Topic: CoinTerra announces its first ASIC - Hash-Rate greater than 500 GH/s - page 64. (Read 231002 times)

it's also distinctly possible that for each chip they will tune them to run at different Ghz depending on how many of the hashing cores and dies are actually working - having a range of 1.1 Ghz to 1.4 Ghz (and potentially even up to 2 Ghz) gives the Cointerra engineers considerable latitude to have their firmware automatically tweak each chip to get to the target 2 Th for the entire unit.

Will

I guess you're not familiar with electromigration, then.

An engine has moving pistons, constantly moving against the cylinder walls..
I think if you can overclock anything, and its stable, its not going to wear out the same as a car..

Theoretically, a porsche GT3 can do the "cannonball run" in 14 hours and 45 minutes.   But it hasn't.   But it could.   I will be happy cruising along with my unit at 2TH/s.   Very happy.

Judging from what's said to have been ordered, I'm thinking the power throughput might be a problem there (for 3.84 TH/s). But that's not to say there won't be some nice overclocking headroom.

Thanks Will, you made me smile a bit  

Perhaps they have known this for some time, and that could be why they aren't lowering their prices on the may batch from $6k

Interesting update.  It seems they have gone from using just three of their dies to assuming they will use all four.

In the FPGA video from last year - http://www.youtube.com/watch?v=jSpNZMYunls  it was stated:

"120 cores per die for a total of 360 cores.  Nominal 500 Gh/sec up to 720 Gh/sec" and "1.4Ghz" (implied top speed for 720 Gh/sec of 2Ghz)

so that was 3 x 120 x 1400 = 504,000 Gh per chip, with four chips = 2.016 Th/sec which meets their specification, back last year.

Now, it seems they are running at only 1.1 Ghz:

120 x 1100 = 132 Gh/s on each die (as said on their post) x 4 = 528 Gh/s per chip, with four chips = 2.1 Th/sec

It's interesting to see their plans from using just three dies have changed to underclocking from their nominal 1.4Ghz to 1.1Ghz and using four dies.  Possible their yield from the first wafer was higher than they expected (they expected to only be able to use three dies), or it's possible that the power profile of running at a lower clock speed were better.

Either way, this is probably good news for cointerra customers, given the potential to clock each of the four cores at 1.4Ghz would give a projected yield of each TerraMiner of 1400 * 4 * 4 * 120 = 2.68Th - and if they are able to clock each die at the 2 Ghz they said in their FPGA video, the power systems can support it, and the yield on the dies is good, then a possible top speed of 2000 * 4 * 4 * 120 = 3.84 Th/sec

All this is speculation, of course.

Will

So by the sounds of it, the chips are capable of reaching their targeted hash rate, and now they're going to start testing the entire batch?  

Very exciting.   Even more exciting that an asic company is updating people in an effective manner.   Great Job CT and looking forward to more progress reports.

http://cointerra.com/engineering-update-asic-bringup-diagnostic-testing/

new update for Jan 7th...

No. Did I say you said it? No.
Please read carefully.


Sigh.

I never assumed 100s of PH wil be coming online in 6 months, please look up the definition of a hypothetical.
Moreover, the hypothetical didnt state that 100's of PH would be online in 6 months, merely that during the next 6 months we would see them being sold and being deployed. Note the present continuous.
Finally I am saying the hypothetical is at least plausible, and more plausible than the other hypothetical I gave of only a few dozen PH. Considering you agree at the current rate without anyone known to be shipping 28nm gear, we could already be on track to add 50PH its hard to see how you are arguing against the plausibility of that number multiplying considering about 10 vendors are expected to start shipping latest gen miners in the next months and currently there are none (known).

 How can ghash.io / Bitfury's recent turn be described as anything other than approaching behavior similar to a cartel ?

did i say 12 PH/s?


Exactly 100 PH per year, no problem with that. Nevertheless that somehow sounds alot different than your previous post:



Oh, yes i do think that the pace will pick up. Who doesnt? Nevertheless its still bullshit to assume "hundreds of PH/s coming online next 6 months".

Cartels wont work for bitcoin miners. Mining is a zero sum game, so asic vendors are competing with themselves even if there was no one else.  So even a monopolist can not continue to charge prices that will make their product appear unable to ever ROI. Every TH they add to the network reduces the market value of their unsold products.

Its a hypothetical, not a prediction. 12 PH seems even more absurd. FWIW we seem to be adding ~3-400 TH daily at this point, before anyone we know is shipping. If that were to continue you are already looking at 100PH per year. You dont think the pace will pick up once HF, CT, Bitmine, BFL etc will start shipping?

hundreds of PTH/s?

lol!

Wikipedia


Lets hope  

While competition is a factor, you cant talk about prices without talking about future difficulty. And future difficulty is a direct result of sales volume. That is the big unknown (aside from BTC price obviously). If all asic vendors combined only sold a few dozen PH, then Im sure there is room for prices to go up, at least for some time. If otoh, over the next 6 months we will see 100's of PH are being shipped and deployed, and BTC price remains roughly where it is now,  I dont see much of a market for $3/GH miners anymore.

Bitfury raised their prices as well for some mad reason. Probably because they knew they had XXX number of people with semi populated kits that need the extra blades.

Personally I don't like the Cointerra team.
I don't like their faces.
But if Cointerra proved they have working units, I will jump into it inclusive the hosting option.