[GLBSE] BFLS.RIG - BFL Hardware mining & Sales

[Epoch]

I've ran the calculations for both BFLS and RIG, based on an upgrade to SC holding to an identical power envelope.

For BFLS:

FPGA Single   
power: 80 watts
cost:   $600.00   
shares: 200 per Single
cost/share:   $3.00   
      
SC Single      
hashrate: 60000   Mhps nominal
power: 300 watts (a guess for now)
upgrade cost: $700.00   
Amount of SC Singles for same power as FPGA:   0.267   
upgrade cost from 80W FGPA to 80W SC: $186.67   
extra shares created per FPGA Single to SC upgrade:   62 (assigned to Inaba)
hashrate/share after upgrade: 61.0 Mhps

For BFLS.RIG:

FPGA Minirig:
power: 1250 watts (a guess)
cost:   $15295
shares: 6058 per minirig
cost/share:   $2.52   
      
SC Minirig:    (this is the original 1.5Thps unit; I can re-run these for the 500Ghps unit if it makes more sense)   
hashrate: 1500000 Mhps nominal
power: 7500 watts
upgrade cost: $14705 (a guess; does anyone remember the original number?)
Amount of SC minirigs for same power as FPGA:   0.167   
upgrade cost from 1250W FGPA to 1250W SC: $2450.83
extra shares created per FPGA minirig to SC upgrade:   970 (assigned to Inaba)
hashrate/share after upgrade: 35.57 Mhps

It is worth mentioning that, after the upgrade to SC, the hashrate of a BFLS share will significantly exceed that of a BFLS.RIG share (by 71%). If anyone can contribute more accurate numbers let me know and I'll update these.

UPDATE 1: edited for 7500W minirig SC (1.5Thps)

[Inaba]

300w estimate on the Singles is a good estimate.  The minirig will be 500 GH/s +/- 10% for about 2500w or so, or 7500w for 1.5 TH/s.

[Epoch]

Yes, that makes sense (5W per gigahash/s). I've updated my post using 7500W per 1.5Thps for the minirig SC. Looks like people holding BFLS will be in a much better position than BFLS.RIG holders (61Mhps/share vs 35Mhps/share respectively) ... which is why it remains important to keep the 2 share groups distinct.

(the reason for this large difference is that the FPGA Singles use about double the power per gigahash than the FPGA minirig boards. Thus they can be converted to a lot more SC hardware while keeping within the same power envelope).

[Vbs]

Guys, I still think there's something we're missing? Aren't we seeing the problem in reverse?  

For example, an FPGA single has ~10.40 MH/s/W, FPGA Mini-Rig ~18.80MH/s/W while a Single SC has ~200.00 MH/s/W. I did the following estimates using current diff=15,605,633, $/BTC=100.56 and $/kWh=0.15 and a monthly time frame.



The FPGA single has currently a ~10.9% power cost overhead, the FPGA mini-rig ~6% and the SC single will have a ~0.57% power cost overhead.

[Vbs]

(Updated above post with FPGA mini-rig values)

(...)
Overall, I would say the average for a Minirig is closer to 23 - 23.5 GH/s not 25, on average.  Some overperform, but most are underperforming on average due to cards going flakey.

For a Single SC (60GH/s@300W) to have a ~10% overhead in power costs, the difficulty has to be around 275 million ($32.87 power cost/month at $0.15/kWh vs. $329.03/month revenue @ $100.56/BTC). Assuming $/BTC goes up, the power costs in $ go down proportionately.

[Epoch]

Looks like gigavps has an SC minirig hashing already:

https://bitcointalk.org/index.php?topic=237710.msg2514804#msg2514804

Any ETA of when BFLS will start getting SC equipment?

[RHA]

I missed the posts from last week, so I can answer only now.

Inaba, you worries unnecessarily. You should simply order/allocate more power for each coming new SC device. 10% of the increased revenue will easily cover the new costs. Vbs has proved it.

