Topic: [CLOSED] ASICMiner Prisma 1.4th/s - 1.47 BTC - page 6. (Read 49994 times)

"My main reason for the first post was trying to figure if a 260 oc is better then a 200 uc."
Roger that, why we're doing testing at various different clock rates.
Think it's a forgone conclusion that H/W error rate will rise with increased clock speed.

There is a fine line balancing act between initial cost, cost of operation, and difficulty increases.
I'll stick my neck out here and say that any gear can post excellent W/GH/s if de-clocked enough.
Heck, look at what guys are doing with de-volting and de-clocking Bitmain S1's.
Problem with de-clocking is hash rate falls with it and given the network difficulty increases with reduced hash rates it may never pay for itself.

I would re-write your list with each line item carrying an equal weight to read:
1) initial cost ($/GH/s)
2) profit potential (W/GH/s)
3) reliability/dependability
4) heat + safety

Regarding reliability/dependability, the amount of effort required to cause a miner to restart mining (after say a power failure), and the length of time required between miner reboots.
The Bitmain S1's are the epitomy of dependability. Plug'em in, config'em, OC'em, and forget'em.

"I did find that my power went from 985-90 to 1033-1049 watts.  the meter jumps back and forth with the readings."
Your meter might be sensitive enough with a fast enough sample rate to be seeing the power fluctuations caused by varying power requirements during hashing.
 
"So temps may play a big part in power draw."
In our observation across many different machines, incoming air temp most definitely causes changes in power consumption.
As a rule-of-thumb, the warmer the air the more trons they use. Colder is cheaper.

"I keep getting the feeling this gear will need too much care to be safe."
Like much of the new releases of mining H/W, it takes the community/individuals a butt load of trial 'N' error to find the "sweet spot" for operating parameters.
I think with the kind of chip densities the Prisma has this "sweet spot" will be harder to find.

At this juncture I'd like to give a "shout out" to friedcat for designing the Prisma, to AM for manufacturing it, to CG and Canary for selling them, and to gekkoscience for making the adapters boards that power these units.
All of the above are "pushing the envelope" in search of better, more profitable, cheaper mining technology.
Thanks Guys, couldn't do it without you.

"For your farm with multiple units emmersion cooling may be best."
In bulk (16-32 hash boards/tank), immersion cooling makes sense.
On a "onesy-twosy" basis it drives the initial cost to the point of non-profitability.

 

So i have got 1/3 of prismas on Friday so far. But now to be honest i dont want to get the rest . I thought after so many Asic miners was made these would work without much hassle.
But its exactly the opossite . Was much easier to set up 5 gpus litecion mining rig than this crap.
I was trying to make this work for 2 days without any luck. Most of the time i get these stupid "BET0: Unexpected value received" or sync errors.
Boards are connected in the right order with the switches being 100% correct.
I managed to get it running twice on pull power , just randomly , suddenly was running fine. Sometimes it boots up one or two boards , or just few chips, keeping 100ghs - 500ghs hashrate....

I have tried these tricks with switching to RM cgminer and back to AM , no help.
Turning cgminer off , unpluging cable , plugging back and running cgminer , nothing.
Imaged Rpi with original minera , compiled cgminer , same problem.
Tried raspbian with compiled cgminer , no change.
Tried even different computer , with windows, Linux mint and Ubuntu. Still getting the same errors.
Changing all the cables , trying different PSU's , plugging in different Usb slots on the Rpi.
Tried running the boards separately , sometimes they boot up , sometimes they dont , but never on full power.


So i guess there is not much hope to run this crap , unless someone has miraculously manage to get this work?

happens .  I have other points besides safety.

I have run this overnight and I am at 104f-105f in a closed space. ( a 235 square ft garage )

in terms of 105 f that is 40 c   once I open a door temps drop to 75f or 23.89c

my power draw is 985 watts  that is 985/1201      0.8201 watts per gh  this power number include all my  extra fans

if I remove the  extra fans I drop to 975  watts that is 0.811 watts .

 I don't knock oc but if your oc gives you .9 watts vs  my .811 watts ( due to the more power costly fans) besides heat which looks to be no problem what about

power cost?  

not to mention my error rate is .88% and your error rate is 6.62 %

My main reason for the first post was trying to figure if a 260 oc is better then a 200 uc  .

1)power
2)heat + safety
3)profit

Now when I set this up last night power draw was 985-990 watts.

 It ran 11 hours with the garage door closed temps rose from 75 f to 105 f. temp seems to get stable at 105f
I did find that my power went from 985-90 to 1033-1049 watts.  the meter jumps back and forth with the readings.   So temps may play a big part in power draw.  I just opened my garage door and I will see if the temp drop down lowers the power down.

I keep getting the feeling this gear will need too much care to be safe.

For your farm with multiple units emmersion cooling may be best.

 


this is the unit with garbage door open letting in 43f or 6c air


in 2 hours time temps dropped a lot and power dropped a lot

OOPS on the previous post, just noticed your temps are in F not C.

MY BAD!!

