HashFast announces specs for new ASIC: 400GH/s

[jjiimm_64]

I am still going to do my calcs on 700....  anything over that would be a bonus

Ignoring power costs, my projections at 700GH on Jan 1st, and 4x (2800 GH) on Feb. 3rd give a total lifetime loss of 17BTC.  Which is a much smaller loss we were looking at before.

With 800GH the loss narrow to 11BTC.

With my calculations we should get our btc back with 700 on Jan 1 and 2800 on March 1 by July-August (when network hash rate will be around 65Ph/s)

Hmm... I like your numbers better.
Are you using your own spreadsheet?

its NOT 2800, it is 3500 Gh

mpp is up to 4x more hashrate.. original max of 2000 (1 bj and 4 modules).  so your calcs should be 3500 not 2800

[r1senfa17h]

I am still going to do my calcs on 700....  anything over that would be a bonus

Ignoring power costs, my projections at 700GH on Jan 1st, and 4x (2800 GH) on Feb. 3rd give a total lifetime loss of 17BTC.  Which is a much smaller loss we were looking at before.

With 800GH the loss narrow to 11BTC.

Can you explain your calculations? I like this calculator, but only see it generating 16 btc if it hashes at 800gh/s with 25% increases per difficulty adjustment (hopefully diff is less!).

Edit: I forgot to add MPP. But I'm not convinced it will be anything more than 4 x the 400gh/s they advertised. They'll probably just ship 2 extra modules to everyone on March 1st, which increases my estimate to ~23.5 btc generated over its lifetime.

[JWU42]

I am still going to do my calcs on 700....  anything over that would be a bonus

Ignoring power costs, my projections at 700GH on Jan 1st, and 4x (2800 GH) on Feb. 3rd give a total lifetime loss of 17BTC.  Which is a much smaller loss we were looking at before.

With 800GH the loss narrow to 11BTC.

With my calculations we should get our btc back with 700 on Jan 1 and 2800 on March 1 by July-August (when network hash rate will be around 65Ph/s)

Hmm... I like your numbers better.
Are you using your own spreadsheet?

its NOT 2800, it is 3500 Gh

mpp is up to 4x more hashrate.. original max of 2000 (1 bj and 4 modules).  so your calcs should be 3500 not 2800

I guess you guys can't be bothered to actually read their website? 

What if a super-duper Exa-hash machine shows up?  

The MPP coverage extends to up to 400% of the capacity purchased. For a Baby Jet, with a nominal 400 Ghash/s capacity, the MPP is limited to sending an additional 1.6 Terrahash/s of additional ASIC hashing capacity, for a total of 2 Terahashes!

SOURCE: https://hashfast.com/miner-protection-program/

[jelin1984]

Final
Hashfast

Will delivered more than 400 gh per miner?
Or not?

[perezoso]

Oh, no.  With all the bullshittery and lying from Hashfast about October, if they don't produce 4 boards for the MPP, I'll go ballistic.

But here I present a simple recipe - one that I've been promoting for a while - for happiness, peace, and love between Hashfast and its customers:


Hashfast Happiness (serves each order for 1 Baby Jet under the MPP)

Ingredients

• 1 reliably working Baby Jet
• 4 "Golden Nonce" brand miniboards, also working
• Two tablespoons of improved marketing ethics
• A dash of contrite humility

First, deliver working Baby Jet to each destination immediately, add at least one dash of humility to improve palatability for hungry customers.

Second, unambiguously commit to delivering the four (4) working, fully-populated MPP "Golden Nonce" brand miniboards at the earliest possible opportunity, in any event no later than 1 March (but hopefully much earlier).

Third, self-administer two tablespoons of improved marketing ethics.  This means no more October false promise bullshittery, and may mean replacing Mr. Animal Porn in the marketing department.

Mix all together, perform in accordance with commitments, and enjoy the warm glow of happier customers and a restored reputation.

[bobsag3]

I have to admit that 800GH/chip, if stable, would be impressive to say at least. I wonder how the sierra is supposed to dissipiate 3kW of heat in its 4U. Not even recent and ultra high density datacenters can handle 30kW/rack.

Would probably need to just space the units in the rack and limit it to like 14KW/rack densities.

Not aware of many DC that can accommodate 14kW.

And mine.

