Bitcoin Forum

Is this http://peernova.com/product/petaone-rack/ (http://peernova.com/product/petaone-rack/) the wet dream of every miner - or just some huge PR-buzz which will turn out as hot air?

Same team as highbitcoin.com - I'd say google around for that - make up your own mind.

a rack is usually 42U tall (and are up to 48U tall) and if they're claiming 50 TH per rack, then thats slightly more than 1 TH per U.  this is for silicon that they may not have taped out yet (its not been announced) and thus may not available til later this year, presumably a few months from now, even if they were to tapeout today.   highbitcoin/peernova has yet to make their first silicon and many companies who received first silicon didn't get them working at their anticipated spec so there are risks of course that the silicon isn't always 'right first time', and even if it 'works', it may be grossly higher power than estimated (most asic companies have suffered this problem as bitcoin mining asics are notoriously difficult to accurately simulate with existing power estimation tools)

Spondoolies next generation SP30 product has already been pre-announced and pre-orders taken for it... and its approx 6 TH in a 2U rack device, thus it would comfortably beat the highbitcoin density spec.   their sp30 product is about 4x their currently shipping product (sp10).  i mention the 4x number, because it feels like about the right stepping from one generation to the next and is probably a theme that we will see from other bitcoin asic providers.

kncminer's current product is 4U and in no way gets close to the density (700 gh) so its hardly worth mentioning, but their next gen product - the neptune, was announced to be at least 3 TH/s in a 4U so its getting close to the required density, and that should be available by June (they say), and kncminer has a habit of over-delivering so its likely their neptune will be a 4 th/s miner and that they might hit the density of 1 TH/s per U that this would require.

cointerra's current product is also a 4U and is 1.6 th/s (or 2.0 th/s next month with the improved boards)...  which isn't as high a density (but only half isn't that far away), and presumably cointerra's next product will probably also be in the ballpark.  if their next gen one is more than 2x the density of their current one then they'll be there, and it seems clear that no company is just going to aim for a 2x improvement from one generation to the next so expecting more than 2x is likely.

all of this talk of density is relatively academic as datacenters don't charge too much for increased rack space use thus density isn't so important, but instead, charge proportionally for how many KW's are used, thus fitting in less rack space doesn't save much cost but being lower power, does save a lot of running expenses.

Have you been hiding in a cave aerobatic? "KNC has a habit of overdelivering." WHA? That statement couldn't be further from the truth!

Even at .1 watts GH if they can reach Spondoolies Gen 3 batches,  it would need 5000 watts to operate.  Where would people plug that into their house.  50amp 115v, or 30amp 240v circuit.

 yeah but they say it is .5 watt per gh  so that rack is 50th or   500 watts a th at .5 watts  times 50 = 25000 watts

 You could put it on the market July 1st and it is a money loser.  never mind that it burns 25,000 watts.  how about cooling it.

it gives off 125,000 btu of heat

 you need a 10  ton 100000 btu unit to cool it. along with serious fan work  that is 6000 more watts.

http://peernova.com/product/petaone-rack/

PetaOne™ Rack Features

Performance

Each PetaOne Rack delivers 50TH/s
Power efficient design consumes an industry low 0.5W per GH/s
Mechanical

Standard 42U, with 24” wide open frame rack
Incorporates up to 36 standard 1U blades (19” wide)
Cooling Requirements

Conventional air-cooling, with hot aisle enclosure preferred
Input Air temp: 10-30 °C, Output Air Temp: < 55°C
Software

Remote management, diagnosis and troubleshooting
Supports mining for Alt coins, easy upgradeability
We are accepting pre-orders for early July delivery!

sounds like the dryer plug might have other uses  :D

my cave is very comfortable, thanks ;-)

sorry i know of late they haven't been living up to their previously set reputation...  and that the frankenjups were a bad idea from the start... but i was still expecting the neptunes to be more than 3 TH when they're eventually delivered.  or do you think that assumption is no longer valid?