No clue.
When we initially fired them up we were powering them with EVGA 1300's feed with 120V with a Kill-A-Watt monitoring watts.
Stats were ~1250W @ 240, ~1300W @ 260, and ~1400W @ 270.
Don't have a power meter that will measure @ 240V, hence my cluelessness.
Now powered with DPS1600BB's feed 240V which have about the same efficiency as the EVGA's feed 120V.
So the power draw @ a given clock rate will be in the ballpark.
Hence my use of the "~" character indicating approximate.

If any one knows of a cheap power meter that will run @ 240V I'd sure like to know.
Because I'd sure like to have one in my inventory of test equipment

Gross power draw is not the only consideration when considering safety.
Heat is what causes these things to fail (sometimes violently).
Excessive heat is what caused all the units to fail with all the ugly pictures posted in this forum.
More precisely, not being able to dissipate the heat is what caused the failures.
Configured and mounted vertically, as in the pictures I posted, heat is being managed, keeping chip, board, and cap temperatures within tolerable and safe ranges.
What we are doing is finding out what it takes to safely run these things at elevated clock rates on air.

This elevated clock rate testing speaks to the second part of your post which concerns itself with W/GH/s which is used to predict the most cost effective air cooled clock rate to operate at.
Very likely operation with air cooling @ stock (240) or de-clocked rates will be the most cost effective air cooled operating mode.
But one doesn't know unless tested in the real world to develop stats.

We're a business. We bitcoin mine to make a profit. We are not hobbyists. So finding the most profitable mode of operation is our goal.
We may determine with this Prisma analysis that there is more cost effective air cooled H/W.
The next phase of our testing will be these Prisma boards mounted in 2 Phase Open Bath Immersion Cooling Tanks.
But developing an air cooled base line is the first step.

What follows is a snapshot of our RPi controlled Prisma clocked @ 260.
In my feeble mind 2 things jump out at me versus the image you posted of your RPi controlled Prisma.
1) There is a difference of hashing rate ~300 GH/s.
2) There is a difference of operating temperature of ~70C (158F)

Almost 100% sure that the temp indicated by Minera is the temp of the RPi near one of the vertical connectors on the RPi B+.
The RPi that is controlling ours sits in the incoming air stream and is specifically positioned there to give an indication of incoming air temp.
The RPi that's controlling yours is (by it's own internal measurement) operating some place hot enough to boil water.
If the Rpi is that hot, how hot is the Prisma? and the components contained thereon.
While not having a pic of your configuration, Your Prisma may be operating at or near a thermal failure point. Even in a de-clocked operating mode/state.
Just food for thought.



While the bulk of our air cooled hashing power is/are Bitmain products, we have quite a few different hashing units in our environment. Mostly acquired for immersion cooling testing.
None but the Prisma have the board densities suitable/optimum for OBI cooling. Likely due to the requirement of using air as a thermal transfer medium (which is a poor choice at best).
We have high hopes for the Prisma hashing units primarily because they prove challenging/difficult to cool on air.

IMHO, the Prisma was specifically designed for a Data Center environment with all that brings to the table (Chilled water, adequate air flow, stable abundant power, dislocation from combustible materials, fire suppresion, etc. . .) and not for the home hobbyist that likely will operate these in a bedroom or garage.

But you know what they say about opinions, "Opinions are like assholes, everybody's got one and they all stink."
Maybe I'm just the stinkiest around.

In the previous tirade, I'm not trying to belittle, deride, or criticize your config, just trying to offer some observations/solutions that possibly might prevent a catastrophe.

Best of luck.

Sry was not on if your log shows BETO errors and no hash do this it works everytime please go to settings in minera go to where the CGam fork dropdown is choose rockminer fork save and restart miner then go back and reput onto CGam fork save and restart miner then click dashboard look at log all clear and you start mining you will get the BET0 errors anytime you change something with your pool however once failover is set it does failover and returns with no errors  

Can you enable debug logging from the settings page and restart cgminer? You should get some extra detail around why the boards aren't being properly initialized.

I tried the AM update via the command that CrazyGuy gave and no luck.

I've switched white cables, pi power supply, changed pi, pi usb, tested 1 board at a time with dip switches set for 'off' and nothing.  If I had a board issue, the miner should be able to run with alteast 1 of the 4 boards.  Pools appear to be working.  I even tried running with Phillipma1957 pool info.  So far nothing.

The last thing that I haven't tried is a different uart connector since I only have 1.

Thanks for the help everyone.  This miner is very painful to get going and from what I've seen with other people's catching fire and burning, a rather crappy miner at that.

at planetcrypto  what is the watts for that. >>>> if you get 1600gh and I get 1280gh  my way has to be a lot safer since the watts are low.


I am working on low clock right now I am at freq 200


great hash numbers over 1280gh and error numbers under  1%  >>>>>>>>>> I have 2 watt meters and they vary so much I suspect they are both wrong. one is too low and one is too high.

my lower meter read 1025 watts. I run extra fans so this unit is pulling about  1000 watts to do 1250 gh .  if accurate I am getting .8 watts per gh.

the high meter reads 1100  which is 1075 after extra fans  if acurate I am getting .86 watts a gh.

 it is about 40f outside I am going to open the garage door which will drop the temp from 104 to 80 I want to see if the cool air helps get better numbers.

