If I photoshopped Spondoolies over Cointerra, it would look very similar to products I have from you. Are you two working together in any way?
Since we have mutual NDA between the companies, I can't answer in any way. I'll just say that it's absurd to deduct it based on a "Q1" product page. I've noticed the similarity also and was amused. But come on, read the product description. Do you see us shipping without embedded controller? |
|
|
|
... weird that TSMC works with them ...
I've mentioned TSMC because of 16ff+ related information (I've been asked spec and date). I'm not breaking an NDA when saying that Cointerra is working with GUC. Read the product page again. |
|
|
|
Under NDA with Cointerra and TSMC. I'll just state that the product and page looks familiar somehow  Do your research and learn from the past. Q1 sounds familiar also. |
|
|
|
bla bla bla ...
Due to the diligent work of AM minions, our little thread is now #3 in replies rank. Thank you. As I wrote before, FC in person is a great guy. BitQuan seems like a capable company. I like them. Can't say the same about some of their SHs Guy  |
|
|
|
You can consider this off topic, but I think I must say this.
After working with almost any available for third parity developers ASIC chips for the last year We can state that the RockerBox is state of the art in terms of delivering the advertised parameters , communication and flexibility.
Also the Spondoolies team could not be more responsive in order to assist us during the development .
Thank you. RockerBox will improve with each batch starting from October, as we continue to tweak the production parameters. |
|
|
|
Why is the SP20 less efficient than the SP35?
I thought they used the same ASICs?
SP20 1100W / 1700 GH/s = 0.65J/GH
SP35 3500W / 6000 GH/s = 0.58J/GH
Indeed same ASICs. The SP20 should give better results then what we published. We'll get the faster ASICs in 2 weeks and revise the spec. Also, iit depends on the PSU efficiency. |
|
|
|
|
From initial testings with the slow ASICs we currently have, we see great over clocking potential in the SP20.
The limiting factor is cooling, and we're working to improve it. Experimenting with various enhancements to the heatsinks.
|
|
|
|
Thanks, I've alerted Coindesk. Guy |
|
|
|
...
Deleted. Please avoid personal attacks. |
|
|
|
Each of the 4 PCIe connector power 2 RockerBox ASICs, independently. Power diode from each power the management board. In theory, one PCIe connector is enough to run SP20 at 1/4 speed.
Make sure the PSU PCIe connectors can handle 300 Watt.
So if you put an ohm meter across the 12V of one PCI-E and the ground of another, you're saying there won't be any signal back? They're absolutely independent? How were you proposing to put a third PCI-E port on the boards? Absolutely independent. I wasn't purposing a third PCI-E port on the boards. If we'll see a need, it can be implemented, for example PCI-E port per ASIC (total 8, 4 per board) |
|
|
|
Does anyone know if the 4 power inputs on the SP20 are independent? ie: can I use multiple separate PSUs w/o causing a problem? I looked on the product page but didn't see anything in the specs either way.
Guy (Spondoolies-tech) already posted about it: you can use 2X600W PSU or 4X300W PSU according to him (perhaps he was joking about the latter one) with the only caveat that PCIE connector has to be rated for ~300W (ATX)-this is actually in bold red in TOS for Sp-20. This information is typically not readily available, but can be had from manufacturers on request. Seasonic X-1050 and EVGA 1300 have this type of connectors. First-according to Guy, second-according to the post on the KNC forum that I cited beforehand. You can use 2x PSUs but NOT 4x. Each pair of PCI-E connectors is electrically separate across 12V, but within the pair its not. I think you are confusing the AntMiner S3 with the SP-20 I asked Spondoolies directly and they said, "Yes, the power connectors are independent. You can use 4 small PSUs, assuming the PCIe connector can handle 300 Watt" Each of the 4 PCIe connector power 2 RockerBox ASICs, independently. Power diode from each power the management board. In theory, one PCIe connector is enough to run SP20 at 1/4 speed. Make sure the PSU PCIe connectors can handle 300 Watt. |
|
|
|
In general, to win in 2015 you'll need in descending order of importance:
1) Team or access to team with years of circuit design experience in advance nodes.
2) Access to cheap electricity - sub 0.03 $/KWh or less
3) Cheap production
or a nice SHA-256 exploit Always helpful, assuming you don't abuse too much |
|
|
|
1. Did you miss the announcement of avalons 28nm? (The 3rd chinese company to successfully make a 28nm chip)
2. Are you sure absolutely nobody in china can get electricity below $0.03/kwh? Is there a reason the farms need to be built in china?
1. No. I've been expecting it for a month now. It's a bit difficult to "argue" with you when you don't understand what I've meant by "circuit design" and gives standard cell design as a counter example. 2. Yes. I did my homework. |
|
|
|
The russian self taught asic designer?
Valery (both of them) won't appreciate being called Russian. But yes, BitFury is #1 on my list. They have enough $ to hire top world talent. They have proven execution track record. |
|
|
|
Is bitmain selling out batch after batch of S3's considered struggling?
Where did I say that? You've missed the point that I was talking about 2015. The nature of the business is that it resest itself with each generation. |
|
|
|
Is Israel the only country with people who have experience in advanced nodes?
Definitely not. |
|
|
|
These competition points raised by Sptech could be valid, but i think he forgot about a major factor that may undermine the whole industrials at this point: btc price..
Whatever cheap electricity or manufacturing costs you could get, peofitability will be driven by btc price obviously. And with it stabilisīng or tanking even more, some rushers might get squeezed and go bankrupt.
Interesting times ahead.
Under current BTC value, the bounty is $1.2BN until the next block incentive halve. For sake of argument (only ), let's assume you're able to develop 0.2 $/GHs or 0.1 $/GHs (or below ) hardware. How does the math look like ? |
|
|
|
No love for the little guy anymore?
Always. It depends on our relative advantage. In 2015, it might be in a form of a mining facility "bond". |
|
|
|
number 2 is pretty irrelevant for miner manufacturing. Question regarding number 1. How strong is IP protection since you still have to get the chips made over there anyways?
My reply is regarding 2015, which will be all about mega industrial farms or franchise / JV type deployments. Access to multi MW low cost electricity will be the limiting factor. China lacks cheap electricity and free MWs. No issue with IP protection, manufacturing is done at Taiwan (TSMC) at least for us. |
|
|
|
I have respect to AM and Bitmain past achievements, but they're not high on my list. I don't believe China will be able to compete in 2015
Can you elaborate as to why? I'm under NDA with various parties, so I can't give concrete information. In general, to win in 2015 you'll need in descending order of importance: 1) Team or access to team with years of circuit design experience in advance nodes. 2) Access to cheap electricity - sub 0.03 $/KWh or less 3) Cheap production From what I was able to learn, the Chinese teams (including some publically unknown) has only (3) above. Some of them are rushing to bleeding edge nodes (TSMC 16 or Samsung/GF 14) but it's not enough. |
|
|
|
|
Show Posts