Each 200 BFLS shares were entitled to 200/440 of the 60 GH/s after upgrading to Single SC.
At the beginning you intended to replace each one Single to one Single SC (for ease of calculations lets forget for a moment about conversion to Minirigs).
It would give for your disposal additional 240/440 of the power. I think you should keep to that.

The alternative is to use lesser number of Singles SC, eg. half of them - replacing each two old Singles with one Single SC.
Each 400 BFLS shares would be then entitled to 400/440 of the Single SC hashing power, and you would get 40/440 of it.

With Minirigs you should proceed similarly.

The important point is to keep the level of hashing power per share equal to what was promised to be after upgrading. The price of the shares was driven by this expectancy.

[Epoch]

RHA makes some good points ...

[Vbs]

(...)

For BFLS (Singles):
- currently 1 Single represents 200 shares and is worth $599. 832Mhps per Single.
- 1 BFLS share is thus worth $3.00 and corresponds to 4.16Mhps.
- a Single SC upgrade costs $699 yielding 60,000 Mhps.
- Inaba would assign 433 shares total for each SC Single: 200 shares (the original $599) would stay with the current owners. The additional 233 shares (the $699 upgrade cost) would be assigned to Inaba.
- each BFLS SC (Single) share would still be worth $3.00 as before ($3 x 433 shares is $1299, the full price of a Single SC); shareholders would not lose (or gain) any value.
- each BFLS SC share would produce 138.4Mhps (60000 / 433 shares).

For BFLS.RIG (MiniRig):
- currently 1 MiniRig represents 6058 shares and is worth $15295. 26400Mhps per MiniRig.
- 1 BFLS.RIG share is thus worth $2.52 and corresponds to 4.4Mhps.
- a MiniRig SC upgrade costs $14604 yielding 1,500,000 Mhps.
- Inaba would assign 11842 total shares for each SC MiniRig: 6058 shares (the original $15295) would stay with the current owners. The additional 5784 shares (the $14604 upgrade cost) would be assigned to Inaba.
- each BFLS.RIG SC share would still be worth $2.52 as before ($2.52 x 11842 shares is $29899, the full price of a MiniRig SC); shareholders would not lose (or gain) any value.
- each BFLS.RIG SC share would produce 126.6Mhps (1500000 / 11842 shares).

(...)

Following RHA, my proposition to keep power costs at a minimum would be to trade only the "necessary" singles. Inaba told me he has almost 40 fpga singles atm, so that's around 3200W.

If he were to "upgrade" all the 40 fpga singles, those would give 2.4TH (40*60GH), of which,

• 2.4TH * 200/433 = 1.108TH would go to BFLS shareholders
• 2.4TH * 233/433 = 1.292TH would go to Inaba

So "only" 1.108TH is needed to fulfill all BFLS shareholders, which is about 40*200/433= 18.48 asic singles, consuming ~5543W.

Since we need an integer number and there is an overhead to pay for power, 19 asic singles should be enough (consuming a total of 5700W), since those remaining 0.52 (worth 227 shares, around 31455GH) yield at the moment $2629.81/month, while the power costs for all 19 asic singles (5700W) should be around $625/month ($0.15/kWh).


The underlying math is simple,

For BFLS: [#new asic singles]*433 - [#existing fpga singles]*200 = [#new BFLS shares assigned to Inaba]. So, with 19 asic singles, Inaba will get 19*433 - 40*200 = 227 new shares.

For .RIG: [#new asic minirigs]*11842 - [#existing fpga minirigs]*6058 = [#new BFLS.RIG shares assigned to Inaba].

Or, alternatively,

For BFLS: [#new asic singles]*433 - [#existing fpga singles' shares] = [#new BFLS shares assigned to Inaba].

For .RIG: [#new asic minirigs]*11842 - [#existing fpga minirigs' shares] = [#new BFLS.RIG shares assigned to Inaba].

Using this proposal, it's just a matter of Inaba deciding on the number of new shares he wants/needs to cover for the overhead cost margin of running the hardware.