Ambient Air Temp for the following pics:


Temp of back bracket on vertical Prisma:


Vertical Prisma Temp was taken off of:


Rear hole closed with Aluminum Foil Tape:


Hashing Data of OC'ed Prisma, Hash Rate ranges from 1400GH/s to 1800GH/s, Average is ~1600 GH/s:

Also look here at post #506,https://bitcointalksearch.org/topic/m.9489615 I sorta had the same problem, more user error on my part though. CrazyGuy helped me out, read to post 511. Also updating like CrazyGuy just said cant hurt.

Run this command to pull the latest and rebuild cgminer-AM

sudo /var/www/minera/./build_miner.sh cgminer-AM


ASICMiner checked in some updates to their fork about 5 hours ago surrounding Block erupter initialization. Worth a shot.

  my guess is that  is out of date the tubes came out in aug if that card is pre aug the software is wrong.

I'm using CG minera build on a sdcard that I've been using for months without issue on the r-pi.  I tried using the Rockminer, AM switch and reboot and get the same no hash.  I've tried it multiple times with different boards and no change at all.  

The usb adapter sees the miner fine.

The errors I'm getting appears that the BE chips have wrong addresses:

Without getting all esoteric, horizontal the back bracket was typically 176F @ stock clock (240) and unit was sporadically rebooting.
Mounted vertically as in the photo with the box fan below on high (~2000-2500 CFM) the back bracket is is a chilly 110F-112F.

There has always been a 50F delta T between the front chips and the back chips. And that Delta remains in this config.

Ambient air temp oscillates between 70F and 80F due to the powered roof vents cycling. And air is drawn in from the outside (4F-20F air) through a screen door.

I'm almost 100% sure that the temp reported on the dashboard is the temp near one of the vertical connectors on the Pi.

We used some beefy zip ties to hold it to the wire mesh.

The first time the fan cycled it scared the shit outta' me cuz' it got real quiet and I was sure I blew sumpin' up.
Yes they are louder than hair dryers and they are the loudest miner in the DC.

A little birdy whispered in my ear and told me that 70% of the heat dissipation in a Prisma is through the heat sinks the other 30% is out the component side of the board.

Be back l8r with pics.

edit----------the post above may be your answer.

okay try just using my pool info and none of yours.  

and what sd card image are you using?

if my pool alone does not work I have to think you have a bad sd card image and / or a bad usb adapter

back to my under clock tests.

this is freq 210 the garage down is open gear is cool gives me 1260-1280

 but power is lousy  1180 watts .925 watt per gh  .. I am beginning to not believe the meter

I had the same BET 0 errors and switching to cgminer (rockminer fork) under settings, then save and reboot, then switch back to cgminer (ASICminer fork) save and reboot seemed to fix it. Also people said just rebooting to the Asicminer fork several times cleared up the BET 0 errors. I just has this problem last night when I removed them from being daisy chained and plugged each prisma into its own port on the RPI B+ model using the USB to UART adapters. Hope this helps you

I enabled the debug log and I'm getting these errors:
 [2014-11-15 19:32:18] blockerupter: Read (0/1)
 [2014-11-15 19:32:18] BET0: Unexpected value 00 received
 [2014-11-15 19:32:18] BET0: BlockErupterRead (amt=0 err=-7 ern=25)
 [2014-11-15 19:32:18] blockerupter: Read (0/1)
 [2014-11-15 19:32:18] BET0: Unexpected value 00 received
 [2014-11-15 19:32:18] BET0: BlockErupterRead (amt=0 err=-7 ern=25)
 [2014-11-15 19:32:18] blockerupter: Read (0/1)
 [2014-11-15 19:32:18] BET0: Unexpected value 00 received
 [2014-11-15 19:32:18] BET0: BlockErupterRead (amt=0 err=-7 ern=25)
 [2014-11-15 19:32:18] blockerupter: Read (0/1)
 [2014-11-15 19:32:18] BET0: Unexpected value 00 received
 [2014-11-15 19:32:18] BET0: BlockErupterRead (amt=0 err=-7 ern=25)
 [2014-11-15 19:32:18] blockerupter: Read (0/1)
 [2014-11-15 19:32:18] BET0: Unexpected value 00 received


Interesting that my CS number of 'pool details' is increasing.

*EDIT*

After entering CG minera command from the other thread, I'm still getting errors:

[2014-11-15 19:45:56] BET0: Sent work 109 to board 0
 [2014-11-15 19:45:56] BET looking for and found BET 10c4:ea60
 [2014-11-15 19:45:56] BET looking for BET 10c4:ea60 but found 0424:ec00 instead
 [2014-11-15 19:45:56] BET looking for BET 10c4:ea60 but found 0424:9512 instead
 [2014-11-15 19:45:56] BET looking for BET 10c4:ea60 but found 1d6b:0002 instead

All right.  Testing now:

Pool shows alive with your info.

Changed r-pi to 1.1A power supply

changed to different white cable.

No hashing so far

* changed boards and no hashing on different board.