HashFast announces specs for new ASIC: 400GH/s

[iCEBREAKER]

HF was the first full custom 28nm ASIC.  KnC used a shortcut.

That's why we ordered from them, if you can remember that far back.   

Hmm full custom you say? That's impossible! HF chip can't be a full custom chip. If it is the it must be the worst implementation ever!

Ok RS, if "full custom" is the wrong term please tell us the right one, because you know what I meant.   

[1714120516]

1714120516

[1714120516]

1714120516

[1714120516]

1714120516

[1714120516]

1714120516

[RoadStress]

HF was the first full custom 28nm ASIC.  KnC used a shortcut.

That's why we ordered from them, if you can remember that far back.   

Hmm full custom you say? That's impossible! HF chip can't be a full custom chip. If it is the it must be the worst implementation ever!

Ok RS, if "full custom" is the wrong term please tell us the right one, because you know what I meant.   

There are different type of making a chip. Most common ones are "full custom" and "standard cell". Bitfury has a full custom 55nm chip, all the rest are standard cell. Google is your friend my friend. Search and see the differences. HF chip must be a standard cell one, NOT full custom!

[iCEBREAKER]

There are different type of making a chip. Most common ones are "full custom" and "standard cell". Bitfury has a full custom 55nm chip, all the rest are standard cell. Google is your friend my friend. Search and see the differences. HF chip must be a standard cell one, NOT full custom!

OIC, you are just trolling and trying to make me think I'd forgotten my ASIC vocabulary.

KnC's first chip was standard cell.  HF's GN1 is full custom.  That's why it is so much faster than the competitions!

750 GH on a single chip.  Still an unbeaten record.   

[Inaba]

There are different type of making a chip. Most common ones are "full custom" and "standard cell". Bitfury has a full custom 55nm chip, all the rest are standard cell. Google is your friend my friend. Search and see the differences. HF chip must be a standard cell one, NOT full custom!

OIC, you are just trolling and trying to make me think I'd forgotten my ASIC vocabulary.

KnC's first chip was standard cell.  HF's GN1 is full custom.  That's why it is so much faster than the competitions!

750 GH on a single chip.  Still an unbeaten record.   

It's actually 4 chips on a single substrate.  So 750/4 is the actual speed of the chip, or 187.5 GH/s per chip. 

[RoadStress]

OIC, you are just trolling and trying to make me think I'd forgotten my ASIC vocabulary.

KnC's first chip was standard cell.  HF's GN1 is full custom.  That's why it is so much faster than the competitions!

750 GH on a single chip.  Still an unbeaten record.   

"Fastest" chip from the fastest mining company bankruptcy in the Bitcoin ecosystem.

How many bitcoins are you willing to bet on GN1 being full custom?

[iCEBREAKER]

"Fastest" chip from the fastest mining company bankruptcy in the Bitcoin ecosystem.

How many bitcoins are you willing to bet on GN1 being full custom?

Why put "fastest" in scare quotes?

Are you too cranky to acknowledge the simple fact that HF's 750GH/s/chip (whether full or partial custom) is still the world record holder by a large margin?

What does bankruptcy have to do with performance specs?  Did the GN1 suddenly get slower when HF ran out of money and time?   

You appear to be letting your butthurt emotions color your view of objective facts of hardware output.

Do you need a crying towel or pouting handkerchief?   


PS  Inaba, it's one chip with four dies.  Chip = [die(s) + substrate package].

[Syke]

What does bankruptcy have to do with performance specs?  Did the GN1 suddenly get slower when HF ran out of money and time?   

It became irrelevant when they couldn't build a working miner out of it.

[jimmothy]

Are you too cranky to acknowledge the simple fact that HF's 750GH/s/chip (whether full or partial custom) is still the world record holder by a large margin?

PS  Inaba, it's one chip with four dies.  Chip = [die(s) + substrate package].

Wrong once again.

[iCEBREAKER]

Are you too cranky to acknowledge the simple fact that HF's 750GH/s/chip (whether full or partial custom) is still the world record holder by a large margin?

PS  Inaba, it's one chip with four dies.  Chip = [die(s) + substrate package].

Wrong once again.

If I'm wrong, please tell the class what competing die(s)+substrate object offers higher performance than HF's.  Come on, don't be shy.  You can do it!  Oh wait, no you can't.   

Both definitions are correct, but in common usage the complete die(s)+substrate package constitutes a chip.

Since you want to be a pedant, let's just agree that HF's GH per sq mm of wafer is the highest of any design, regardless of how obnoxiously narrowly you want to define "chip."   

[Gleb Gamow]

Are you too cranky to acknowledge the simple fact that HF's 750GH/s/chip (whether full or partial custom) is still the world record holder by a large margin?

