[ANN] Spondoolies-Tech - carrier grade, data center ready mining rigs - page 129.

philipma1957

legendary

Activity: 4354

Merit: 9201

'The right to privacy matters'

Quote from: bigblind on March 03, 2015, 10:41:23 AM

yes I've set them to 3. But temperatur was never over 85° on the downclocked ones, they are pulling fresh air and it is maybe 6°C at the moment outside.

I think it really might be a lose connection on the cable. Yesterday I moved the two SP20s. Maybe during that one of the connectors got lose.

it could be loose connection but 3% is low and 600 watts on two cables is high.

I have run 2 sp20e's on 1 evga but I staggered the loads and upped the fan. Hard to tell which was the cause here. Maybe a little of both?

But I do believe 3% fan and 600 watts on two cables is high. I also think that the connection may have been less then ideal due to movement you mentioned.

bigblind

sr. member

Activity: 378

Merit: 252

yes I've set them to 3. But temperature was never over 85° on the downclocked ones, they are pulling fresh air and it is maybe 6°C at the moment outside.

I think it really might be a lose connection on the cable. Yesterday I moved the two SP20s. Maybe during that one of the connectors got lose.

Here you go, that one is the badboy Cheesy

Temperatures are fine, arent they?

TracerX

hero member

Activity: 918

Merit: 1002

Quote from: Pt0x on March 03, 2015, 08:52:10 AM

Quote from: philipma1957 on March 03, 2015, 07:52:28 AM

Quote from: bigblind on March 03, 2015, 06:19:29 AM

most likely the 3% fan speed and the max loaded wires are the cause.

Maybe he was using the auto fan feature?. Personally I think it lacks a target temp setting or a min/max speed threshold.

The auto-fan only steps down to six, so he went in and manually set the fan to run at three.

Guy Corem

donator

Activity: 1414

Merit: 1051

Spondoolies, Beam & DAGlabs

To FC and David, I hope all is well in BitQuan.
I hope you'll emerge stronger and wiser from those trying times.

Guy

Pt0x

sr. member

Activity: 266

Merit: 250

Quote from: philipma1957 on March 03, 2015, 07:52:28 AM

Quote from: bigblind on March 03, 2015, 06:19:29 AM

most likely the 3% fan speed and the max loaded wires are the cause.

Maybe he was using the auto fan feature?. Personally I think it lacks a target temp setting or a min/max speed threshold.

mavericklm

hero member

Activity: 742

Merit: 500

if contact is perfect and overload is over all wires, then it's always the middle ones that fails first!

here is just bad contact between the male and female connector

philipma1957

legendary

Activity: 4354

Merit: 9201

'The right to privacy matters'

Quote from: bigblind on March 03, 2015, 06:19:29 AM

posted it already in the Spondoolies-Tech Thread, but here again:

Can anybody tell what just happened?
I just checked my miners and saw a lower Hashrate on one of them. One of the ASIC Loops seemed down.

Checked the cables and saw this:

As you can see one of the Pins seemed to have burned?! The ASIC hasnt been running for a week. It was even underclocked. (to about 620watts).

EDIT::

A switch of the cables seems to have fixed it.
It was a 2x8 Pin cable, powering 2x 6 Pins on the SP20. Might it be that the cable itself has been overloaded?

How do you connect your EVGA Supernovas to 2x Sp20s? It has only 6 VGA Connectors.

what were the hash rates of each machine on the evga 1300 g2?

and what were the wall watts of each machine on the evga 1300 g2.

I see that with the 3 left it reads 444 wall watts or 445 which means it was about 600 or 595 with 4 working that should be okay. but I see you set the fan at 3% .. so the wire got too hot for 2 or 3 reasons.

1) max watts of 580 to 620 watts.
2) really slow fan speed
3) did you reverse the fan this would really warm the connectors at the sp20e side.

most likely the 3% fan speed and the max loaded wires are the cause.

I would hook 2 sp20e's to one evga 1300 g2.
run one at 1050gh with 10% fan the one with the 2 headed pcie wires
run one at 1150 with 6% fan the one with 4 pcie wires.

also do 800 pixel photo not 640 very hard to read numbers.

wh00per

hero member

Activity: 572

Merit: 500

Quote from: mavericklm on March 02, 2015, 03:35:36 PM

just a reminder:

Quote

12U, 16kw, 10 replaceable hashing boards.

I'm really curious how did SPT power 16kW of hashing boards. Generally, the rackware is also sized accordingly to the power input so that you will max out the power supplies on it. They need an intellingent power supply if they still do the BIST aproach ..

Now with 16KW and the 2000W power supply used in SP35 it'll be an ugly/noisy beast with 8x PSU with their little whining fans. and I'd hate to use 8x 208V power cables in a 9x (3 phase) PDU just for 1 unit. In the end the savings would be in installing only 1 controller per 10 hasing boards and a little mechanical work to manufacture a larger case versus more smaller ones.

If the cases are as deep as SP3X the rear chips will still overheat .. and with a "backplane" in the rear, one will have to vent the units sideways or vertical. Crap.

I'd put the "backplane" on a side .. and make the units as deep as 2 ASIC space (half deep of a SP20 .. or OK .. the PSU depth at max). Instead of using high velocity fans, I'd rather use the largest they find. I'd include 2 port network cards (like a small switch) in each unit, to be able to daisychain them.

If you go industrial .. I'd look at the PLC designs .. and their extension cards.

bigblind

sr. member

Activity: 378

Merit: 252

posted it already in the Spondoolies-Tech Thread, but here again:

Can anybody tell what just happened?
I just checked my miners and saw a lower Hashrate on one of them. One of the ASIC Loops seemed down.

Checked the cables and saw this:

As you can see one of the Pins seemed to have burned?! The ASIC hasnt been running for a week. It was even underclocked. (to about 620watts).

EDIT::

A switch of the cables seems to have fixed it.
It was a 2x8 Pin cable, powering 2x 6 Pins on the SP20. Might it be that the cable itself has been overloaded?

How do you connect your EVGA Supernovas to 2x Sp20s? It has only 6 VGA Connectors.

opentoe

legendary

Activity: 1274

Merit: 1000

Personal text my ass....

Quote from: RoadStress on March 02, 2015, 02:55:39 PM

Quote from: Dr Charles on March 02, 2015, 02:21:55 PM

8U at 10kw 16-20TH would be perfect for me as well. I would look to run 1 or 2 of a unit at that size.

I disagree!

8U at 10kW and 50TH/s would be perfect!

Be nice, but I'd hate to have to pay shipping. Once you get up to those sizes it turns into commercial shipping which is terribly expensive. It almost makes sense not to even buy hardware and just buy bitcoin. I've installed a few 8U and larger size servers and when full can weigh several hundred pounds. Just tack on $900-$1200 for shipping.

Biffa

legendary

Activity: 3234

Merit: 1221

Quote from: opentoe on March 03, 2015, 03:53:16 AM

Quote from: klondike_bar on March 02, 2015, 11:44:44 AM

Quote from: opentoe on March 01, 2015, 10:27:13 PM

Spondoolie, you know just like everyone else a home user can't buy in bulk, and if that's all you are going to do in the future then the single consumer is off your marketing list. You for some reason just don't want to verbally say it. It is your company to do with what you want, but bulk sales to just a few investment groups isn't going to help bitcoin in my opinion.

If an SP100 comes out thats ~8U, 10kW, 16TH I would be able to operate 1-2 of them. Its definitely not bulk, but it would be a practical size for the serious (ie: >$5000 investment) miner since it reduces on the cost of using multiple controllers and shipping boxes to otherwise ship 3-4 SP3x units

@SP-T: Would be nice to see a clever upgrade path given for old SPT equipment, namely the power supplies. For example, a 750W PSU that could swap out with the 1050W PSU in an underclocked SP10 miner. that 1050W could then be used in newer gear like an SP100.

That was my point I was trying to make before. I don't think they will sell someone just 1 or 2 or 3 or 4. Only bulk, maybe a min of 15-20 or something. Their business model will be for deep pockets only. %90 can't just spend $20k or more at the drop of a dime. I mean their SP31 that still says in stock has had the same price for quite a while now. I doubt they're even wanting to sell those individually.

Thats not how I read what they have said. But I doubt they will come out with a "home" miner again. They will stick to their datacenter optimised models going forwards. But I think they will sell single units, its not like they do free shipping, you pay for the shipping so its no harder for them to ship one than ship 3 or 5 or 10.

opentoe

legendary

Activity: 1274

Merit: 1000

Personal text my ass....

Quote from: klondike_bar on March 02, 2015, 11:44:44 AM

Quote from: opentoe on March 01, 2015, 10:27:13 PM

Spondoolie, you know just like everyone else a home user can't buy in bulk, and if that's all you are going to do in the future then the single consumer is off your marketing list. You for some reason just don't want to verbally say it. It is your company to do with what you want, but bulk sales to just a few investment groups isn't going to help bitcoin in my opinion.

If an SP100 comes out thats ~8U, 10kW, 16TH I would be able to operate 1-2 of them. Its definitely not bulk, but it would be a practical size for the serious (ie: >$5000 investment) miner since it reduces on the cost of using multiple controllers and shipping boxes to otherwise ship 3-4 SP3x units

@SP-T: Would be nice to see a clever upgrade path given for old SPT equipment, namely the power supplies. For example, a 750W PSU that could swap out with the 1050W PSU in an underclocked SP10 miner. that 1050W could then be used in newer gear like an SP100.

That was my point I was trying to make before. I don't think they will sell someone just 1 or 2 or 3 or 4. Only bulk, maybe a min of 15-20 or something. Their business model will be for deep pockets only. %90 can't just spend $20k or more at the drop of a dime. I mean their SP31 that still says in stock has had the same price for quite a while now. I doubt they're even wanting to sell those individually.

philipma1957

legendary

Activity: 4354

Merit: 9201

'The right to privacy matters'

Quote from: klondike_bar on March 02, 2015, 06:50:35 PM

Quote from: RoadStress on March 02, 2015, 05:20:40 PM

Quote from: klondike_bar on March 02, 2015, 05:16:26 PM

if using rockerbox that would mean about 20-30TH depending on the number of chips and their voltages (likely 25TH)

using a gen2 28nm would probably result in 10-20% more hashrate/watt. using a 20nm or 16nm would improve efficiencies by about 2.5-3x

Don't underestimate SP-Tech design capabilities. Bitfury showed to all that there is plenty of squeezing left from the 28nm chips. I expect from SP-Tech to at least match them!

IMO theres about a 20% improvement possible via a good redesign. maybe even 30% if the new chips can run at <0.6V. another 10-20% improvement might come from improving the voltage regulator design.

I doubt they will be able to do better than 0.3w/GH @wall with a 28nm design, and that would be near the lowest hashrate. at full speeds 0.5-0.6W/gh is likely

If my sp20 did .3 watts and 700gh I would be happy.

But I earned enough that it would be cheap on a $$ per gh number. Ie 200usd with shipping for a 700gh machine that does .3 watts.

klondike_bar

legendary

Activity: 2128

Merit: 1005

ASIC Wannabe

Quote from: RoadStress on March 02, 2015, 05:20:40 PM

Quote from: klondike_bar on March 02, 2015, 05:16:26 PM

if using rockerbox that would mean about 20-30TH depending on the number of chips and their voltages (likely 25TH)

using a gen2 28nm would probably result in 10-20% more hashrate/watt. using a 20nm or 16nm would improve efficiencies by about 2.5-3x

Don't underestimate SP-Tech design capabilities. Bitfury showed to all that there is plenty of squeezing left from the 28nm chips. I expect from SP-Tech to at least match them!

IMO theres about a 20% improvement possible via a good redesign. maybe even 30% if the new chips can run at <0.6V. another 10-20% improvement might come from improving the voltage regulator design.

I doubt they will be able to do better than 0.3w/GH @wall with a 28nm design, and that would be near the lowest hashrate. at full speeds 0.5-0.6W/gh is likely

Fatman3001

legendary

Activity: 1554

Merit: 1014

Make Bitcoin glow with ENIAC

Quote from: klondike_bar on March 02, 2015, 05:16:26 PM

Quote from: mavericklm on March 02, 2015, 03:35:36 PM

just a reminder:

Quote

12U, 16kw, 10 replaceable hashing boards.

if using rockerbox that would mean about 20-30TH depending on the number of chips and their voltages (likely 25TH)

using a gen2 28nm would probably result in 10-20% more hashrate/watt. using a 20nm or 16nm would improve efficiencies by about 2.5-3x

20nm won't let you do anything near that. There is little reason to stop by the 20nm mark.

sjc1490

hero member

Activity: 546

Merit: 500

Quote from: RoadStress on March 02, 2015, 05:20:40 PM

Quote from: klondike_bar on March 02, 2015, 05:16:26 PM

if using rockerbox that would mean about 20-30TH depending on the number of chips and their voltages (likely 25TH)

using a gen2 28nm would probably result in 10-20% more hashrate/watt. using a 20nm or 16nm would improve efficiencies by about 2.5-3x

Don't underestimate SP-Tech design capabilities. Bitfury showed to all that there is plenty of squeezing left from the 28nm chips. I expect from SP-Tech to at least match them!

According to CoinDesk (12/1/2014) article that is exactly what Guy said.

Spondoolies-Tech has decided to stick with the mature 28nm node for its third-generation ASIC, despite the fact that the company's first FinFET chip will ship a couple of months after rivals will have launched their 16nm FinFET ASICs.

"We have our own third-generation design, still 28nm, which will be comparable to their 16nm designs," said Corem.

Edit: Reference:
http://www.coindesk.com/spondoolies-tech-ceo-talks-new-asics-blockchain-lottery-device/

RoadStress

legendary

Activity: 1904

Merit: 1007

Quote from: klondike_bar on March 02, 2015, 05:16:26 PM

if using rockerbox that would mean about 20-30TH depending on the number of chips and their voltages (likely 25TH)

using a gen2 28nm would probably result in 10-20% more hashrate/watt. using a 20nm or 16nm would improve efficiencies by about 2.5-3x

Don't underestimate SP-Tech design capabilities. Bitfury showed to all that there is plenty of squeezing left from the 28nm chips. I expect from SP-Tech to at least match them!

klondike_bar

legendary

Activity: 2128

Merit: 1005

ASIC Wannabe

Quote from: mavericklm on March 02, 2015, 03:35:36 PM

just a reminder:

Quote

12U, 16kw, 10 replaceable hashing boards.

if using rockerbox that would mean about 20-30TH depending on the number of chips and their voltages (likely 25TH)

using a gen2 28nm would probably result in 10-20% more hashrate/watt. using a 20nm or 16nm would improve efficiencies by about 2.5-3x

ElGabo

hero member

Activity: 635

Merit: 500