Topic: Goliath Miner - page 6. (Read 10805 times)

+1

That's because I find his product interesting but it doesn't fit into a professional setting. His design is close enough (IMO) that it could be adapted to serve both markets. Win-win. I get to fit that in my business plan and he finds more customers without losing his current base.

Maybe there are some potential clients from his customer base that aren't so aware of the other bitcoin mining hardware.  Why don't you just let him be? Let him sell 1Ghash for 1 mil $. If he finds customers then his win, if not then his loss. Not yours.

Good for you!

But I'm still scratching my head at your targeting the retail market with such a product.

This ^

I think their power values and unit sizes should be revised. Besides, unless you make a dedicated server room with proper cooling and cabling, there is no way to run a 6KW heater in your home. That's just crazy/stupid (or both) and generally unwise/dangerous.

I could put this in context that as far as I understand batch3 from Avalon are charging 75 BTC for 66GH/s in their batch3 which I believe is running late and no ETA. On todays conversion that is about £73/GH/s so is that so very different from our £80/GH/s? If it doesn't sell we will simply put it into our own mining operation.

We are also not going to try and compete with manufacturers that have little or no standing cost base and might not be around in 3 months when you have a problem.

I might also speculate that BFL has problems because they charged too little for their miners and their 2.5X recent-ish price rise is an indicator of that. Avalon too have also raised prices through batch 1 to 3 presumably for similar reasons. We have been in business now for 24 years and I am look forward to celebrating 25 still in reasonable financial health and that means charging realistic prices as far as we are concerned.

Costs can be argued about what is good or bad. I have not been following other Avalon based designs in any depth so anything I say is based on limited knowledge. I think the Burnin boards are using a USB interface and that is one we wanted to get away from. It's not great for a big design as we found out with CM1. When you get to 256 COM ports you are stuffed in Windows and maybe Linux too. The CM3 design will have a controller running Linux and miner software so doesn't need a host as such with those issues. We do have a USB interface but that is for special cases or maintainance.

The K384 I don't know much at all so can't make a comment on what they are doing.

With Avalon chips there are some particular problems with offering warranties. I'm not saying this as any insult or other bad mouthing of what Avalon have done but wearing my pro hat they are a supplier with next to no track record and barely have any normal business presence or support. Now if you said to most companies like us they wouldn't touch those chips with a 100 mile bargepole. It fails all the business due diligence tests that are normal. So what does that mean in practicality? It means that you might have poor quality in the design or manufacturing. It might also mean non-working chips arriving. It might mean chips don't appear at all. As a business that represents a huge risk in warranty and litigation. It might also mean a lot of support work and costs. Probably worst of all it could be pissed off customers. I certainly don't want Enterpoint being described in the same BFL is currently. So that is the context we set our prices for CM3. Remember also that CM3 is mainly being done as a service to customers because they asked for it and we can do it in tandem with CM4.

ROI is always going to a difficult one. ROI also changes depending on whether you consider it in flat or BTC. A year ago people talked of an exchange rate that might some day exceed $100/coin. We have been 2.5X that over the last few months. That change has made a big difference in flat profit levels. If we went back to $10/coin flat profits might not be a profit.

In BTC you might say 1/10 of last years BTC earnings is now crap. At best Bitcoin mining is a gamble and you back you horse i.e. equipment to make you money. So in context in buying our equipment we think most customers have, or will, make money and hopefully that will continue to be the case.

We don't say "OMG OP" for any other reason than to help you. We are potential customers saying, if you dont lower price, we wont purchase. You can listen or not, but companies pay big bucks to here why people aren't buying.

While I understand that, the model you provided looks incredibly similar to a single person commercial build around... so I don't want to bitch and did some calc.

The burnin variation including avalon chips comes for me, at around 250€ including chips, VAT, manufacturing in factory that caters from small timers to the german military.

Wages in germany are ridiculously expensive.

