Goliath

[yohan]

This thread is now defunct. See newer threads on Goliath, Cairnsmore2, Cairnsmore3, Cairnsmore4 and Cairnsmore5.

[1714792080]

1714792080

[1714792080]

1714792080

[1714792080]

1714792080

[1714792080]

1714792080

[meowmeowbrowncow]

Re:  We wont release tech details, but feel free to speculate.

Achieving competitive hash rates with FPGA seems to be so unlikely that I can imagine the worst kind of speculation.  Not to mention how an assumed FPGA based system would compete with future ASIC die shrinks from competitors.

A complete lack of technical details doesn't inspire confidence.  What you are proposing should be believable.  Especially since you are asking for equity funds or a loan.

[2112]

As to the technology it's more valuable to us in other markets so handing out our ideas and knowledge in any shape or form isn't a flyer. That includes even giving a hint of what we are doing.

Hurrah for more speculation! Bit-serial GaAs hashers?

Could anyone ping bitfury for his guess? Anyone else wants to speculate?

[pieppiep]

Massive parallel vacuum tubes?

[CoinHoarder]

I'm not sure how small or large of an investor you're looking for, but I should be receiving a $8000 refund from BFL very soon.

If you could convince me that doing business with you is a better idea than buying an ASIC from another manufacturer, I would probably be interested. But of course, I need more details.

[pazor]

massive mining accelerator by hyperthreating of fpga cores?
 

[spiccioli]

I don't know what you're going to do, but I've got several of your CM1s and I like them even with their usb problems.

I like them even more when I think of the way you made them, the openess in the process and so on.

So, if you're going to sell some goliath units, given that I'm a nerd at the core, I'm in.

If, OTOH, you're going to sell shares or something like shares, I don't know.. I like to own my own gear and I like to have it hashing and tinkering with it.

spiccioli.

ps. I think that right now enterpoint is the only trustworthy company in the bitcoin scene.

[meowmeowbrowncow]

Re:  We wont release tech details, but feel free to speculate.

Achieving competitive hash rates with FPGA seems to be so unlikely that I can imagine the worst kind of speculation.  Not to mention how an assumed FPGA based system would compete with future ASIC die shrinks from competitors.

A complete lack of technical details doesn't inspire confidence.  What you are proposing should be believable.  Especially since you are asking for equity funds or a loan.

At this stage we are not asking people to believe us. You can believe or not we don't actually care at this stage whether you believe. As yet we are not even asking for money merely polling to people to find out if whether our favoured selling mechanisms would be viable. But speaking of blind faith look at the people that sunk huge amounts of money in the existing ???ASIC manfacturers without even test die being available and arguable track record in the ASIC field. Before we ask for money you will be able to see the inital small system results in the network hash rate. It will be big enough to be noticed. We think we can build so much hashing power that it would make the market unstable and we plan to control the hashing power release to avoid creating a problem. The questions should be more of how much investment gets me what hash rate? and maybe when can pay my money and have my share allocation?

As to the technology it's more valuable to us in other markets so handing out our ideas and knowledge in any shape or form isn't a flyer. That includes even giving a hint of what we are doing. What I can say is that it is a culmination of work we have done even before we got into Bitcoin and it is very innovative. We just took the last 8 months to take these ideas forward to being a very serious Bitcoin mining contender.

Yohan

So, if you are going to suggest more valid questions (not my opinion, but I digress) why don't you answer them rather than being rhetorical?

edit
Regarding blind faith.  It's plain to see at this point many customers are not happy with purchases on faith to be delayed many months.  I think less of making purchase decisions without data or without accurate data would be a positive change.

As for me.  Pics or it didn't happen.

[Slipbye]

This is my personal opinion only but if you could offer a nominal trade in for CM1 owners towards this goliath so that there is a bit of love share wise compared to joe blogs in the $/mh shares then i'm in.

One last personal request is this time I would like a cuppa coffee while watching flashing lights on goliath for half hour just so I can feel like my money is invested in something my eyes can see even if my brain wouldn't know wtf it is lol.

Slipbye

[2112]

I like to own my own gear and I like to have it hashing and tinkering with it.

I'm somewhat familiar with this type of deals. There will be no sales whatsoever. You could however negotiate a lease agreement. The equipment would have to be colocated in a secure datacenter (this means IBM, Experian, Wells Fargo, etc. datacenter; not Joe's Datacenter) and maintained only by an approved operator. You will however be able to see it, touch it and evaluate it using a mutually approved methodology, etc. Only certain failure/defect statistics will not be available.

Basically an appropriate operator and datacenter are assurances against reverse engineering and/or spying. This type of arrangements are not unusual in some markets. Think of them as a sort of "technology escrow". Operator assures that the technology vendor doesn't see the data you pass through their equipment and technology user doesn't open the box and reverse engineer it.

[2112]

Before we ask for money you will be able to see the inital small system results in the network hash rate. It will be big enough to be noticed.

This is interesting point, if done outside of market hours where high-frequency trading goes on. In that case Enterpoint may just use an idle time on the FPGA trading machines. There would be no capital outlay for them whatsover, only the electricity cost.

[rocks]

Yohan,

I think Cairnsmore1 was a great product for the FPGA era and respect your ability with boards, etc.

At the end of the day you have some type of semiconductor chip to perform SHA2 hashing. This chip is either a standard FPGA, structured ASIC, eASIC, cell based ASIC, full custom ASIC, or something else.

What type of chip is by far the #1 consideration factor in determining the likely success and longevity of a project. I don't see how people could possibly evaluate the project's potential without this information.

For example: Do you have a wafer piggy back ASIC run available to produce 50 custom ASIC chips at 65nm, but not enough to sell? If so, yes I would be very interested in this project. Are you putting together boards with 100 FPGAs on them and making massive FPGA arrays? If so, no I am not interested.

[rocks]

Before we ask for money you will be able to see the inital small system results in the network hash rate. It will be big enough to be noticed.

This is interesting point, if done outside of market hours where high-frequency trading goes on. In that case Enterpoint may just use an idle time on the FPGA trading machines. There would be no capital outlay for them whatsover, only the electricity cost.

Taking advantage of un-utilized FPGA time would have been great for 2012. However once Avalon sells batch #4 at end of 2013 and the network hash rate is 10x-50x where it is today, even free FPGAs will not break even. Just as free CPUs did not break even after GPUs became mainstream.

As per my note above, the type of semicon chip being used is by far the #1 consideration IMHO. Knowing this, and the source of the chips, is required to make a real analysis of long-term potential of the project.

[Photon939]

I would want my own hardware, I don't want to screw around with buying shares or a lease agreement.

I like being able to work on my own equipment, if I wanted to buy shares of a mining company I can do that elsewhere.

[meowmeowbrowncow]

Yohan,

I think Cairnsmore1 was a great product for the FPGA era and respect your ability with boards, etc.

At the end of the day you have some type of semiconductor chip to perform SHA2 hashing. This chip is either a standard FPGA, structured ASIC, eASIC, cell based ASIC, full custom ASIC, or something else.

What type of chip is by far the #1 consideration factor in determining the likely success and longevity of a project. I don't see how people could possibly evaluate the project's potential without this information.

For example: Do you have a wafer piggy back ASIC run available to produce 50 custom ASIC chips at 65nm, but not enough to sell? If so, yes I would be very interested in this project. Are you putting together boards with 100 FPGAs on them and making massive FPGA arrays? If so, no I am not interested.

@Rocks.  Enterpoint is surpassing lack of disclosure levels that we've seen so far with the likes of BFL.  They aren't offering any technical details.

Yes, a technical premise on which to judge Enterpoint's Goliath longevity would make for a smarter investment decision.  Yohan has basically said no.

What's even more interesting is how they plan to protect investment in the face of favorable competing technologies.  The OP mentions some sort of compensation based on one's investment duration if Enterpoint's Goliath operation becomes less competitive.  That kind of scheme shows they don't have a plan for that.

