Author

Topic: ASICMINER: Entering the Future of ASIC Mining by Inventing It - page 374. (Read 3917543 times)

legendary
Activity: 1904
Merit: 1007
Quote
for small miners I'm still waiting to see 4 hashratio's miners stacked like the 4 SP10

Not exactly 4 but this might be what you're looking for.(10th/s)



I see that you are in a good mood if you are posting rotated images just to please me. I will give you credit for that Smiley
legendary
Activity: 1036
Merit: 1000
Nighty Night Don't Let The Trolls Bite Nom Nom Nom
put it in a sack, drown the fu*ker and watch it survive and spawn again.

well done to the team for getting us this far in 1 year since gen1
hero member
Activity: 770
Merit: 509
Quote
for small miners I'm still waiting to see 4 hashratio's miners stacked like the 4 SP10

Not exactly 4 but this might be what you're looking for.(10th/s)

legendary
Activity: 1904
Merit: 1007

I am talking more about mass deployment/rack storage where space is critical. It really shouldn't matter if you're mining at home since you have the space and not enough heat to matter.

I don't have a video since these are brand new but heres a pic:

Also have no idea how big they are. 3u is just my guess.


So for mass deployment/rack storage where space is critical you are saying that 3U? is much better than 1.25U? There is a contradiction in that statement.

I will clarify:

IF you can cool 40kw/rack then you can have greater density with SP10 over a hashratio machine.

However the majority of datacenters cannot cool anywhere near 40kw/rack. More like 10-20kw/rack meaning SP10's would need to be spaced ~3u apart.

So without extreme cooling, heat density for both devices are about the same.

But if you are going to go with extreme cooling, you might as well go with 2 phase immersion cooling.

Let me clarify:
- for small miners I'm still waiting to see 4 hashratio's miners stacked like the 4 SP10
- for big miners: well anyone investing a lot of money into miners will surely find a way to cool a 40kw/rack. If they won't then they need to space them apart, but I am sure that you can fit 2 SP10 miners in a 3U space. It leave just enough space between the miners. Even if they space them bigger, you are still trying to compare a miner which is available since March with one that nobody has except the Chinese manufacturers.
legendary
Activity: 1260
Merit: 1002
My question is, from my reading if the design is good a fully custom chip outperforms standard cell chips. If that's wrong why would anyone do full custom design?

Premium. Profit. But less quantity and more unstable.
hero member
Activity: 770
Merit: 509

I am talking more about mass deployment/rack storage where space is critical. It really shouldn't matter if you're mining at home since you have the space and not enough heat to matter.

I don't have a video since these are brand new but heres a pic:

Also have no idea how big they are. 3u is just my guess.


So for mass deployment/rack storage where space is critical you are saying that 3U? is much better than 1.25U? There is a contradiction in that statement.

I will clarify:

IF you can cool 40kw/rack then you can have greater density with SP10 over a hashratio machine.

However the majority of datacenters cannot cool anywhere near 40kw/rack. More like 10-20kw/rack meaning SP10's would need to be spaced ~3u apart.

So without extreme cooling, heat density for both devices are about the same.

But if you are going to go with extreme cooling, you might as well go with 2 phase immersion cooling.
legendary
Activity: 1904
Merit: 1007

I am talking more about mass deployment/rack storage where space is critical. It really shouldn't matter if you're mining at home since you have the space and not enough heat to matter.

I don't have a video since these are brand new but heres a pic:

Also have no idea how big they are. 3u is just my guess.


So for mass deployment/rack storage where space is critical you are saying that 3U? is much better than 1.25U? There is a contradiction in that statement.

When someone tells you that SP-Tech has better power consumption you are bragging about AM's chip being faster. When someone tells you that SP-Tech has better miner density you are bragging about price. Always derailing from the subject. It seems that you are trying so hard to compare SP10 miners which are available since March with AM miners which were just released (well not really because they have chips in hands for almost 2 months now) and which don't seem to exist out in the miners hands.
full member
Activity: 224
Merit: 100
I wish Spoondoolies would also go public so I can have some shares during the AM takeover.  Grin
hero member
Activity: 770
Merit: 509
Gimmick for what? Here is a picture with 4 stacked with no problem(thank you tzortz):


3U is the final case for the hashratio's miners? Show me a video of one running in a miner's farm.


I am talking more about mass deployment/rack storage where space is critical. It really shouldn't matter if you're mining at home since you have the space and not enough heat to matter.

I don't have a video since these are brand new but heres a pic:

Also have no idea how big they are. 3u is just my guess.

donator
Activity: 1414
Merit: 1051
Spondoolies, Beam & DAGlabs
My question is, from my reading if the design is good a fully custom chip outperforms standard cell chips. If that's wrong why would anyone do full custom design?

I'm not feeling comfortable posting here. It's a legit question, so I'll answer. Please continue further discussion via PM

Sure, fully custom chip done correctly will outperform standard cells chips. The shining example is BitFury rev2 55nm UMC ASIC
To get a killer ASIC, you need to have better SHA256 implementation, coupled with custom design and some more elements I won't discuss.
Such a killer ASIC could achieve 0.1 W/GHs and sub 0.05 $/GHs on 28nm process
legendary
Activity: 1904
Merit: 1007
As you say, ASICMINER isn't selling mining machines directly. Most of us are invested in ASICMINER because we want in on the profits from chip sales and franchising. The mining income at this point is going to be dwarfed by the income from those other two sources, and that's fine by me. Comparing Spondoolies and ASICMINER seems like an apples-and-oranges comparison to me.

Yes franchising underpreforming ASICs is the right way to profit from chip sales.



Since spondoolies + fanboys are all here I thought I'd ask, have any of you even come close to ROI on your $4/gh miners?

Must be painful to see asicminer machines shipping now as low as $1/gh especially if you paid $1/gh to preorder the sp30 which will be delivered in months.

First of all asicminer isn't shipping anything. They are selling chips to third-party. Secondly I still haven't seen any of these miners in the wild while my SP10 miner is working hard to make me happy. Thirdly I'm hashing with 1.5Th/s in a small 1.25U box. How much space/cables/PSU/etc is needed for asicminer machines to reach 1.5Th/s?(not to mention the power consumption)

You do realize the 1.25u case is a gimmick? It is impossible to cool 1kw/1u so you need to spread them out. That means both the 1.25u sp10's and the (3u?) hashratio machines both take up the same space.

But for those wanting a machine that sounds like a chainsaw, SP10 wins hands down.

Gimmick for what? Here is a picture with 4 stacked with no problem(thank you tzortz):


3U is the final case for the hashratio's miners? Show me a video of one running in a miner's farm.
full member
Activity: 216
Merit: 100
Don't let the nam-shub in your operating system.
My question is, from my reading if the design is good a fully custom chip outperforms standard cell chips. If that's wrong why would anyone do full custom design?
hero member
Activity: 770
Merit: 509
Since spondoolies + fanboys are all here I thought I'd ask, have any of you even come close to ROI on your $4/gh miners?

Must be painful to see asicminer machines shipping now as low as $1/gh especially if you paid $1/gh to preorder the sp30 which will be delivered in months.

First of all asicminer isn't shipping anything. They are selling chips to third-party. Secondly I still haven't seen any of these miners in the wild while my SP10 miner is working hard to make me happy. Thirdly I'm hashing with 1.5Th/s in a small 1.25U box. How much space/cables/PSU/etc is needed for asicminer machines to reach 1.5Th/s?(not to mention the power consumption)

You do realize the 1.25u case is a gimmick? It is impossible to cool 1kw/1u so you need to spread them out. That means both the 1.25u sp10's and the (3u?) hashratio machines both take up the same space.

But for those wanting a machine that sounds like a chainsaw, SP10 wins hands down.
hero member
Activity: 714
Merit: 500
As you say, ASICMINER isn't selling mining machines directly. Most of us are invested in ASICMINER because we want in on the profits from chip sales and franchising. The mining income at this point is going to be dwarfed by the income from those other two sources, and that's fine by me. Comparing Spondoolies and ASICMINER seems like an apples-and-oranges comparison to me.
legendary
Activity: 1904
Merit: 1007
Since spondoolies + fanboys are all here I thought I'd ask, have any of you even come close to ROI on your $4/gh miners?

Must be painful to see asicminer machines shipping now as low as $1/gh especially if you paid $1/gh to preorder the sp30 which will be delivered in months.

