Are there issues at the data centre, all of mine are running really slow like 450GH 80GH slow  Yes. We're checking it now. We'll compensate for the the down time. |
|
|
|
Very similar to our video with Gadi, up to UI presentation starting 4:30 (of course, much more complicated setup ...) Congratulations on the one click upgrade at 7:06. Seems we have started a trend  btw, you forgot to add the auto discovery feature http://myminer.ioNot so bad power consumption, We'll meet again in one month  |
|
|
|
Are my posts bring deleted here? Because on the "sister" board usuing mch better software doesn't allow it and I see my posts there. Any time I talk about price and make sense about it, the post disappears.
We never remove posts. Guy |
|
|
|
I'm seriously looking at buying a few SP30s. Can someone tell me how far along the design of this product has come? Are there units already shipped? Are there prototype results being published or is the SP30 using much of what the SP10 has already proven to work?
I'll save Spondoolies some time and answer what can be answered based on publicly available information: The first SP30s are scheduled to ship in July. The ASIC was taped out before or on May 29th. They should be getting their chips back from the fab pretty soon, but AFAIK they don't have silicon yet. The new design, Rockerbox, is a relatively straightforward die shrink of their old Hammer 40nm ASIC down to 28 nm, and as such is unlikely to have any new major problems. It should use the same serial protocol as Hammer, so there isn't much risk in the new PCB designs either. They have everything finished for the SP30 except for the ASIC. There are a few photos that have been circulated on this thread ( https://bitcointalk.org/index.php?topic=521520.msg7347228#msg7347228) with SP30s fully assembled except for the ASICs themselves. No results have been published that I'm aware of except from the pre-tapeout simulations. Expect around 6 Th/s +/- 10% at 2.7 kW. Based on performance of the SP10s and the common design, the SP30s are likely to have better performance in cold rooms than in warm rooms, and to also perform better if they're run on 208V or 230V rather than 115V. Fair summary. The Tapeout was done on May 14. http://www.spondoolies-tech.com/blogs/news/14226621-rockerbox-2nd-gen-asic-tape-outSame team as our 1st gen. Same vehicle - Global Unichip Taiwan (The chairman of GUC is the vice chairman of TSMC: http://www.guc-asic.com/1-0.php?MD=5) Same CM - Flextronics Israel. Guy |
|
|
|
|
The amount of emails is out of control. Please always cc support@ or info@ so we'll be able to track and eventually get back to you.
Please be patient.
Guy
|
|
|
|
Passive 2-phase immersion cooling is often misunderstood. There is no need for a chiller, it uses a simple dry tower, which is essentially the same thing as a radiator in the front of your car or your water cooling system of a PC. And it also works in very hot climates. The systems are designed to work with hot water (aka hot water cooling). It takes very little energy to cool water from say 70C to 50C even when it's extremely hot outside like in Arizona or other hot places. (this hot water can in fact also be used to heat up buildings like schools or offices) http://www.electronics-cooling.com/2014/02/bitcoin-2-phase-immersion-cooling-and-the-implications-for-high-performance-computing/ The PUE of less than 1.01 includes the _total_ cooling process, including device and building cooling. There is nothing else required. There is no trickery. In fact it's less than that. If you'd use wet towers, or a river near by, it would be lower. But why waste the water and waste time with an open system (needs frequent cleanup)? Thank you for the explanation. We'll continue the discussion in HK 2014 |
|
|
|
I'm under friendly NDA. I'll just comment that it's BS. |
|
|
|
One more thing I want to add:
I believe I know this market quite well. I'm aware of all the 5 big players deployment plans. Locale and deployment methods.
Beside AM, which isn't really a player, none of them are considering IC DT. It's simply way too expensive.
The only one that came close, was 21e6. They experimented with oil cooling and decided eventually to do air cooling.
Disclaimer: My visibility into the Chinese market is very limited.
Guy
Got to go to work.
Who exactly are the big players? Does 70PH+ not qualify? What about bitfurys 20mw DC which will use immersion cooling? I won't reply to the first question. Show me reference to BitFury 20mw DC and IC please. |
|
|
|
|
One more thing I want to add:
I believe I know this market quite well. I'm aware of all the 5 big players deployment plans. Locale and deployment methods.
Beside AM, which isn't really a player, none of them are considering IC DT. It's simply way too expensive.
The only one that came close, was 21e6. They experimented with oil cooling and decided eventually to do air cooling.
Disclaimer: My visibility into the Chinese market is very limited.
Guy
Got to go to work.
|
|
|
|
As for resale, it will still be possible. Shipping blades will be cheap and there is a conversion to air cooling.
Untrue. To take advantage of IC DT, you need to design your boards specifically for that purpose. You can't use them outside.
Although I don't know who you would be selling hardware to. Maybe someone who pays $0.01/kwh?
You'll always find someone to sell it to. It depends on the timing and cost. Maximizing ROI is done by selecting the best timing. With IC DT you can't do that. You end up throw away old HW, which eats considerably into your possible profit. |
|
|
|
... Are you saying the ambient air temp will not be cool enough?