So I don't really understand how someone can create a quality product (Including ridiculous German warranty) while doing all this and still make good money. We are talking 44€ vs. 95 Euros. I could understand, 50, 60, even 70 Euros. But more than a double premium on someone who has already hopefully more than a 200% markup?


What I want to say is, I like to invest in premium products. I have money coming towards bitcoin. For me as a customer, what advantages does your product get me compared to other products? I did understand this with GPUs, since your FPGA beats GPUs and others were similarly priced. With Avalon Asics, I don't get the premium.

Then again, sure, at current difficulty you will break even in one month, at further difficulty at 40-50 million you might do it in two to three, totally acceptable. But what advantage does your board have over the Bitburner boards? Or the K384?

To put this into context this is a solution for more less now and we do see CM3 and CM4 being superceeded by CM5 and later "engine" boards. In Bitcoin product lifetimes are always going to be very short and the design we are doing is influenced by that. 90% the work that goes into the early CM3/4 systems will be reused in CM5 onwards and it is entirely possible in time that a single rack could even be comprised of 1 each of CM3, CM4, CM5, CM6, CM7, CM8, CM9 and CM10 to be extreme. That forward progression is our big design aim and the underlying technology can change very quickly to the latest and still work with older elements of a rig. With firmware updates the controllers we supply should be able to run with the newer technology as well. Eventually the standard may have to change say because of some restriction in I/O bandwidth but when we get to that point we will attempt to support a migration usage for "old" boards. That might be a faster Controller or one that has say 10G Ethernet. Whatever the challenge is we will try and bring the older boards along with the new designs.

If you are going to try and fine tune to use 100% the power supply of a DC you are likely to generate costs much more elsewhere in your rig in some other inefficiency. Some would argue that the DC itself is an unnecessary cost and that is why we want to make it capabile of running at home. We also think the potential for heating homes is much more useful than convenience maybe of a DC. It is certainly slightly greener. Also if you are in the camp argueing for hashing to be distributed a lot more then DCs are bad news. You could have 10000 miners basically in same building and they don't not know that. Lose that one building by whatever reason and you lose a big amount of the network. In our own DC plan we have done some thinking about all of these aspects.

Thanks again for the replies Yohan, the following is simply feedback from the perspective of my needs (and is not mean to be critial of the project).

I am completely on board with the density focus of Goliath. The issue I am having is for home builds density is not much of an issue, just look at how most people stacked large CM1 builds. That is the easy way to do it in the home. As a result, I have always seen Goliath as a potential DC build, and I suspect others did as well.

Where I see density being useful is in a DC environment to lower total costs. The issue here is power draw limits density. In the Goliath configuration only 8 boards are supported within a single rack due to power, that is not very dense. So in a DC you end up paying for space that is not used.

Another issue is a DC build should EXACTLY draw the power paid for. If you have 10A assigned to a 10U appliance, then that appliance should draw exactly 10A. If it draws 9.9A it is wasting 0.1A. In the build described above 2 boards drawing 800 watts is not DC efficient, and 3 boards at 1200 watts trips over the limit, which makes it not cost effective.

Overall I am begining to come to the conclusion that builds based on Avalon chips or FGPAs draw too power to be long-term cost effective in a DC environment due to the limit on density. However, my guess is builds based on next gen 50nm Avalon's or BFL chips will be able to achieve an attractive density matched to DC power availablity.

In other words, if we had ASICs that draw half or a quarter of the power of current Avalon chips, then one could really take advantage of the density Goliath offers. That would then get interesting for a large DC build-out.

Being that CM4 uses 1.5 times the energy of Avalon units with similar hashing power I'd say that it contains some high end FPGAs that Enterpoint can buy with a substantial discount because they're ordering a ton of them for some other project they're doing.

spiccioli.

Hello yohan again. Thank you very much for answering my questions. I  just have 1 more regarding the orders Would it be possible to put money down now for an order and put a deposit down on 1x unit and pay as my group comes up with the funds or pay for a basic system and have this on like a buy now build up process a bit like I want an order for the big unit and have this sent in parts ie pay for part working system that provides hashing power and continue to pay for parts and have them shipped.. I can go through back details and invoice for an order within the next week or less and go through payment details and setting some sort of contract out.