[MrTeal]

Massive parallel vacuum tubes?

The best part about this is that you can really drive them hard, and even if you do have a hash collision or orphaned block it's not nearly as dissonant or annoying as it is with ASIC based miners.

[DeathAndTaxes]

Given you haven't provided any specs, details, anything it is just rhetorical/academic but please don't provide no worthless mining bonds (which seems to be what you are intended to offer).  I would point out to anyone who may not realize that with a mining bond one doesn't even need the hardware.  Just price them at a point where you guestimate the bond will be worthless long before you paid out more in dividends then you collected on face.  Given global hashpower only rises it is a pretty safe bet.

If you want to run a mining company then run a mining company.  i.e. a registered corporations with officers and directors on file (and subject to real prison for fraud).  Pay your taxes, open books, the whole nine yards.  You know real business.  Come up with a real business plan and at least a prototype which shows you can achieve the hash/$ and hash/J necessary to compete.

Anything less is just a tired scam in a long series of already been done scams.

[2112]

Taking advantage of un-utilized FPGA time would have been great for 2012. However once Avalon sells batch #4 at end of 2013 and the network hash rate is 10x-50x where it is today, even free FPGAs will not break even. Just as free CPUs did not break even after GPUs became mainstream.

As per my note above, the type of semicon chip being used is by far the #1 consideration IMHO. Knowing this, and the source of the chips, is required to make a real analysis of long-term potential of the project.

I think you needlessly extrapolate the current FPGA technology used in the Bitcoin miners against the future process shrink of ASIC miners.

If Enterpoint has capital-outlay-free access to the most recent FPGAs (28nm) then I think they will be quite competitive with ASIC in 110-90-65nm range; based purely on the electricity rates.

We think we can build so much hashing power that it would make the market unstable and we plan to control the hashing power release to avoid creating a problem.

What else could cause "unstability" in the Bitcoin market aside from part time operation? Pretend for a moment that Enterpoint had already delivered (and received payment) for a high-frequency trading hardware that is operated 5*4hours/week during the window where both NYSE and LSE are open.

Long term the type of particular chip doesn't really matter. What matters is the continuous access to the new fabrication nodes and ability to share the development costs between Bitcoin finance and the classical finance.

For a homorous take on a really-long-term prognosis, see the 2nd link in my signature.

[rocks]

If Enterpoint has capital-outlay-free access to the most recent FPGAs (28nm) then I think they will be quite competitive with ASIC in 110-90-65nm range; based purely on the electricity rates.

As someone who used to be in both the ASIC and FPGA fields, and have done many many designs between the 180nm and 45nm nodes, my bet is a full custom ASIC at 110nm will be significantly more power efficient than a 28nm FPGA. The majority of the FPGA's electricity burn is related to routing logic, a full custom ASIC is much more routing efficient and will use significantly less electricity, even at 110nm.

The general run of thumb is an FPGA is 10x slower, 10x higher area (die cost), and 10x higher electricity.

Yohan needs some sort of custom chip if this project will have any legs now that I fully believe Avalon will prove legit.

[2112]

The general run of thumb is an FPGA is 10x slower, 10x higher area (die cost), and 10x higher electricity.

This is a fair point. I don't have any hard data pointing otherwise.

I did however discuss SHA-2 implementations with somebody up-to-speed. His best guess was that the power-reduction factor will be 2.5-5 times. We've come to such conclusion after noticing that the SHA-2 has unusually high toggle rate  (especially in D-type flip-flops) and uses a lot of carry-look-ahead chains (in multi-level adders). Also, the sea-of-hashers (not unrolled) implementation would use comparatively little of the long-distance routing resources.

The area reduction should be easily 10 times, especially after realizing that neither JTAG chains nor even the reset signals are required in the optimized hashers.

The timing reduction should fall somewhere in between the above two.

Last, but not least, all open-source FPGA bitstreams were speed-optimized, not power-optimized. I don't know which optimization strategy was used by the Avalon group, but I'm betting that they've choosen the default (timing closure) due to their tight time-to-market constraints.