First of all asicminer isn't shipping anything. They are selling chips to third-party. Secondly I still haven't seen any of these miners in the wild while my SP10 miner is working hard to make me happy. Thirdly I'm hashing with 1.5Th/s in a small 1.25U box. How much space/cables/PSU/etc is needed for asicminer machines to reach 1.5Th/s?(not to mention the power consumption)
hero member
Activity: 770
Merit: 509
Since spondoolies + fanboys are all here I thought I'd ask, have any of you even come close to ROI on your $4/gh miners?

Must be painful to see asicminer machines shipping now as low as $1/gh especially if you paid $1/gh to preorder the sp30 which will be delivered in months.
legendary
Activity: 896
Merit: 1001
AM is not the latest crap in bitcoin world.  Wink



aaaactually, i beg to differ.

AM - 0.554 J/Gh at 11.52 Gh/s.
Spondoolies - 0.58 J/Gh at 7.5 Gh/s.



Not my thread, but I need to set the record straight.
Our ASICs can achieve above 10 GHs. It was just one corner measurement.
At the end, it's system efficiency. I saw three examples of systems based on AM3 in the wild, all of them above 1.1 W/GHs.
Out system is about 0.85 W/GHs at full power.
We've taped out our 2nd gen ASIC, and we'll be below 0.45 W/GHs at the system level in two months.

AM3 can't compete on 28nm, since it's "custom design", which means it will take at least 6 months to produce 28nm. Well in time for our 3rd gen.
Finally, we can compete on the price as well.

Can you guys imagine FC showing up in this guy's thread and commenting about how they are doomed?


Friedcat won't.  He a man of few words as results matter, not rhetoric.
legendary
Activity: 1904
Merit: 1007
Rockxie claimed that AM's ASIC does 11.52 GH/s using 0.72V and consumes 0.554 J/Gh.

http://blog.rockminer.com/#!/2014/04/13/Testing_Results_Of_BE200.md

You are giving too much credit to a single ASIC test board measurement, which was probably done to ease the pressure from the shareholders. The chip can have 0.1J/Gh in a test environment, but as a miner I care about how much power is my miner sucking from the power outlet and it seems that all AM gen3 miners need at least 1.1W/GH (~1.1J/Gh/s to be correct). My Oct/Nov Jupiters are having that power consumption and I mined a lot of coins up until now. Even I know now that chips can be slow-slow, fast-fast etc and I'm not a tech guy. It's obvious that the average chips perform much worse than what was shown more than 1 month ago. Even the blog post has "Tips:this result is not very accurate just for reference." at the end so I wouldn't take that as granted.
full member
Activity: 224
Merit: 100
Not my thread, but I need to set the record straight.
Our ASICs can achieve above 10 GHs. It was just one corner measurement.
At the end, it's system efficiency. I saw three examples of systems based on AM3 in the wild, all of them above 1.1 W/GHs.
Out system is about 0.85 W/GHs at full power.
We've taped out our 2nd gen ASIC, and we'll be below 0.45 W/GHs at the system level in two months.

AM3 can't compete on 28nm, since it's "custom design", which means it will take at least 6 months to produce 28nm. Well in time for our 3rd gen.
Finally, we can compete on the price as well.

Can you guys imagine FC showing up in this guy's thread and commenting about how they are doomed?

Laughing way too hard about this right now!  Cheesy

He should have the courtesy to also visit the various Gen3 system builders and let them know too.
hero member
Activity: 686
Merit: 500
A pumpkin mines 27 hours a night
Not my thread, but I need to set the record straight.
Our ASICs can achieve above 10 GHs. It was just one corner measurement.
At the end, it's system efficiency. I saw three examples of systems based on AM3 in the wild, all of them above 1.1 W/GHs.
Out system is about 0.85 W/GHs at full power.
We've taped out our 2nd gen ASIC, and we'll be below 0.45 W/GHs at the system level in two months.

AM3 can't compete on 28nm, since it's "custom design", which means it will take at least 6 months to produce 28nm. Well in time for our 3rd gen.
Finally, we can compete on the price as well.

Can you guys imagine FC showing up in this guy's thread and commenting about how they are doomed?

Laughing way too hard about this right now!  Cheesy
Jump to: