Pages:
Author

Topic: [ANN] Sfards:SF100, the first 28nm Dual-Mode Miner is accepting pre-order now - page 62. (Read 129811 times)

hero member
Activity: 840
Merit: 1000

@J4bberwock - is that just a footprint for the chip (or circuitry too)? How many layers did you get that on?

Just the footprint for now, I'll have a look at 1 or 2 chips board this weekend, and if it works well, maybe a 18-20 chips in string design board

I don't expect to have it working in less than 4 layers 2oz. 6 layers may be easier/necessary as I like to have large pwr/gnd planes in my boards.
hero member
Activity: 518
Merit: 500
@TheRealSteve - take your own advice .... even better, mind your own business.

@J4bberwock - is that just a footprint for the chip (or circuitry too)? How many layers did you get that on?
hero member
Activity: 686
Merit: 500
FUN > ROI
Perhaps the whole lot of you could respect OP's thread and keep your feud to the threads that revolve around whatever grievances you have with each other?  This isn't the first thread that stuff has spilled over into, but I hope that you could decide to make it the last.

Speaking of the datasheet...
@sfards: on the next revision, could you make sure that the pin names between the pin list, the ball map and the SV schematic are consistent?  Creating a component for a BGA is fun enough without the mismatches Smiley
( example: AVDD_OSC (list/schematic) vs VDDA_OSC (ball map) , ESDSUB_ANA (schematic/ball map) vs unlisted )
hero member
Activity: 840
Merit: 1000
.....
Really neat, but we still don't know if this is at the wall yet?
It's a chip datasheet. "At wall" depends on many factors. DC/DC converter and PSU efficiency. So, if thats not specified all data covers only chip.

up to 25% extra power draw is to be expected at the wall with a gold rated PSU, depending on the design:
10-12% for the PSU, and 10-15% for the power converter that will drop your +12v to 0.6-0.9v

When using string design, you'll save almost all the loss from the power converter.
That's what you can expect from the "24 chips design" they announced at the same time as the 1 chip design.
I still believe that 18-20 chips would be a better choice if they go the string design route.

If some of you are interested in kicad footprint and library for the SF3301/GC3328 chip, I finished mine a few minutes ago.
I'll double check everything tomorrow and share with those interested.





legendary
Activity: 1666
Merit: 1185
dogiecoin.com
The commonality of components doesn't make them cheap. High amp DC-DC is expensive.

Only if you don't know how to design them properly.
Huh That's like saying cars cost less to buy if you're a good driver.

No, it's not. A poorly designed power supply can end up costing a lot more than a well designed one. The car analogy is nonsense.

You're arguing that off the shelf components cost more 'if you design them wrong', which doesn't make sense. If however you were trying to say that 'if you design the system wrong, and use more components than you should...', then sure but that's not what we were discussing.

We already know if that if you use 10 gold ingots to wedge open a door, its going to cost more than if you were to use 5 gold ingots to do the same job, its just a given. What we were actually discussing was the merits of using gold ingots to wedge open a door in the first place.


DC to DC components do have a cost, but the design as a whole can be more efficient (power efficiency and ROI wise) than without.

No one said anything about comparing high amp DC-DC to no DC-DC. But now we are, are you actually trying to argue that string design on small chips makes them less efficient and more expensive?


Big difference between being openly hostile and slagging the competition and having a vigorous debate about the potential of this chip.

Says the guy who entirely just ignored the entire debate to attack me? Roll Eyes
legendary
Activity: 1029
Merit: 1000
.....
Really neat, but we still don't know if this is at the wall yet?
It's a chip datasheet. "At wall" depends on many factors. DC/DC converter and PSU efficiency. So, if thats not specified all data covers only chip.
hero member
Activity: 924
Merit: 1000
I am surprised also by the comment about the cost of such a DC2DC gadget. It seems like that, and an adjustable frequency gadget are pretty common these days, on a whole range of mining devices.

Only if you don't know how to design them properly.

Keep your drivel to the bitmain threads!
DC to DC components do have a cost, but the design as a whole can be more efficient (power efficiency and ROI wise) than without.

Let the man speak. It's good to have discussions so that the community at large can get a variety of opinions, then make up their own mind as to what's information and what's drivel. Maybe that way some of the drivel might disappear.

And you're absolutely right about the overall efficiency of a product, very good insight.

See? You've probably to taught a good few readers to think differently about design.

If you ignore Adam then you can just cut to reading what people that are not bent on shilling for another company have to think. Big difference between being openly hostile and slagging the competition and having a vigorous debate about the potential of this chip.

Thanks to those that cut through his bullshit. Just stop quoting him so I don't have to read it.

sr. member
Activity: 441
Merit: 250
I am surprised also by the comment about the cost of such a DC2DC gadget. It seems like that, and an adjustable frequency gadget are pretty common these days, on a whole range of mining devices.

The commonality of components doesn't make them cheap. High amp DC-DC is expensive.

Only if you don't know how to design them properly.
Huh That's like saying cars cost less to buy if you're a good driver.
Keep your drivel to the bitmain threads!
DC to DC components do have a cost, but the design as a whole can be more efficient (power efficiency and ROI wise) than without.

Let the man speak. It's good to have discussions so that the community at large can get a variety of opinions, then make up their own mind as to what's information and what's drivel. Maybe that way some of the drivel might disappear.

And you're absolutely right about the overall efficiency of a product, very good insight.

See? You've probably to taught a good few readers to think differently about design.
hero member
Activity: 518
Merit: 500
I am surprised also by the comment about the cost of such a DC2DC gadget. It seems like that, and an adjustable frequency gadget are pretty common these days, on a whole range of mining devices.

The commonality of components doesn't make them cheap. High amp DC-DC is expensive.

Only if you don't know how to design them properly.
Huh That's like saying cars cost less to buy if you're a good driver.
Keep your drivel to the bitmain threads!
DC to DC components do have a cost, but the design as a whole can be more efficient (power efficiency and ROI wise) than without.
sr. member
Activity: 441
Merit: 250
I am surprised also by the comment about the cost of such a DC2DC gadget. It seems like that, and an adjustable frequency gadget are pretty common these days, on a whole range of mining devices.

The commonality of components doesn't make them cheap. High amp DC-DC is expensive.

Only if you don't know how to design them properly.
Huh That's like saying cars cost less to buy if you're a good driver.

No, it's not. A poorly designed power supply can end up costing a lot more than a well designed one. The car analogy is nonsense. Most electronic engineers are not familiar with high current low voltage power supply design, its a specialist area and needs careful selection of components, not to   mention circuit board tracking that provides very low resistance paths (and connections). But the world is full of 'experts', is it not, and that's why shitty designs often result. People simply don't accept their limited knowledge of technical subjects and so don't read enough.

legendary
Activity: 1666
Merit: 1185
dogiecoin.com
I am surprised also by the comment about the cost of such a DC2DC gadget. It seems like that, and an adjustable frequency gadget are pretty common these days, on a whole range of mining devices.

The commonality of components doesn't make them cheap. High amp DC-DC is expensive.

Only if you don't know how to design them properly.
Huh That's like saying cars cost less to buy if you're a good driver.
sr. member
Activity: 441
Merit: 250
I am surprised also by the comment about the cost of such a DC2DC gadget. It seems like that, and an adjustable frequency gadget are pretty common these days, on a whole range of mining devices.

The commonality of components doesn't make them cheap. High amp DC-DC is expensive.

Only if you don't know how to design them properly.
legendary
Activity: 1456
Merit: 1000
Not that I am a fan of litecoin, but being able to mine both at the same time is a pretty cool feature. If all ASICs use this technology in the future then the litecoin network will become bigger and bigger... Maybe LTC isn't dead and useless after all?

I was a huge fan of scrypt at one time.  That was until the gear dropped over night on worth.  I almost lost some serious money.

Luckily I rented whole time I had machine so I did ROI..... but  was tightest margin I've ever had.   

I'm more interested in BTC specs personally.
hero member
Activity: 682
Merit: 500
Not that I am a fan of litecoin, but being able to mine both at the same time is a pretty cool feature. If all ASICs use this technology in the future then the litecoin network will become bigger and bigger... Maybe LTC isn't dead and useless after all?
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
I am surprised also by the comment about the cost of such a DC2DC gadget. It seems like that, and an adjustable frequency gadget are pretty common these days, on a whole range of mining devices.

The commonality of components doesn't make them cheap. High amp DC-DC is expensive.
if you want hi amp low ripple and 92% or more efficiency they cost.

somewhere I read a 20 chip machine  if it can not clock under the .6 volts  the low setting will give  .4 watts at the wall like the sp20's do on super low clocks.

not impressed so far but maybe I am missing something or just a bit burnt out by  asic gear.

legendary
Activity: 1666
Merit: 1185
dogiecoin.com
I am surprised also by the comment about the cost of such a DC2DC gadget. It seems like that, and an adjustable frequency gadget are pretty common these days, on a whole range of mining devices.

The commonality of components doesn't make them cheap. High amp DC-DC is expensive.
alh
legendary
Activity: 1846
Merit: 1052
My simplistic read of the Power/Freq/Voltage table suggests that they filled in values for some "reference" points. I am not sure we should interpret that to mean that ONLY .6V and .8V are allowed. The example of the SP20 suggests that it has a "variable" range DC to DC converter, within limits. The same goes for frequencies.

I am surprised also by the comment about the cost of such a DC2DC gadget. It seems like that, and an adjustable frequency gadget are pretty common these days, on a whole range of mining devices. Obviously there are variety of "non functional" combinations (e.g. Min volts and Max Frequency), as well as "poor efficiency" (e.g. Max Volts and Min Frequency). Beyond that it tends to be a limit of software on the extent folks are allowed to twist the Freq/Volt "knobs".

Not surprisingly, the Low Volts and Low Frequency setting yields the best efficiency.

This seems like pretty standard ASIC stuff doesn't it?
legendary
Activity: 1029
Merit: 1000
Thanks, datasheet downloaded. So many voltages....
legendary
Activity: 1666
Merit: 1185
dogiecoin.com
Can't figure out what that FBB voltage is...
forward body bias - see chip datasheet

Quote
Body bias involves connecting the transistor bodies to a bias
network in the circuit layout rather than to power or ground.
The body bias can be supplied from an external (off-chip) source
or an internal (on-chip) source. In the on-chip approach, the
design usually includes a charge pump circuit to generate a
reverse body bias voltage and/or a voltage divider to generate
a forward body bias voltage. Reverse body bias, which involves
applying a negative body-to-source voltage to an n-channel
transistor, raises the threshold voltage and thereby makes the
transistor both slower and less leaky. Forward body bias, on the
other hand, lowers the threshold voltage by applying a positive
body-to-source voltage to an n-channel transistor and thereby
makes the transistor both faster and leakier. The polarities of the
applied bias described above are the opposite for a p-channel
transistor.

So it saves power by utilising a non 0V ground?
hero member
Activity: 686
Merit: 500
FUN > ROI
Can't figure out what that FBB voltage is...
forward body bias - see chip datasheet
Pages:
Jump to: