Topic: Hardware for BitCoin (Read 825 times)

Doesn't surprise me as it is a good way to push the hardware to be the best it can be instead of locking operating parameters at what works for all. Thass why OC'ing used to be so popular with Ants and others- quite often chips could be ran faster than factory spec. My point was that the 16/14 node made it a requirement so Bitmain and others can use more of the chips they get vs locking specs low enough to accommodate laggards.

Auto tune PREdates the 14/16nm miner generation.

 Spondoolies SP20 (28nm) and probably most if not all of the other Spondoolies models.

Ja. As I've often said and pointed out in the A7 thread (about the Premium market): Having had one open to replace the fan I can say Canaan did things very well in their design. Their version of auto-tune is just another example of that.

And before you ask.... Autotune.....
Thank moving to 16/14nm chips for that. In short the production processes and chip yields are still far from stable. Even now going on 2 years after the GPU and mining chips were released for sale to the public the chips have performance specs all over the map not only batch-to-batch but also highly variable between dies from the same wafer. As a result, some run fast, some run slow, some like high Vcore, some like it low...

Started in Feb 2014 with 2 little 10GHs Jala's from Butterfly Labs. Never ever thought that in less than 4 years that would grow to 240+THs.....

What I do: Research, design and build industrial lasers for the past 40+ years.... Everything starting from the power on the wall to putting beam on target. That is where my knowledge of current chip tech comes from: we make some of the very critical machines that make chips....

A prime example, about 9 years ago the LED lights you now see everywhere started to become much much cheaper - and brighter. One key piece of the puzzle leading to that was missing and in 2006 I was in Taiwan installing some equipment when our customer there called me into a management meeting with Phillips-LumiLEDS They wanted to know if....

It so happens that in 2005-2006 a few new laser and optics technologies were gaining traction and I realized how they could be combined to work together: My answer was well, yes I know how it can be done. As a result of that chance meeting and the very deep pockets of our customer in Taiwan, this is what I came up with that for one, changed the face of lighting forever: Press Release Even now almost 10 years since we delivered that 1st system no one sells systems anywhere close to the specs ours do 24x7x365. They simply have not figured out the minutia in the voodoo I do so well

That very same process I came up with is used to make bleeding-edge nterposers for CPU's and other chips. In essence an interposer supports the die, provides an excellent thermal path for it and fans out the microscopic connections between the die to something larger and more spread out for connecting to the final chip package pins and the outside world.

Damnit... the Avalons are still the obvious choice for me, then...

I appreciate that you're always willing to take the time to explain things so well. How long have you been doing this and am I correct in assuming your professional career was in a related field?

Continuing on about Auto-tune: The Avalons use it as well however.....
a.Bitmains version only runs when the miner is booted. After that those are the settings it runs at. Period. Considering I often go well over 3 month between reboots well, things and conditions WILL change...

b. Bitmain gives the end used zero input to how they want the miner to perform.

Canaan on the other hand:
A. From what little Canaan has said about it, their auto-tune continuously runs and constantly tweaks Vcore and chip speeds. Awesome miner shows those speed tweaks and monitoring power in shows dips/peaks matching in time what Awesome shows. They make mention of it in PR about their 'True Hash Rate' or whatever the acro is. Point is, the hash rate they advertise is the MINIMUM it will produce at default voltage setting of -0 and a specified air inlet temp. So far all of mine run a bit faster than advertised

B. The Avalons also allow a modicum of power/speed control by giving us 4 choices of base line core voltage to be used. Because of how their constant auto-tune works the end result is that using a setting of +1 increases the chip core voltage allowing the chip to run at a higher speed meaning faster hash rate at the expense of drawing more power and running hotter.

Using a setting of -2 drops core voltage and since that voltage directly controls how fast the chips will operate before generating too many errors, auto-tune now lowers the speed to keep the chips happy (and cooler/less power).

The Avalons still allow users to tweak the voltage, correct? I'm expecting I'll probably need to; especially during the summer months, depending on how well I get the ventilation set up.

After maybe the s3 over clocking became a Not Good Thing to do and starting with the s9 batch 16 Bitmain started using Auto-tune and took away all user control of frequency. They took away voltage control with the s7's. So yes, all stock settings.

The problem with over clocking is that not only do the miners get even more bloody hot but more to the point, power vs hash rate efficiency starts to drop like a stone. The crappy thing about Bitmains Auto-tune is that we cannot under clock anymore to lower operating temps (and power bill). I turned my earlier ones down about 750GHs dropping the hottest chips from 106C and pulling 1,375W to being around 98-99C and 1,250w.

Do you attempt to overclock at all? Is it really that hit or miss on stock settings?

After my 31 August order from Canaan (still no btc address) and the 741 group buy that seems to have stalled out due to their payment requirements, I'm attempted to change to Bitmain and run the S9s instead.

With Bitmain is is literally a roll of the dice on reliability. Heavily loaded in our favor perhaps but nonetheless the S9 does not give me a fuzzy-warm feeling like even the s7 did. Out of 11 of them starting from batch-1 on up I've had a total of 3 board failures so far.

Not bad but out of the 20 s7's I used to run -- 1 DOA board and 0 failed boards once running. Going way back to their s1, s2, s3, s4, and s5's -- I had zero failures.

Their R4.... Have 3. 2 of them are perfect, 1 had board fail after 25hrs. Reports of failure abound here.

Have 4x T9's - running perfect and very few reports of problems.

I had no clue that failure rate is huge with some batches. In that case, you recommend buying a few from each batch rather than buying a lot from just one batch? There are so many moving parts to this....

Joke: Yes I'm just as powerful as the S9.   Just gimme paper and a pen, and I'll start hashing.

 

Okay, if we talk seriously:
S9's are the most efficient but not easily available and failure rate is huge for some batches.
AvalonMiners are not the best with power efficiency but are rock solid. Second best availability at the moment.
Ebit E9 Plus are in the same category with power efficiency as Avalonminers. There is not much data for Ebit miner quality. Best availability at the moment.

Are these Ebit's, Canaan's, Ebang, Hagss etc better or even comparable to the Bitmain S9's?

And there is Ebang with their Ebit E9 Plus miner.
http://miner.ebang.com.cn/

For Bitcoin the choices are limited: Bitmain and Canaan are pretty much the only game in town unless you got mega$$ in which case a Bitfury container might be applicable.

-oh right, good call Hagss, Ebang too.

Hi.

What kind of different hardware is availble these days for mining Bitcoin? It could be good if you just could post the hardware that is used. Don't mind if it is profitable. I see alot of diffent hardware, but I don't know if it is for Bitcoins or AltCoins.

Br
Valnurat