HashFast announces specs for new ASIC: 400GH/s

[ninjarobot]

I personally am planning on stripping my machines and putting them in cases of my own design.  If I could get my order as parts a week earlier (by not shipping Canada and then back to the US)... that would be just fine with me.

Exactly. The only thing that matters is the modules. The rest is just cowbell.

[1713911480]

1713911480

[1713911480]

1713911480

[1713911480]

1713911480

[pajak666]

http://www.youtube.com/watch?v=5rl6L8nt3Hc&feature=youtu.be

lol @ 0:27 that assembly machine just dropped capacitor:) well done!

[DeathAndTaxes]

Exactly. The only thing that matters is the modules. The rest is just cowbell.

Agreed.  Honestly I wish companies would just sell modules because so far none of them have had particularly good complete rig designs.

BFL case of fans?  We don't even need to go any further.
KNC uneven airflow and rats nest of cables because we decided to make the case too small to fit the PSU internally (I get why they didn't want to ship a PSU but come on make the case wider).
HF radiators on the front.
Cointerra lets get almost everything perfect and then have the power supply intake on the side which will be restricted airflow at best in a datacenter rack.

Now none of these are fatal flaws but honestly just give me the boards.  It cuts down on shipping, it cuts down on assembly, and let me build my own rig.

[-ck]

Anybody know if the HF miner firmware can be updated by the user after delivery?

Yes

[DeathAndTaxes]

I'm talking about the sierra. Yes, the chip level heat to be dissipated will be lower, but there will be around 1000W of heat anyway in there. Having the units in a real datacenter will help a lot, but anyway, putting the radiators where they are makes no sense at all.

Agreed it is dumb but honestly it isn't going to make that huge of a difference.  The chip itself (as opposed to entire system) isn't going to use more than 300W (or less) each so you are talking 900W spread out over a large area with significant airflow.  More airflow means lower temp rise (less thermal energy per unit of air).  The air will be warmer but it isn't going to be 20C burn your hand and melt the capacitors warmer.

[iCEBREAKER]

Ice, please tell me the sense of putting the radiators in the front of the sierra case, so that the whole board has to endure the hot output air of the radiators. I feel bad for the PCBs thinking that they will be exposed to a 1500W heater.

This is the level of proper designed cases HF has to offer.

David, I can't explain the thermal management features of the Sierra case to you, because you are incapable of understanding them.  Even after watching the airflow test video.   

No matter what i say, you will simply continue to whinge and moan and pick nits which do not exist. 

EG, you love it when Cointerra uses side vents ("a proper case") but hated it when HashFast announced the same thing:

Please tell me that there is a good reason to put the PSU air intake there, in a spot that once in a rack, no air can come trough.

And if such a good reason doesn't exist, i HAVE to conclude that it is only a rendering of a non existing case for something that will not ship anytime soon done and approved from people that doesn't have any idea of what they are doing.

In short David, you know nothing about hardware, mining, or Bitcoin.

You are a prematurely disgruntled high-maintenance customer and risked more than your poor mother can afford to lose.

PS  How's your lawsuit coming?  Have you got Johnnie Cochran, David Boies, F. Lee Bailey, Robert Shapiro, and Alan Dershowitz to file a class-action yet?   

[cedivad]

Don't worry Ice, the lawsuit is coming out great. However, before spending 20k$ for a retainer, i would just like to know how much HF screwed us.
But again, don't worry, i will keep you updated.

[Visesrion]

@HashFast: FiRST TESTS:  UP AND RUNNING! http://t.co/naqM73X1Wg #bitcoin #bitcoinmining


We’ve had an extraordinarily gratifying nine days of testing on the chip and motherboard. On Dec 1, the substrates finally became available and Golden Nonce became closer to a reality. Below is an update from our engineering team on how the chip has performed in tests. (In a word — excellent!)

“We chose a methodical and incremental approach, beginning with a free-standing environment with external power supplies and control logic. The device and substrate combination proved functional outside the motherboard. Low speed mining was performed to establish a baseline of current consumption vs frequency. A baseline of functionality (100%) in conjunction with cgminer was also established, giving us a place to stand should difficulties arise in any of the PCB-related bring up steps.

Next, the chip was placed on a blank motherboard with no other components. Power was provided using an external supply. Control was provided by external FPGA and microprocessor, and a liquid cooler was attached. This allowed testing of the device at about 1/2 speed, proving that motherboard power planes and substrate connectivity are good.

A motherboard with the complete 400 amp power supply but no ASIC was populated and tested under various loads. Power supply basic function was proven.

The last step was to bring it all together. We were pretty confident that it would spring to life — which it did! And from there the fun began. Clock speed was ramped, measurements taken, and finally we ended in full-speed operation of a single die, at temperature, with plenty of overclocking headroom left and an excellent looking power supply. Even the emergency thermal shutdown proved functional — saving the day when a water pump was accidentally left unpowered!”