[Inaba]

I'm fine with swapping out for the same power footprint, as that is ultimately the root of the problem.  Space can be managed most likely, but power is much more difficult to manage - 19 singles would consume almost double what power consumption is now, which is right around 3000w. 

Let me see what I can shuffle around and see if I can somehow fit almost 20 singles in somewhere.

[Vbs]

I'm fine with swapping out for the same power footprint, as that is ultimately the root of the problem.  Space can be managed most likely, but power is much more difficult to manage - 19 singles would consume almost double what power consumption is now, which is right around 3000w.  

Let me see what I can shuffle around and see if I can somehow fit almost 20 singles in somewhere.

Maybe now it would be more beneficial to you (to increase power efficiency) to just convert all those singles into minirigs? Using Epoch's values, a "1.5TH asic minirig" is 11842 .RIG shares, so a real 500GH asic minirig is 3947.3(3) shares.

So, [#new 500GH asic minirigs]*(3947+1/3) - [#existing fpga singles]*200*3/2.52 = [#new BFLS.RIG shares assigned to Inaba].

With three 500GH asic minirigs, that's 3*(3947+1/3) - 40*200*3/2.52 = ~2318.19 .RIG shares to you; existing BFLS shareholders get converted to .RIG shares = 3/2.52*#BFLS shares.

(another option is using 2*500GH minirigs + 4*SC singles, for example)

Edit: Asic MiniRigs behave internally as 8 SC singles (480TH), see my post below.

[Epoch]

SC minirigs and SC Singles have roughly the same power efficiency. Space requirements are different.

[Sukrim]

Oh, I just looked it up now that this thread reappeared and it seems I have still one single unclaimed share in BFLS.RIG...

Any way to still claim that?

[Vbs]

SC minirigs and SC Singles have roughly the same power efficiency. Space requirements are different.

True.

I was checking giga's post (https://bitcointalk.org/index.php?topic=237710.0) and it seems each 500GH minirig behaves internally like 8 SC singles with a max of 480GH (8*60GH), so if space is not a problem the SC singles should be the best choice (more redundance with larger area for heat dissipation, lesser work for fans).

So, if Inaba prefers to convert everything to SC singles (or do a mix, with one asic minirig counting as 8 sc singles), using the .RIG referencial (more granularity),

(hardware # based): [#new SC singles]*433*3/2.52 - [#existing fpga singles]*200*3/2.52 - [#existing fpga minirigs]*6058 = [#new BFLS.RIG shares assigned to Inaba]
or
(share # based): [#new SC singles]*433*3/2.52 - [#existing fpga singles' shares]*3/2.52 - [#existing fpga minirigs' shares] = [#new BFLS.RIG shares assigned to Inaba].

This assumes all BFLS shareholders get their shares converted to .RIG shares, using [#shareholder's new .RIG shares] = 3/2.52*[# shareholder's actual BFLS shares].

Edit: Added formula.

[Epoch]

Just to clarify: The conversion factor of existing BFLS shares to 'new shares' would be diffferent than the conversion factor of existing BFLS.RIG shares to 'new shares'; there should not be a 'blanket' conversion if we want to be fair (I own approximately equal amounts of both so this doesn't affect me much, but it would negatively impact those holding more BFLS than BFLS.RIG). I assume we want to be fair.

BFLS is worth more than BFLS.RIG in terms of $; even more so if we are constrained to the same power envelope simply because a BFLS share represents significantly more power than a RIG share and can be converted to significantly more SC hardware/hashrate/income.

My calculations a few posts up show the details of this.

[Vbs]

Just to clarify: The conversion factor of existing BFLS shares to 'new shares' would be diffferent than the conversion factor of existing BFLS.RIG shares to 'new shares'; there should not be a 'blanket' conversion if we want to be fair (I own approximately equal amounts of both so this doesn't affect me much, but it would negatively impact those holding more BFLS than BFLS.RIG).

