I know. We've been having a conversation about other vendors in relation to how the hashrate will be when Spondoolies releases their SP30 in August. So by pointing out that BA, KNC, and others are shipping we're discussing the profit potential of the SP30.
Of course now instead of bashing the other companies that some of them are shipping, you don't want to hear about it. Whatever...
Its a valid discussion in my mind, at this point it won't change my pre-order for the sp30s I ordered but it hopefully it illustrates (not like you guys don't already know) the importance of a timely delivery on your 28nm hardware. If it is delayed to NOV DEC we will be in bad shape. Not saying it will happen though. ? We'll ship on time and on spec. Btw: When will May 2nd batch start to be shipped? Any information on this? Early next week.
|
|
|
I know. We've been having a conversation about other vendors in relation to how the hashrate will be when Spondoolies releases their SP30 in August. So by pointing out that BA, KNC, and others are shipping we're discussing the profit potential of the SP30.
Of course now instead of bashing the other companies that some of them are shipping, you don't want to hear about it. Whatever...
Its a valid discussion in my mind, at this point it won't change my pre-order for the sp30s I ordered but it hopefully it illustrates (not like you guys don't already know) the importance of a timely delivery on your 28nm hardware. If it is delayed to NOV DEC we will be in bad shape. Not saying it will happen though. ? We'll ship on time and on spec.
|
|
|
Typically this is done at the very start of chip production and then evaluated, to later on only build optimal chips. Which circumstance has lead spondoolies to receive slow-slow chips in a later batch i dont know.
We're going over the production WAT report to find out. The system was designed to work with all the ASICs corners, except fast-fast. The production test remove fast-fast ASICs. Some SP10 will be a bit better, some will be a bit worse than the average. Will the 5% be credited for every sp10 customer of the May batch?
Yes, I think it's fair. Next week systems might have better numbers, we really don't know yet.
|
|
|
Hi Everyone,
We wanted to notify you of some deviations we found in the current SP10 batch.
The early May batch ASICs we got are of a slow-slow corner ASIC. We've discovered it on Saturday and we're working hard to adapt our firmware.
We've added an overclocking feature to the firmware. Zvi will describe its use in our technical blog.
The measured performance of the SP10 units built with these ASICs is 1.45 - 1.46 THs at 20C temperature. To achieve these numbers, some tweaking of the new settings is recommended:
Increase the initial operational voltage to ~0.69 Increase the max current to 1260W. It will also be possible to control the fan speed and find the optimal compromise between performance and noise.
Users that don't use overclocking Expect around 20 minutes of ramp-up time for the system to achieve its full hashing rate. The hash rate will start at 1.3 TH/s and rise to around 1.44 TH/s after 20 minutes.
Compared with SP10 units that were built with typical corner ASICs running in turbo mode, the W/GHs ratio will be about 4% higher. Note however that both machines will enjoy the same W/GH ratio in the quiet mode.
Both Group Buy and non Group Buy miners are built with those ASICs. Group Buy customers interested in additional SP10 will get 5% coupon.
Guy
Guy, I appreciate your honesty. While this is not the best news it shows your commitment to transparency and keeping customers updated. I noticed this "corner ASIC" being discussed once before. Can you briefly, in layman's terms, describe the difference between a "corner ASIC" and a "non-corner ASIC"? Disclaimer: My background is software. There are manufacturing variants between each wafers lots. It's classic bell shape distribution. Most of the ASICs wafers falls in typical corner. Some falls under "slow" (less leakage, lower frequency at a given voltage). Some falls under "fast" (leakier, higher frequency at a give voltage). To compensate for the "slow slow" corner, we had to increase the voltage supplied to the ASICs by the DC2DC. Zvi can elaborate on it. This is why we quote +- 10% when giving performance numbers.
|
|
|
Hi Everyone,
We wanted to notify you of some deviations we found in the current SP10 batch.
The early May batch ASICs we got are of a slow-slow corner ASIC. We've discovered it on Saturday and we're working hard to adapt our firmware.
We've added an overclocking feature to the firmware. Zvi will describe its use in our technical blog.
The measured performance of the SP10 units built with these ASICs is 1.45 - 1.46 THs at 20C temperature. To achieve these numbers, some tweaking of the new settings is recommended:
Increase the initial operational voltage to ~0.69 Increase the max current to 1260W. It will also be possible to control the fan speed and find the optimal compromise between performance and noise.
Users that don't use overclocking Expect around 20 minutes of ramp-up time for the system to achieve its full hashing rate. The hash rate will start at 1.3 TH/s and rise to around 1.44 TH/s after 20 minutes.
Compared with SP10 units that were built with typical corner ASICs running in turbo mode, the W/GHs ratio will be about 4% higher. Note however that both machines will enjoy the same W/GH ratio in the quiet mode.
Both Group Buy and non Group Buy miners are built with those ASICs. Group Buy customers interested in additional SP10 will get 5% coupon.
Guy
|
|
|
This was probably covered at some point, but it is difficult to look through 94 pages:
What is the projected power usage for SP-30 (for 120 and 240V)? I am in US. I don't want to order it and then struggle to run it.
2500 - 2700, under clocked will be possible. We offer hosting plan in the US Hosting plan in US-is this $1200/3mo? I remember seeing something re WA, but only for people with =>30 SP-30. WA hosting plan was posted somewhere for 290$ per month, 6 month commitment, 100$ setup cost. This will be only available come summer, until now spondoolies can only help people with large orders get to DCs in the US. Correct
|
|
|
This was probably covered at some point, but it is difficult to look through 94 pages:
What is the projected power usage for SP-30 (for 120 and 240V)? I am in US. I don't want to order it and then struggle to run it.
2500 - 2700, under clocked will be possible. We offer hosting plan in the US
|
|
|
The last difficulty adjust was the lowest in almost a whole year, since June 2013. Maybe we will reach a plateau before SP30 ships?
That's that I was thinking too. Besides Ants and SP10s is there anyone else shipping in big numbers? KNC may start shipping neptunes any day now, they taped out in february. There is no single indication about that. Yes there is: The fact that they taped out in february. Chip fab only takes so long, it's been three month so they could receive the chips any day now. They have everything else ready, last fall they only needed a couple of weeks or so from they received the chips to when they started shipments. Why do we discuss KNC at our thread ? The facts: 1) They've TO at the beginning of March, not February. Don't believe everything you read in a press release 2) 20nm with double patterning takes 4-5 months for masks creations Edit: Please continue KNC related discussion at KNC thread
|
|
|
there are orders to fulfill (SP10 May batch) before these orders start shipping, correct?
Yes
|
|
|
SP: can you comment on how many units can be built and shipped in a day?
The SMT is done on one day, the assembly is done on another. About 40-70 miners a day
|
|
|
Everything on time. Talking about "time": When will the May-batch be sent out? Shipping started
|
|
|
Seems that August will be July batch...? No, July batch was already sold out.
|
|
|
SP30 enclosure sample arrived
Awe - I just noticed, no view of the backside? For blaiser: .... next pics with boards and chips ? what's about the 6.5TH/s claimed in another thread ? PCBs without ASICs on the 22th Performance will be 5,700 TH/s - 6,600 TH/s, depending on power modes. +-10% estimation, also depending on ASICs corners.
|
|
|
SP30 enclosure sample arrived
Awe - I just noticed, no view of the backside? For blaiser:
|
|
|
Nothing else in the file. Just double up or even triple up on the security details you have for verification of identity with them; be it security number or codes or passwords or call backs. There is enough video of each of you talking out there that with today's technology anyone determined to try can manipulate anything. Link removed. We're working with one account manager who know the authorized people personally. No way someone may fool him. When ordering a unit and selecting to pay with Euro or $ you get the account details. Our address and the bank branch address is public information. What other sensitive information did you see in the authorization pdf ? Thanks for the sound advice.
|
|
|
Link removed. We're working with one account manager who know the authorized people personally. No way someone may fool him. When ordering a unit and selecting to pay with Euro or $ you get the account details. Our address and the bank branch address is public information. What other sensitive information did you see in the authorization pdf ?
|
|
|
I'd love to see a spondoolies USB stick
We sent Hammer samples to several USB stick makers after they've requested. We don't have the bandwidth to do it ourselves. Same for smaller form factor miner. Napkin math shows a USB 3.0 would allow running it at nominal speed of 7.5ghash An analysis we did for Hammer awhile ago (not tested): Under 600 mV, 4 ASICs, 15W with good DC2DC efficiency we may achieve 25 GHs.
|
|
|
I'd love to see a spondoolies USB stick
We sent Hammer samples to several USB stick makers after they've requested. We don't have the bandwidth to do it ourselves. Same for smaller form factor miner.
|
|
|
|