The coming days should be equally rewarding as systems come together and we move on to all four dies. Look for more updates soon.

[MrTeal]

@HashFast: FiRST TESTS:  UP AND RUNNING! http://t.co/naqM73X1Wg #bitcoin #bitcoinmining


We’ve had an extraordinarily gratifying nine days of testing on the chip and motherboard. On Dec 1, the substrates finally became available and Golden Nonce became closer to a reality. Below is an update from our engineering team on how the chip has performed in tests. (In a word — excellent!)

“We chose a methodical and incremental approach, beginning with a free-standing environment with external power supplies and control logic. The device and substrate combination proved functional outside the motherboard. Low speed mining was performed to establish a baseline of current consumption vs frequency. A baseline of functionality (100%) in conjunction with cgminer was also established, giving us a place to stand should difficulties arise in any of the PCB-related bring up steps.

Next, the chip was placed on a blank motherboard with no other components. Power was provided using an external supply. Control was provided by external FPGA and microprocessor, and a liquid cooler was attached. This allowed testing of the device at about 1/2 speed, proving that motherboard power planes and substrate connectivity are good.

A motherboard with the complete 400 amp power supply but no ASIC was populated and tested under various loads. Power supply basic function was proven.

The last step was to bring it all together. We were pretty confident that it would spring to life — which it did! And from there the fun began. Clock speed was ramped, measurements taken, and finally we ended in full-speed operation of a single die, at temperature, with plenty of overclocking headroom left and an excellent looking power supply. Even the emergency thermal shutdown proved functional — saving the day when a water pump was accidentally left unpowered!”

The coming days should be equally rewarding as systems come together and we move on to all four dies. Look for more updates soon.

If you want any further insight into why D&T is saying that it was impossible for HashFast to have shipped their miners in October, you don't have to look any further than the fact that by their own admission they did basic power supply function and load testing in the start of December. If they were really anywhere near ready to ship back in October, hell even if they were ready but the silicon itself got delayed, that power supply and board should have been extremely well characterized in the month and a half since their first estimated ship window.
There's simply no reason why everything on that board shouldn't have been vetted long before the chips got mounted on the substrates, barring something unexpected in the ASIC forcing a respin.

[Syke]

A motherboard with the complete 400 amp power supply

That's gotta be one really big power supply! I wonder if all 400 amps are on a single +12V rail.

[defcon23]

sill testing .... then still dont shipped.

[DeathAndTaxes]

A motherboard with the complete 400 amp power supply

That's gotta be one really big power supply! I wonder if all 400 amps are on a single +12V rail.

400A @ <1 Volt.  Simulating the output of the voltage regulators.  The chip will probably be running at 0.7V to 0.9V so a 240W to 360W bench power supply.  Unlike an ATX power supply it will be adjustable.

[bob marley]

@HashFast: FiRST TESTS:  UP AND RUNNING! http://t.co/naqM73X1Wg #bitcoin #bitcoinmining


We’ve had an extraordinarily gratifying nine days of testing on the chip and motherboard. On Dec 1, the substrates finally became available and Golden Nonce became closer to a reality. Below is an update from our engineering team on how the chip has performed in tests. (In a word — excellent!)

“We chose a methodical and incremental approach, beginning with a free-standing environment with external power supplies and control logic. The device and substrate combination proved functional outside the motherboard. Low speed mining was performed to establish a baseline of current consumption vs frequency. A baseline of functionality (100%) in conjunction with cgminer was also established, giving us a place to stand should difficulties arise in any of the PCB-related bring up steps.

Next, the chip was placed on a blank motherboard with no other components. Power was provided using an external supply. Control was provided by external FPGA and microprocessor, and a liquid cooler was attached. This allowed testing of the device at about 1/2 speed, proving that motherboard power planes and substrate connectivity are good.

A motherboard with the complete 400 amp power supply but no ASIC was populated and tested under various loads. Power supply basic function was proven.

The last step was to bring it all together. We were pretty confident that it would spring to life — which it did! And from there the fun began. Clock speed was ramped, measurements taken, and finally we ended in full-speed operation of a single die, at temperature, with plenty of overclocking headroom left and an excellent looking power supply. Even the emergency thermal shutdown proved functional — saving the day when a water pump was accidentally left unpowered!”

The coming days should be equally rewarding as systems come together and we move on to all four dies. Look for more updates soon.