[jspielberg]

Correct, I meant, 700 delivered in Jan 1 and 2800 delivered in March 1 for a total of 3500 in march 1. My calculations were for 3500 after March 1

I was using my own spreadsheet:
https://docs.google.com/spreadsheet/ccc?key=0Auya3iRE6az1dDc0UVgwMU52YVpTazVjSHByOGNiWHc&usp=drive_web#gid=8

I have been pretty impressed with the coinplorer mining simulator.
https://coinplorer.com/Hardware/Simulate

[jjiimm_64]

I am still going to do my calcs on 700....  anything over that would be a bonus

Ignoring power costs, my projections at 700GH on Jan 1st, and 4x (2800 GH) on Feb. 3rd give a total lifetime loss of 17BTC.  Which is a much smaller loss we were looking at before.

With 800GH the loss narrow to 11BTC.

With my calculations we should get our btc back with 700 on Jan 1 and 2800 on March 1 by July-August (when network hash rate will be around 65Ph/s)

Hmm... I like your numbers better.
Are you using your own spreadsheet?

its NOT 2800, it is 3500 Gh

mpp is up to 4x more hashrate.. original max of 2000 (1 bj and 4 modules).  so your calcs should be 3500 not 2800

I guess you guys can't be bothered to actually read their website? 

What if a super-duper Exa-hash machine shows up?  

The MPP coverage extends to up to 400% of the capacity purchased. For a Baby Jet, with a nominal 400 Ghash/s capacity, the MPP is limited to sending an additional 1.6 Terrahash/s of additional ASIC hashing capacity, for a total of 2 Terahashes!

SOURCE: https://hashfast.com/miner-protection-program/

then there is some ambiguity

400% additional to what purchased.. i purchased 400gh boards  1600/400 is 4more boards

they could go the scammy route and say here are 2 boards for the 1600..... BUT they aleady budgeted 4 boards for the mpp, so it does not cost them extra to send the 4 boards

[jspielberg]

My spreadsheet is obviously more complicated... as I am not basing it on anything other than the historical difficulty.  So far this exponential model has undershot reality for the last 6 months.

I am using wolfram alpha to get the components for a best fit exponential.

Here is the raw data in wolfram:
http://www.wolframalpha.com/input/?i=exponential+fit+%7B%7B13%2C+3651012%7D%2C%7B37%2C+4847647%7D%2C%7B156%2C26162876%7D%2C%7B200%2C65750060%7D%2C%7B222%2C112628548%7D%2C%7B243%2C189281249%7D%2C%7B308%2C908350862%7D%2C%7B319%2C1180923195%7D%7D

adjusted r^2 looks compelling to me at 0.999978

With this model and 700GH/s, I get ~5BTC for January.
And with 3500 GH/s Feb. 3rd forward I get an additional ~27 bringing the total to ~32 for the next 10 or so months

Obviously the super-exponential can't continue forever, so I put in a linearity cutoff saying that after that point hashpower will be added linearly to the network.  I optimistically chose 250M daily difficulty rise (which if the exponential continues will hit around March).

A bit complicated... but seems to modeling reality fairly closely (though the linearity threshold is admittedly a very air picked number which has a large impact on profitability)

[jspielberg]

The daily calculation is intentional. I do daily because you are still competing with a growing number of other participants.

I believe the effective difficulty is higher every day because there are an increasing number of competing hashes being added daily which dilute your hash power.  That happens even if their isn't a difficulty adjustment.

Even with the daily difficulty recalc --- this model has continually undershot reality... so I haven't bothered trying to make it paint a more rosy picture when it is already optimistic.

I like your spreadsheet, and I am using the same hack of using 3600 daily coins rather than using a proper hashrate/difficulty calculation, but I think using reported wafer orders is going to undershoot a lot.  I don't expect chip producers have much incentive to report the actual number of wafers they are having produced.  Easier to just model based on what the actual network reports.

[ImI]

The daily calculation is intentional. I do daily because you are still competing with a growing number of other participants.

I believe the effective difficulty is higher every day because there are an increasing number of competing hashes being added daily which dilute your hash power.  That happens even if their isn't a difficulty adjustment.

Wrong imo. Added hash power wont dillute your hash power. Added hash power will dilute bitcoin supply.

The one block every 10min rule is flexible. That means that for a short period of time there will be one block every 6 minutes for example and not every 10 min.