My last question that has just spring to mind is about the CM3 and CM4 is their details of what these will provide will they be a bit like the jalas that BFL have like 5GH and 20GH and so forth or is their no details on this yet.

It is commercially too sensitive to for this level of detail to be revealed in open disscussion at this point. Outside of the performance offered, unit power, and what the cost is to customers that won't ever be discussed by us in this forum before customers have units in their hands. However you are free to speculate as much as you like.

Ok so what could be done is that systems are adjusted to fit a DC capabilities. So the obvious way is to fit less cards per system. So 10A at 120V is only 1200W and that is assuming a good power factor. We are maybe looking at each CM3 card taking about 400W so 2 or 3 cards per power input might make sense. If a Goliath rack was fitted with 4 PSUs you could power 8 boards taking 4 x 10A feeds you have available. If you really want to push the space then 1.5 racks is just about possible or stay comfortable at 8 boards or 1 full rack.

The big difference with Goliath is that it is designed to go big and it has a good packing density compared to K16.  You also get the pain of manufacturing removed by paying us to do that in the cost. When you start working at these sizes it becomes a very different design and manufacture problem compared to a K16. That is not in any way being critical of K16 and it has it's place in the mining field.

Over the last 16 months that we have been involved in Bitcoin mining we have learned a lot and I think we still are learning more. We did CM1 last year and that has proved to be a very good product, popular with customers, and with excellent reliability except when placed in torrent of water. Outside of the early phase deliveries the number that have come back to see us on warranty I can count on one hand. That product whilst very simple has taught us a lot in what needed to be done in the Goliath Miner. CM2 which nobody outside of Enterpoint has seen in the flesh has also helped in furthering our ideas and developments. Another little known secret is that we have our own mining software and that might be cut into these miners if we think that is the best way to go. That may offer product stability over third party builds that we have no control over "updates". We will talk more about this aspect as we integrate the design in July.

So what is the CM4? Does it have ASICs? If so from where? Or is it made out of FPGAs somehow brought way down in price (and power requirement?)?

-MarkM-

+1 on that regarding DC

Hi Yohan,

Thanks for providing your responses so far. I was one of the people who expressed early interest in this project, so here is my situation with additional questions. I suspect many others considering Goliath are in a similar situation.

Basically once I gained confidence in the Avalon chip order process and open sourced boards I jumped on board and now have several K16s on order. Open sourced K16 boards are offered at a very attractive Gh/$ price point, which is perfect for home builds because you just stack them in the garage with spacers and fans... However it is only reasonable to build a home rig in the 1~3kW range.

So, if I plan to expand I pretty much have to go into a data center. My options in a DC are :
1) Home build 4U, 8U, etc racks with open-sourced K16 boards
2) Purchase Goliath racks

As a result I need to understand :
a) How Goliath would fit into a DC. This mainly means the minimum block size and power draw so that I can configure that and fit it into a DC environment with their pricing options. For example 12U blocks drawing 120V 10A each.
b) How Goliath compares to a home build option.

The "high-power" DCs in my area seem to top out at 120V 40A per rack. To optimize this I was considering building 8U appliances drawing 10A each for 40A total in a single rack with four 8U appliances in each rack. Working in blocks of 5A or 10A increments seems to be best here.

Thanks,

We are looking to take orders asap so we can secure enough silicon for everyone interested. It is riding off the back of one our other orders and we don't often order amounts of silicon this large. So if we miss ordering now there is some possibility we can't offer the current price again. Behind all these we have other solutions planned for the system but what gets offered and the price will depend on many things. So in summary if you want the £80/GH offer for August you have to order asap. Hard order closure will happen once we hit late July / early August manufacturing limits or in approximately 7 days whichever comes first. We will buy a certain amount of extra silicon over hard orders but that will be strictly limited by our cashflow constraints.