PS  Inaba, it's one chip with four dies.  Chip = [die(s) + substrate package].

Wrong once again.

If I'm wrong, please tell the class what competing die(s)+substrate object offers higher performance than HF's.  Come on, don't be shy.  You can do it!  Oh wait, no you can't.  

Both definitions are correct, but in common usage the complete die(s)+substrate package constitutes a chip.

Since you want to be a pedant, let's just agree that HF's GH per sq mm of wafer is the highest of any design, regardless of how obnoxiously narrowly you want to define "chip."  

How does an entity that designed and built the best mousetrap known to man go bankrupt, while inferior mousetrap manufacturers continue to exist, supplying their customers with highly sough-after sub-par products?

[bigasic]

I think this is what is causing bitcoin to stagnate.. too many fucking scams. here we have hashfast, I could sit here all day listing scams, etc. But hashfast really had some potential. I was very interested in them at the beginning.. Im so lucky that BFL already stole all my money so that I couldn't invest, lol...

[iCEBREAKER]

How does an entity that designed and built the best mousetrap known to man go bankrupt, while inferior mousetrap manufacturers continue to exist, supplying their customers with highly sough-after sub-par products?

You act as if HF is the first company in history to go bankrupt despite having a superior product, but you are far too old for that to be the case.   

Did you miss the last 12 months of this thread, where the story was told blow-by-blow and in detail?

You know the answers to your question, but pretend you don't just to create negative insinuations (and drama queen, as usual).

Do you really think anyone is going to buy your fake pearl-clutching pantomime?   

[SolarSilver]

You act as if HF is the first company in history to go bankrupt despite having a superior product, but you are far too old for that to be the case.   

If you consider the HF produced equipment to be a 'superior product' then that tells a lot about the low low standards you live by...

Only one out of 4 units that arrived here performed as spec'd out. The rest had to run crippled for an entire month until there was finally a software fix that disabled one of the cores, making the systems now to run at 3/4 speeds. That is surely a superior yield...

Oh did I forget to mention the superior product arrived over 2 months late, rendering it totally useless?

Why do you keep defending HF like you are Hashfast_CL's retarded little sister? "We got the fastest chip". Who cares. Bottom line  matters: HF did not deliver a working solution to their customers on time. And it turns out that most of it is to blame to incompetence too.

[jimmothy]

You act as if HF is the first company in history to go bankrupt despite having a superior product, but you are far too old for that to be the case.   

Why do you keep defending HF like you are Hashfast_CL's retarded little sister? "We got the fastest chip". Who cares. Bottom line  matters: HF did not deliver a working solution to their customers on time. And it turns out that most of it is to blame to incompetence too.

I think I remember a rumor that Icebreaker actually was Hashfast_CL. It really does make sense now. Nobody other than Icebreaker could completely destroy hashfast reputation and anger customers like Hashfast_CL did.

[jimmothy]

How does an entity that designed and built the best mousetrap known to man go bankrupt, while inferior mousetrap manufacturers continue to exist, supplying their customers with highly sough-after sub-par products?

You act as if HF is the first company in history to go bankrupt despite having a superior product, but you are far too old for that to be the case.  

Yes, HF is the first bitcoin manufacturer in history to spend ~$15m on NRE and design a chip that is only on par with the competition.

BTW where is the document that outlines Hashfasts expenses? I've been looking through the documents here but can't seem to find it: https://cases.processgeneral.com/cases/document/case/5/hashfast-technologies-llc/

Both definitions are correct, but in common usage the complete die(s)+substrate package constitutes a chip.

No both definitions are not correct. A substrate is just the board part of any PCB (other part is components/copper traces). There are no ASICs that do not go on a substrate.

If I'm wrong, please tell the class what competing die(s)+substrate object offers higher performance than HF's.  Come on, don't be shy.  You can do it!  Oh wait, no you can't.  

I'm assuming you mean in terms of gh/s per substrate so here's your useless comparison:

Hashfail - 750 gh/s
Cointerra - 800 gh/s
SP10- 800 gh/s
SP30- 2,250 gh/s
BFL monarch - 800 gh/s

Since you want to be a pedant, let's just agree that HF's GH per sq mm of wafer is the highest of any design, regardless of how obnoxiously narrowly you want to define "chip."  

If anyone is being obnoxious it's you. Nobody cares about how much gh/s you can fit on a chip/pcb. What's important is the $/gh and w/gh for complete hardware not just the chips.

Yes the golden nonce is very good in terms of $/gh (~$0.1/gh I'd guess) but the w/gh is horrible and the amount of money they spent on engineering completely negates the production cost savings.

