛ clockwise from the blue plus





good numbers 10.93 with 13 errors and 6688 good

[aarons6]

hmm so if you can hit freq 200 with .33w/gh with the s5 chip on a special single chip board, wonder if someone can modify/undervolt the s5 to do it?

[bgibso01]

I'm just sitting here shaking my head in amazement.  The next round of miners ought to be crazy efficient.  Keep up the reporting.  This is good stuff.

[philipma1957]

I just wish I had this now I could do some summer mining even with my 18 cents a kwatt