Finally. Locale dependent. Now go read my first post. |
|
|
|
No need for a chiller. Only pump + fans. It works in any climate as far as I know.
This is locale dependent. Can't be otherwise. Simple Thermodynamics. But it can be otherwise. There are novec fluids that boil at a wide range of temperatures. And ? What do you do eventually with the residual heat ? (after all the boiling, transferring and paying for 3M fluids) |
|
|
|
Welcome back!
You never sleep ?
No need for a chiller. Only pump + fans. It works in any climate as far as I know.
This is locale dependent. Can't be otherwise. Simple Thermodynamics. |
|
|
|
DataTank Immersion (3M Novec): <$0.65/W ( PUE 1.01) -> $650/kW... To put this in perspective, Titan has a PUE of 1.25 (25% of electricity go to cooling). Assuming you have nothing to do with the residual heat, how do you handle it ? My assumption is that you still need outside chiller. If you're in an hot locale like Israel, you can't use free air chiller, and your PUE will be much greater than 1.01 I'm guessing the following, correct me if I'm wrong: - For IC DT PUE, you count only the internal looses inside the DataTank, without the external cooling PUE - For Titan PUE (as an example), you count internal + external My point is that if you're deploying in the right locale in terms of electricity cost and weather (e.g. Iceland, Canada or WA state as an example), you can deploy at a much cheaper cost and the internal PUE loses compare to IC DT won't hurt you much. Another issue you don't take into your calculations: With Bitcoin mining, the best strategy, IMHO, is to mine for 3-4 months on new hardware, afterwards upgarde the hardware and sell the old one for nice residual value (25%-50% in some cases). You simply can't do that with Immersion Cooling DataTanks. Guy |
|
|
|
Why are Guy and Jimmothy slugging it out in the KNC thread ? You are both seriously off-topic.
Indeed. Got carried away. Going to sleep now. |
|
|
|
You're clueless, on both of the statement you made.
Quick question: Did you read the IPO terms ?
Quick answer: I skimmed through it. Why? Which statements? ... First payments are expected when the initial hardware is deployed, 3-6 months after successful fund raising ...
I'll admit that I don't know how you can build a 1MW DC for less than $200k.
But how does that change the fact that datatank is offering $0.93/gh and $0.025/kwh where as you are offering $0.85/gh and ~$0.25/kwh starting 3 months later?
Apples and Oranges. The electricity cost is simply the locale. Fill a DT with RockerBox and you'll get much, much better numbers. And as I've explained, you can get much, much, much, much better numbers with other cheap deployments methods. Edit: And on a serious note. The affiliation between AC DT and AM is very loose. It was stated in the IPO for one reason only, get all AM fan boys buy another stock. Cynical but effective move. Working as we can see. |
|
|
|
Guy, if I wanted to a 1mW datacenter using your hardware could you deliver it in a month (30 day window) or would this be $200k and be ready in September or October; or with a large customer could something be delivered sooner along with an earlier batch. If I had any kind of proof that this was possible I might fully be inclined to building out seeing as financially that is no issue and knowledge and resources from our personal company (we are one of the larger IBM partners in the southern US (sales wise)). I looked at the past in doing a large roll out but even utilizing SP10's I could not see it as feasible.
The reality of you having someone hold off 3 months+ compared to filling it with current hardware that is available costly for profits (everyone knows this). You could be cycling hardware as it comes in and out. I should note that my post bashing the fan boys of your hardware is in no way related to you all as a company. We have talked via email and PM multiple times over the last few months and you all are probably one of the most down to earth mining companies, so my beef is not at all with SP-Tech.
Unfortunately not. The ASIC lead time is 3 months - TSMC 28nm allocations. If you know TSMC CEO, it might help, but I'm not sure. Really. They're playing fair with all the customers. PO by the 5th, ASICs 3 months afterwards. I'm not hiding ASICs or miners. All July and August were sold out. We have decent September batch, which is selling fast. We're working to secure funds to cover October batch. I've explained lead time issues in our thread. Guy |
|
|
|
...
So will you be offering sp30s for $3200 shipped in a few weeks?
No. If I'll do that, what will happen to AM SH dividends ? Having you suffered enough ? On second thought, contact sales@ and we'll do a special deal for you. You'll be able to compare it with the all kind of rocks coming out from china  You won't really have a choice. As soon as these datatanks go up for sale you will have to match their price to compete gather preorders. Thanks for the offer but I don't have that kind of money to spend on hardware especially a preorder. And BTW those "rocks from china" are powering a majority of the network (well at least before difficulty exploded) You're clueless, on both of the statement you made. Quick question: Did you read the IPO terms ? |
|
|
|
...
So will you be offering sp30s for $3200 shipped in a few weeks?
No. If I'll do that, what will happen to AM SH dividends ? Haven't you suffered enough ? On second thought, contact sales@ and we'll do a special deal for you. You'll be able to compare it with all kind of rocks coming out from china |
|
|
|
I'd love to know how you can get much below $200k for a 1mw DC. Rent alone should be ~$50-100K/year.
Probably government subsidies AFAIK, it's not the case. It is BitFury case, I believe. |
|
|
|
|
Show Posts
