Pages:
Author

Topic: Antminer S5 or SpondooliesTech SP20 to order - few questions (Read 3613 times)

legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
Today I ordered 4 SP20s!
Guess I will have them here by next week.

Anybody can send me the config for getting it to about 600-650watts? Cheesy


sure

this will give you 595 watts at the wall

you can run 2 on an evga 1300 g2  each will be in the 1100-1150 gh range

make sure to set your firmware to 2.6.7 that allows for auto fans




 fan is 6%  hash  is  1123gh  wall watts = 597
sr. member
Activity: 322
Merit: 250
Good luck with the SP20's.  As an owner of many S5's, I would say probably a good choice to go with the SP hardware.  The build quality of the S5 has gone way down from the S3.  Although it's a more favorable comparison on paper for the W/GH, you can undervolt the SP20 to achieve basically the same numbers, from a product that seems to be much more well built and capable of hashing solidly at those speeds.

I'm down 1/6th of my original purchase of S5's due to various problems - mainly having a whole chain die and leaving the machine running at half speed.  I'm not going to bother replacing them and just ordered a couple SP-35's instead as replacements so as the S5's die I'll just slowly switch to SP gear.
sr. member
Activity: 322
Merit: 250
Today I ordered 4 SP20s!
Guess I will have them here by next week.

Anybody can send me the config for getting it to about 600-650watts? Cheesy
I believe there's a thread for it:

https://bitcointalksearch.org/topic/spondoolies-sp20e-one-impressive-miner-0526-wgh-933536
sr. member
Activity: 378
Merit: 252
Today I ordered 4 SP20s!
Guess I will have them here by next week.

Anybody can send me the config for getting it to about 600-650watts? Cheesy
legendary
Activity: 2128
Merit: 1005
ASIC Wannabe
There is a copper base plate, but it is connected to the aluminum heat sink by a layer of thermal compound, without any mechanical binding, this defeated the purpose of that material

Are those heat-sinks fairly "standard" ? Is it possible to buy higher quality replacements that might work in the SP20's ?

its possible, but not even remotely worth it.

16x heatsinks = $25 if you get a bargain
time spent opening a case, melting the glue holding down the curent heatsinks = 1hr, voided warrenty, likely at least 1 cracked or snapped component
time spend properly cleaning, pasting, and applying a new heatsink to the chips then re-assembling = 2hrs

if you are lucky and do it all properly, you will either be able to reduce fans at full speed by 10-15%, or otherwise perhaps push the unit to 1800/1350W at the most extreme. Thats about a gain of 100GH/150W for the cost of 3hrs + $25.  For reference, $25 put towards the cost of a new $400 SP20 would be effectively worth 150GH/100W

if you're unlucky you will re-assemble the device and find that multiple chips no longer read due to various errors (you cracked the asic, nicked a resistor, w/e), and effectively lose a large amount of your device functionality

its not worth a mod
legendary
Activity: 1652
Merit: 1067
Christian Antkow
There is a copper base plate, but it is connected to the aluminum heat sink by a layer of thermal compound, without any mechanical binding, this defeated the purpose of that material

Are those heat-sinks fairly "standard" ? Is it possible to buy higher quality replacements that might work in the SP20's ?
full member
Activity: 122
Merit: 100
at the end only J/GH counts. at the wall. and undervolting S5 by PSU is topic since 2 month but without results.
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
- can be undervolted to 10-11V to achieve even better efficiency - in the range of 900/360 (SP20 doesnt really get any better than 900/420). This requires uncommon 12-10V voltage regulators though and generally there's 5%+ losses in doing so

can you link to such experiments, where the gains of the S5 are higher than the losses of the DC2DC step down devices?



there is no such experiment posted. 


 once sidehack gets his build to work correctly  it will be 95% so

10/12  is 83.33% divide by 95%  =   87.71 

so a 10 volt setting would  use  87.71 percent of the normal watts. (not  83.33%)

 as sidehack says he is close.

If he gets it to work this gear should allow newer designs to use it.  like a psu it will move from generation to generation of miners.

   It may influence  asic design to be like the  s-5 in string
full member
Activity: 122
Merit: 100
- can be undervolted to 10-11V to achieve even better efficiency - in the range of 900/360 (SP20 doesnt really get any better than 900/420). This requires uncommon 12-10V voltage regulators though and generally there's 5%+ losses in doing so

can you link to such experiments, where the gains of the S5 are higher than the losses of the DC2DC step down devices?

sr. member
Activity: 378
Merit: 252
Okay. Thank you all for your responses. Unfortunately the S5s delivered straight from shops in Germany rose up yesterday in price as up to 489€...
I think the choice will fall for the SP20. As its price even went down from more than 799$ already? The S5 is just getting more expensive and therefore I will suffer a bigger pricecrash once they start to fall off.

I read that there might be a problem with uncentered pins of the PCIe-plugs. Mine are like that. Anybody tried to "mod" the case for them to fit? And what is the difference between batch 1 and batch 2?
legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
First prototype of that adapter is assembled and was tested today, but it's giving me some fits and needs more work. I should have good news soon though.
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'

In other words: which is the better miner, when
  • hashing on same speed (1,1-1,3 TH/s)
  • powering 2 units by one of these PSUs
  • having the same prices per unit to me
  • regarding the noise

Appreciate your help.