BFLS is worth more than BFLS.RIG in terms of $; even more so if we are constrained to the same power envelope simply because a BFLS share represents significantly more power than a RIG share and can be converted to significantly more SC hardware/hashrate/income.

My calculations a few posts up show the details of this.

Absolutely! BFLS shares are worth 19% more ($3/$2.52) than BFLS.RIG shares, that's why each single sc is worth 433*3/2.52 = ~515.48 BFLS.RIG shares.

I've always accounted for that in my formulas above.

[Epoch]

Just to clarify: The conversion factor of existing BFLS shares to 'new shares' would be diffferent than the conversion factor of existing BFLS.RIG shares to 'new shares'; there should not be a 'blanket' conversion if we want to be fair (I own approximately equal amounts of both so this doesn't affect me much, but it would negatively impact those holding more BFLS than BFLS.RIG).

BFLS is worth more than BFLS.RIG in terms of $; even more so if we are constrained to the same power envelope simply because a BFLS share represents significantly more power than a RIG share and can be converted to significantly more SC hardware/hashrate/income.

My calculations a few posts up show the details of this.

Absolutely! BFLS shares are worth 19% more ($3/$2.52) than BFLS.RIG shares, that's why each single sc is worth 433*3/2.52 = ~515.48 BFLS.RIG shares.

I've always accounted for that in my formulas above.

That's valid if our power envelope is unrestricted. If Inaba wants to maintain the SAME power envelope, BFLS is actually worth 71% more than BFLS.RIG. This is because BFLS represents FPGA Singles that can be upgraded to a lot more hashpower than the equivalent in BFLS.RIG (representing FPGA minirigs that use about 1/2 the power of the FPGA Singles for a given hashrate and thus cannot be upgraded to as much SC hardware). See my post here:

https://bitcointalk.org/index.php?topic=89902.msg2476669#msg2476669

[Vbs]

That's valid if our power envelope is unrestricted. If Inaba wants to maintain the SAME power envelope, BFLS is actually worth 71% more than BFLS.RIG. This is because BFLS represents FPGA Singles that can be upgraded to a lot more hashpower than the equivalent in BFLS.RIG (representing FPGA minirigs that use about 1/2 the power of the FPGA Singles for a given hashrate and thus cannot be upgraded to as much SC hardware). See my post here:

https://bitcointalk.org/index.php?topic=89902.msg2476669#msg2476669

Yep, I saw your post, but unfortunately, that solution penalizes heavily BFLS.RIG owners, since their shares are now only worth (35.57MH/61MH) 58% of BFLS shares, while they should be worth ($2.52/$3) 84% of BFLS shares. I hope we really go with an alternative solution.

[Epoch]

That's valid if our power envelope is unrestricted. If Inaba wants to maintain the SAME power envelope, BFLS is actually worth 71% more than BFLS.RIG. This is because BFLS represents FPGA Singles that can be upgraded to a lot more hashpower than the equivalent in BFLS.RIG (representing FPGA minirigs that use about 1/2 the power of the FPGA Singles for a given hashrate and thus cannot be upgraded to as much SC hardware). See my post here:

https://bitcointalk.org/index.php?topic=89902.msg2476669#msg2476669

Yep, I saw your post, but unfortunately, that solution penalizes heavily BFLS.RIG owners, since their shares are now only worth (35.57MH/61MH) 58% of BFLS shares, while they should be worth ($2.52/$3) 84% of BFLS shares. I hope we really go with an alternative solution.

Agreed, but that's what the numbers say under the power budget restriction. Pragmatically, though, I doubt we'll have any trouble reaching a reasonable solution that will make most shareholders (including Inaba, of course) happy.

The key is that we start getting SC hardware hashing for shareholders ASAP!

[Vbs]

I have no doubt we can reach an equitable solution that will make everyone (including Inaba, of course) happy.

The key is that we start getting SC hardware hashing for shareholders ASAP!

May your words echo through the halls of the bitcoin gods!