-- Jez

The single blade looks interesting, and while the sales monkey would love me to order 5 X 208V/30A whips, I think I would rather run a few and expand. I am reserved due to the relationship with cloudpounding, likewise the propaganda is not that impressive with respect to the "staff". Is cloudpounding already running this ASIC? or are they going to magically pull a 2TH/s 1U blade out of a hat come July?

cloudhashing runs mostly cointerra and bitfury boxes.  the highbitcoin system doesn't exist yet, and won't until july at the earliest..  who knows if they've even taped out yet and if they have, july is the best case timescale at this point.. and could easily be august/sept if the chip works first time, and november if it doesn't.

If someone can post me a test rack Ill try it out on the 3 phase for free.

Awesome!  Higher density so you can accrue greater mining losses in the same amount of space.  ;D

Well, if they deliver in Q4, it gives me enough time to let my hair grow and brawl with the most expensive hair dryer in my hometown!

 :D You win the internet for the day!!!!!  :D

I wish there was a "Like" button  ::)

How is this "design" anything new? You can get 50TH/rack now, filling a rack with sp10s.

Come July, you can even get a 115 TH rack with sp30s.

Spondoolies has the added benefit that you dont need to buy a minimum of 1 PH from them.....

I doubt it is for real, more just a wet dream :)

All this will ASICMiner delivers 2PH @ $900/TH in 6 immersion cooled DataTank flat racks that can be reused for their next chip (at a cooling overhead of below 1%). Silent and dust free, no cold/hot aisle required.

Wet dreams indeed  ;D

A "wet-dream-dryer", though.

Actual price might be higher as the immersion cooling tanks also have to be paid for, but yes, in 1~2 months Asicminer will deploy their new systems.

Either way there will be hot air ;)

I have  to agree that it will be a hot air machine.  The .5w a gh  is concerning  the sp30's claim .46w?


   The efficiency numbers are closing in on not much improvement.

 My old school 2012 GPU need 300 watts a gh/ 
Never had a FGPA
But my first asic need an AM usb stick needed 9-10 watts a gh
My next a red  fury needed 2 or 3 watts a gh.
My downclocked undervolted s-1 gets 1.4 watts a gh
My hosted Dragon 1th gets .9 watts a gh

The best running miner now a sp10 gets .7 watts a gh according to your tests on the low power mode.

The nex gen to come a sp30  may get .46 watts.

 we are going flat on this curve

We will remember the days when we could mine at home...

~$900/TH is for a total 2PH ASICMiner system in an immersion cooled container delivered to your doorstep. All you need is connect power and internet to the container and that's it.

2000 * ~$900 = ~$1.8M USD

It's a steal ;D

Your calculations seem off, the container alone is at least 0,6$ per watt, then you have to account for pcb cost, PSU cost, while chip cost alone is atleast another 0,5$ per GH.
1,3-1,5$ per GH seems a more reasonable number.

AM Gen3 chips aren´t more power efficient than 0,7W/GH.

Also keep in mind that Allied Control won´t sell you a container unless you can come up with the around 0,5-0,8 million dollars for the container alone.

My numbers are not off, they come straight from Allied Control. Ask them for a quote of a fully deployed AM BE200 container.

And I assumed you have tested the chips, otherwise you couldn't claim they are not more power efficient than 0,7W/G.

So did anybody contact them for  quote?

I though these guys make the GS1 chip for Cointerra?

wow.. hows that for crossed lines!?

er, no, cloudhashing (now renamed peernova) is a customer of cointerra.  cointerra designed and produced the chips and manufactured the systems containing the chips, and sold them to cloudhashing (amongst others).  cloudhashing then re-sells the gigahashes with a margin to end users.

https://cointerra.com/cointerra-open-silicon-announce-tape-goldstrike1-asic/

Open Silicon = HighBitcoin = PeerNova, no?

So they seem to be making the GS1 chip for Cointerra, but now competing with Cointerra as well.

Same for Cloudhashing if they are indeed PeerNova. This is confusing.

Next thing we learn is that actually the Indian government is behind all this, while the Chinese government is behind the all the A* miners ;)

Goodness me, haven't they got an awful lot of people at the top? 10 VP's/ CE somethings, 3 Directors and 1 Advisor

Wonder what they all do. Between them they must cost that company over $3m a year in salaries, bonuses and benefits. It'll take an awful lot of their machines to pay for them let alone the cost of the actual hardware.

Each company does their own design for their chip.  The RTL design, as its called.

When they've finished with the RTL, they hire a 'back end' designer to do the physical design, which lays out the chip ready for production.

Open Silicon is a contractor.  they do the physical design and are the gateway to the fab.  they act like the fab's agent.

Every asic designer needs to go through a physical design partner to have their chip made.

KnCMiner went through alchip, hashfast went through uniquify, and cointerra went through open silicon.  And now, so did highbitcoin.

this is further confused, because the 'chairman emeritus' (whatever that is) of open silicon is naveed sherwani, who was also originally an advisor to cointerra, and is now president of highbitcoin, so yes, there is some contamination it seems.

this is also a bit weirder... as everyone knows cointerra had an unexpected 20% slower performance and 20% higher power consumption than they originally estimated, and this 'under performance' was introduced at the last minute during the physical design process by open silicon, thus if anything, open silicon was the weak link in cointerra's design process, and now highbitcoin will be using them for their physical design, so we'll see if open silicon makes the same contribution, and we'll see if highbitcoin's chip ends up slower and more power hungry than they expect, as well.

Exactly  :'(  Too much power @ too high a cost per kwh means....no contest....Datacenter Corp mining wins  :'(

They'll just start a mining farm...like the other big boy's  ::)

i dont object to the idea that my mining has to be in a data centre.  it already is, for me.  i own my own hardware but i choose to host it in a professional place that they can keep cool, and give tlc to my machines when they need it (firmware updates, reboots, whatever it takes).

its clear that as the power requirements of these bitcoin miners in total gets higher (as its a never ending growth curve) then its also clear that these won't be in the home very long.  but that doesn't mean it becomes decentralised.   there's still a huge number of asic providers, and a huge number of hosting data centres... so all that has changed is that the mining is done in the cloud.. but it can still be owned, and controlled in a decentralised way

I assume you're not suggesting that open silicon may have 'introduced' more power consumption than needed, so it's probably back to the age old problem of simulation tools being unable to cope with the peculiarities of SHA256.

definitely not saying they deliberately did something bad.  just saying that they were responsible for the back end layout, and they could've done a better job.

actually, you're right... the sha256 designs have been notoriously difficult to simulate because the power estimation tools assume a toggle rate that doesn't match sha256 designs.  almost every single bitcoin mining chip has used more power than they first estimated.  sometimes wildly out like bfl's (10x?).. sometimes just 20% more, like cointerra's & hash fast's (both were assumed to be 0.6 w/gh and both ended up higher)

Several companies also vastly overestimated the DC/DC conversion efficiency, which further compounds chip inefficiencies.

Indeed. It's one thing dealing with 10 amps, but when you're up to 120 then the board layout, Mosfet on resistance and track sizes/thicknesses have to be looked at very carefully and most psu designers simply aren't used to dealing with such unusual parameters.

ha ha ha ha ha ha ha .....  are you fukin kidding me... more pre-orders..

STOP  pre-ordering from these companies miners...  remember bfl... hashlast... cointerra....   


STOP PREORDERING

  in their own advertisment they claim .5watts  vs the sp10's known .7 watts and the sp30's hopeful .46watts.

Asic's are approaching max efficiency   .3watt  just  as a guess.

We need a new way to hash.   a new chip not of asic design.    we went from 300watts a hash to the current .7 watts.

 that is better then 400 to 1.   time for a new game. 



I am sure BFL will come out with the god-like miracle chip   The JOSHER tech.

 the chip will have 1th of hash at 10 watts and use WTF tech to find solve the algorithm.

It will be only 500usd and available this DEC  please place your preorders now.

Dam, soon 50 TH/s will be the new 5 GH/s.   :(

Inevitable I know, but hate where this is all going.

Except you can't power it from a USB port.