So your hashing power still does the same output. The only effect that one should consider is that the retarget interval isnt and wont be for some time the intended 14 days, as this 14 days happen only with the same hashpower output. That leads to approx. 10 days per retarget instead 14 days.

[joshv06]

What about for those of us that bought upgrades?

[DeathAndTaxes]

I have to admit that 800GH/chip, if stable, would be impressive to say at least. I wonder how the sierra is supposed to dissipiate 3kW of heat in its 4U. Not even recent and ultra high density datacenters can handle 30kW/rack.

The Sierra is sold as 1.2 TH/s unit not a 3 chip unit.   IF (and I haven't seen any convincing evidence) they are doing 600 GH/s+ they could drop the Sierra to two modules and just keep the extra hashing power for themselves.  Does anyone think that is unlikely given all the other ways they have screwed over customers so far?

[jspielberg]

I'll create a new spreadsheet based on 12 day difficulty retarget intervals, with the difficulty of the interval based on the exponential value at the time (from wolfram).

Should be pretty easy if you guys feel it will model reality better... but like I said... I anticipate it will still undershoot and end in tears.

[DeathAndTaxes]

The daily calculation is intentional. I do daily because you are still competing with a growing number of other participants.
I believe the effective difficulty is higher every day because there are an increasing number of competing hashes being added daily which dilute your hash power.  That happens even if their isn't a difficulty adjustment.

The last sentence is incorrect.  All that matters is your hashrate vs current difficulty.  If tonight the hashrate of the network doubled you would still earn the same amount tomorrow that you did today.  The rate you earn won't change until the difficulty adjusts.

[DeathAndTaxes]

I'll create a new spreadsheet based on 12 day difficulty retarget intervals, with the difficulty of the interval based on the exponential value at the time (from wolfram).

Should be pretty easy if you guys feel it will model reality better... but like I said... I anticipate it will still undershoot and end in tears.

The retarget period should be based on the % increase in hashpower.
If it undershoot then you are simply not predicting a high enough increase in hashrate each period. 

It makes more sense to raise your hashrate estimate than to keep using an obviously incorrect estimate and then hedge it by computing incorrectly off the incorrect estimate.  Right?
I mean it is like saying I know this model has two errors but they sort of cancel each other out so I am going with two known errors instead of fixing both of them.

[ImI]

The daily calculation is intentional. I do daily because you are still competing with a growing number of other participants.
I believe the effective difficulty is higher every day because there are an increasing number of competing hashes being added daily which dilute your hash power.  That happens even if their isn't a difficulty adjustment.

The last sentence is incorrect.  All that matters if your hashrate vs difficulty.  If tonight the hashrate of the network doubled you would still earn the same amount tomorrow that you did today.  The rate you earn won't change until the difficulty adjusts.

exactly only thing that actual changes is the bitcoin inflation

[perezoso]

What about for those of us that bought upgrades?

+1

[Minor Miner]

I have to admit that 800GH/chip, if stable, would be impressive to say at least. I wonder how the sierra is supposed to dissipiate 3kW of heat in its 4U. Not even recent and ultra high density datacenters can handle 30kW/rack.

The Sierra is sold as 1.2 TH/s unit not a 3 chip unit.   IF (and I haven't seen any convincing evidence) they are doing 600 GH/s+ they could drop the Sierra to two modules and just keep the extra hashing power for themselves.  Does anyone think that is unlikely given all the other ways they have screwed over customers so far?

I disagree.    Here is direct from the home page of their website

The Sierra is an astonishingly powerful bitcoin mining machine, powered by three (3) of HashFast’s Golden Nonce (GN) ASICs. The Golden Nonce is the most advanced mining chip produced to date. Each rig is capable of exceeding 1.2 Thash/s.

[DeathAndTaxes]

I am just saying this company has lied, mislead, backtracked, stealth changed TOS (in violation of law), falsely claimed customer prior to 15 AUG accepted a 31 DEC deadline, etc.

YOU REALLY THINK it is beyond the same company to drop it to two units and ship >1.2 TH/s?  Really?  I mean those customers would at least be getting more hashrate than they bought at roughly the timeline they were promised.  They still come out ahead of Batch 1 customers who at this point even a 800 GH/s rig + 400% of that in MPP delivered 1 Feb isn't enough to break even in BTC terms.