 
Hey HashFast don't ship your Golden Nonsense Baby miner......Just give us our BTC BACK

[aerobatic]

I was at the Inside Bitcoins conference today in Las Vegas...

yesterday the hashfast stand was looking a bit empty, with just the lovely Cara there to hold the fort and fend off the masses.. but today they actually had a Baby Jet to show off.

disclaimers.  it was not plugged in.  it wasn't hashing.  it was actually a bit damaged (a puddle of water can be clearly seen on the board from a leak of the water system).  They said it had been hastily packed and South West Airlines assisted with the rough treatment to cause the water cooling system leak.   Even if they HAD been able to power it up, the chips aren't yet hashing at full rate.  According to hashfast's twitter feed, they've just today got one die to hash (at 100+ GH/s), and have yet to try all four dies at the same time.  Its not so simple, as four dies will use four times the power, which will stress the entire board and all its components so they're taking it slow.

So.. these pictures are mainly to provide a context of what was shown..





The lovely Cara, showing her wares (yeah yeah, i know you're waiting for the punchline but I'm definitely not going to say  '... And the baby jet' ;-)


Which is mainly a large PC tower case, with a nice transparent window showing a single GN chip board...  complete with water cooling system.   The PC's chassis still had its drive bays intact.. these can probably be removed, as they're a tad superfluous.



You can see the puddle of water underneath the pump module...   definitely a good reason not to fire it up.. (and that the board isn't hashing at anywhere close to full speed yet so there'd be little point)



A closeup of the board...




The 'sales sheet' that hashfast was handing out to interested punters.

They also had a nice line in mugs and t-shirts... !

[DeathAndTaxes]

1 J/GH at the wall?  I really hope they can do better because after the delays and broken promises.  1 J/GH would be unimpressive.

Yeah I know they say +/- 20% but that means 0.8 J/GH to 1.2 J/GH and that is what KNC is already getting.

[Puppet]

Even if it was "hastily packaged", its quite worrying its leaking.

Im also a bit puzzled how KnC managed their chip bring up in basically one day, and HF is 10 days in their chip bring up and still hasnt even enabled all dies. "taking it slow" is not what its customers want to hear I think.

[SolarSilver]

Even if it was "hastily packaged", its quite worrying its leaking.

If they ship 550 systems in a week time with UPS, you'll see a two week window of posts with complaints about broken arrivals...

(unless they ship them without the liquid and the end user has to add the cooling liquid but then again, how can you test them at the factory?)

[aneutronic]

@HashFast: FiRST TESTS:  UP AND RUNNING! http://t.co/naqM73X1Wg #bitcoin #bitcoinmining


We’ve had an extraordinarily gratifying nine days of testing on the chip and motherboard. On Dec 1, the substrates finally became available and Golden Nonce became closer to a reality. Below is an update from our engineering team on how the chip has performed in tests. (In a word — excellent!)

“We chose a methodical and incremental approach, beginning with a free-standing environment with external power supplies and control logic. The device and substrate combination proved functional outside the motherboard. Low speed mining was performed to establish a baseline of current consumption vs frequency. A baseline of functionality (100%) in conjunction with cgminer was also established, giving us a place to stand should difficulties arise in any of the PCB-related bring up steps.

Next, the chip was placed on a blank motherboard with no other components. Power was provided using an external supply. Control was provided by external FPGA and microprocessor, and a liquid cooler was attached. This allowed testing of the device at about 1/2 speed, proving that motherboard power planes and substrate connectivity are good.

A motherboard with the complete 400 amp power supply but no ASIC was populated and tested under various loads. Power supply basic function was proven.

The last step was to bring it all together. We were pretty confident that it would spring to life — which it did! And from there the fun began. Clock speed was ramped, measurements taken, and finally we ended in full-speed operation of a single die, at temperature, with plenty of overclocking headroom left and an excellent looking power supply. Even the emergency thermal shutdown proved functional — saving the day when a water pump was accidentally left unpowered!”

The coming days should be equally rewarding as systems come together and we move on to all four dies. Look for more updates soon.

If you want any further insight into why D&T is saying that it was impossible for HashFast to have shipped their miners in October, you don't have to look any further than the fact that by their own admission they did basic power supply function and load testing in the start of December. If they were really anywhere near ready to ship back in October, hell even if they were ready but the silicon itself got delayed, that power supply and board should have been extremely well characterized in the month and a half since their first estimated ship window.
There's simply no reason why everything on that board shouldn't have been vetted long before the chips got mounted on the substrates, barring something unexpected in the ASIC forcing a respin.

THIS^

I'm also troubled by the lack of any wattage readings, no news on this is probably not good news.

Has anyone seen the actual power consumption figures from the tests yet?

[stan258]

Looks like I bought a box that is stupidly shaped and wont fit on anything called normal . I paid in cash so I was willing to keep quiet until now. Like the last damn straw. big brick of poo. Looks like a larger version of bfl jalapeno.... Great I dont need a bigger paperweight.

[ImI]

Great I dont need a bigger paperweight.

Then why don you just give it away as a present?

I am sure you will find somebody here in the forum who is willing to accept your babyjet-paperweight.