[iCEBREAKER]

You act as if HF is the first company in history to go bankrupt despite having a superior product, but you are far too old for that to be the case.   

If you consider the HF produced equipment to be a 'superior product' then that tells a lot about the low low standards you live by...

Only one out of 4 units that arrived here performed as spec'd out. The rest had to run crippled for an entire month until there was finally a software fix that disabled one of the cores, making the systems now to run at 3/4 speeds. That is surely a superior yield...

Oh did I forget to mention the superior product arrived over 2 months late, rendering it totally useless?

Why do you keep defending HF like you are Hashfast_CL's retarded little sister? "We got the fastest chip". Who cares. Bottom line  matters: HF did not deliver a working solution to their customers on time. And it turns out that most of it is to blame to incompetence too.

I care about the fact that HF has the fastest chip because how much that IP fetches at auction directly affects how much we customers will be compensated for their delays.

You don't need to rehash well known history regarding late shipments.  We were all here for that and some of us don't like to wallow in self-pity, as is your preference.

I'm not "defending" anything but the fact that per mm sq of wafer, HF's chip is the fastest.  The chip is the best, although the first boards and early firmware held it back.

Why are you so dead set on beating up on a company that is already in bankruptcy, to the point of straining to put down their great technology?

Are you just mad that I won't join your pity party and whine with you about how nothing HF ever did was any good?   

Yes the golden nonce is very good in terms of $/gh (~$0.1/gh I'd guess) but the w/gh is horrible

Thank you for being reasonable.  Did you know the GN1 is very fuel efficient if you underclock it?  You can get it down to about 0.5GH/w at around half speed.

[SolarSilver]

I care about the fact that HF has the fastest chip because how much that IP fetches at auction directly affects how much we customers will be compensated for their delays.

Oh now we have to feel sorry for good old HF and not mention anything negative, because it will decrease the resale value? "we customers will be compensated" are  you kidding me? By dangling that carrot in front of us you basically think we are just as stupid as a mule?

What is this, Stockholm syndrome? Battered Woman Syndrome?

I'm not "defending" anything but the fact that per mm sq of wafer, HF's chip is the fastest.  The chip is the best, although the first boards and early firmware held it back.

Why are you so dead set on beating up on a company that is already in bankruptcy, to the point of straining to put down their great technology?

"their great technology" is basically now worthless and has no future resale value. Nothing has been done with it over the last 9 months. It missed the boat. "It's dead Jim"

[jimmothy]

I care about the fact that HF has the fastest chip because how much that IP fetches at auction directly affects how much we customers will be compensated for their delays.

Please just stop with the fastest chip nonsense. It's not the fastest chip nor is it the most gh/s to be put on a substrate.

What matters is the $/gh. Yes hashfast has a very good $/gh but you cannot say it's the best because you don't know what everyone else pays.

From what I've extrapolated AM/Bitmain/KNC/Innosilicon/Spondooliestech all pay ~$0.1-0.2/gh at the wafer like Hashfast.

Yes the golden nonce is very good in terms of $/gh (~$0.1/gh I'd guess) but the w/gh is horrible

Thank you for being reasonable.  Did you know the GN1 is very fuel efficient if you underclock it?  You can get it down to about 0.5GH/w at around half speed.

Source?

Hashfast's own misleading chart says otherwise:



That says the GN would need to be underclocked to 220gh/s for ~0.55w/gh at the chip which equates to ~0.7 w/gh at the wall and ~$0.6/gh at the wafer.

Overall the it would be mildly impressive for a $1m standard cell design but it's garbage for a $15m full custom.

[iCEBREAKER]

"their great technology" is basically now worthless and has no future resale value. Nothing has been done with it over the last 9 months. It missed the boat. "It's dead Jim"

"Nothing?" Umm, except for the GN1.5 respin and GN2 design work, plus the Yoli EVO and Habanero boards.

But you just want to cry and bellyache freely, unfettered by inconvenient facts.  There, there.  Let it all out.  Scream into your pillow.  Now do you feel better?   

the golden nonce is very good in terms of $/gh (~$0.1/gh I'd guess)

That doesn't sound "worthless."  Having the masks ready to be fabricated into wafers at TSMC seems like a desirable asset, especially if the price of BTC picks up.

[cedivad]

That doesn't sound "worthless."  Having the masks ready to be fabricated into wafers at TSMC seems like a desirable asset, especially if the price of BTC picks up.

Only if they get the rev4 and rev5 ready after that they have been fooling around with them for months, while we were paying their salaries.

And as you said, only if the BTC price skyrockets and the rest of the competition isn't able to keep up with the demand... That is simply not gonna happen.

So yeah, that's another bullshit post.