...
- can be undervolted to 10-11V to achieve even better efficiency - in the range of 900/360 (SP20 doesnt really get any better than 900/420). This requires uncommon 12-10V voltage regulators though and generally there's 5%+ losses in doing so

What was the lowest voltage someone has verified the chips start at? I had a look at getting a variable PSU in but they were stupidly expensive.

someone tinkered with it and found it really wasnt stable or capable of the advertised 9V = ~0.25w/gh, and that the lowest settings were somewhere around 10.5V or ~0.4w/gh.  I almost want to say it was phillipma, but am not 100% i think its in one of the S5 review unit threads
MR Teal got 9.6 or 9.7 volts 
it is on these threads somewhere.   sidehack is working on a plug n play adapter for a standard atx psu should go to 10 volts and be solid.  still waiting for one for testing.

Right now the sp20 is the plug n play choice.  the s-5 is the choice if you want to get your hands dirty.

1x 11 dollar delta fan which can run pwm via splice to a pwm will overclock to 1275gh  okay for noise 60db

 or splice to  molex 7 volt   will run stock at 1150gh pretty quiet 53db
legendary
Activity: 2128
Merit: 1005
ASIC Wannabe

In other words: which is the better miner, when
  • hashing on same speed (1,1-1,3 TH/s)
  • powering 2 units by one of these PSUs
  • having the same prices per unit to me
  • regarding the noise

Appreciate your help.

...
- can be undervolted to 10-11V to achieve even better efficiency - in the range of 900/360 (SP20 doesnt really get any better than 900/420). This requires uncommon 12-10V voltage regulators though and generally there's 5%+ losses in doing so

What was the lowest voltage someone has verified the chips start at? I had a look at getting a variable PSU in but they were stupidly expensive.

someone tinkered with it and found it really wasnt stable or capable of the advertised 9V = ~0.25w/gh, and that the lowest settings were somewhere around 10.5V or ~0.4w/gh.  I almost want to say it was phillipma, but am not 100% i think its in one of the S5 review unit threads
legendary
Activity: 1666
Merit: 1185
dogiecoin.com

In other words: which is the better miner, when
  • hashing on same speed (1,1-1,3 TH/s)
  • powering 2 units by one of these PSUs
  • having the same prices per unit to me
  • regarding the noise

Appreciate your help.

...
- can be undervolted to 10-11V to achieve even better efficiency - in the range of 900/360 (SP20 doesnt really get any better than 900/420). This requires uncommon 12-10V voltage regulators though and generally there's 5%+ losses in doing so

What was the lowest voltage someone has verified the chips start at? I had a look at getting a variable PSU in but they were stupidly expensive.
legendary
Activity: 2128
Merit: 1005
ASIC Wannabe

In other words: which is the better miner, when
  • hashing on same speed (1,1-1,3 TH/s)
  • powering 2 units by one of these PSUs
  • having the same prices per unit to me
  • regarding the noise

Appreciate your help.

- at about 1.16TH, the units are about equal for power consumption (~580W)
- for 2 units on a 1300W PSU, either would work fine, as you would be running at about 2.35-2.4TH/1.2kW with either system
- for the same price, take the SP20
- noise-wise, you can turn the SP20 down to 10% or less fan speed within its OS. on the S5, you need to physically modify it to provide lower voltage to the fan (ie: use molex 7V adapter instead of PWM)

the SP20 is a far better OS by a long shot. If you want a device thats about 1150-1200GH/600W its the winner.

where the SP20 excels:
- if an altcoin comes out or BTC price sharply rises, you can push the SP20 to 1.65TH/1.2kW if extra hashrate is worth the lost efficiency
- fan control and custom power settings

where the S5 excels:
- overclocking to 1300GH/660W (whereas the SP20 is about 1300GH/700W) or as high as 1400/720 (Sp20 is 1400/800)
- can be undervolted to 10-11V to achieve even better efficiency - in the range of 900/360 (SP20 doesnt really get any better than 900/420). This requires uncommon 12-10V voltage regulators though and generally there's 5%+ losses in doing so


Bitmain's next step is to use the S5 chips in a longer chain to achieve something like 1.5TH/500W. I don't think spondoolies is capable of improving the current rockerbox chip any further than 0.46w/GH unless they also go with a chain design or create a chip that's better suited to lower voltage
legendary
Activity: 1988
Merit: 1012
Beyond Imagination
Just found this post here:

The heat sink on SP20 is just entry level plain aluminum heat sink, they don't have huge surface area like those heat pipe heat sink with many thin aluminum fins, so the airflow must be very high to reduce the surface temperature

I just measured 20% fan speed noise at 1 meter away with my sound meter app, it is still 67db, and I personal feel is close to in car noise on highway (70+-db)

Would be interesting what hashspeed is reachable with these settings.

If the heatsinks are the same as the SP30/35, they've got a 4-5mm copper base do they not? Its inset from the footprint of the main alu heatsink so isn't visible from above. I wasn't able to remove my SP20 heatsinks as they're aggressively glued to the PCBs and would have likely caused damage.

There is a copper base plate, but it is connected to the aluminum heat sink by a layer of thermal compound, without any mechanical binding, this defeated the purpose of that material
legendary
Activity: 1666
Merit: 1185
dogiecoin.com
Just found this post here:

The heat sink on SP20 is just entry level plain aluminum heat sink, they don't have huge surface area like those heat pipe heat sink with many thin aluminum fins, so the airflow must be very high to reduce the surface temperature

I just measured 20% fan speed noise at 1 meter away with my sound meter app, it is still 67db, and I personal feel is close to in car noise on highway (70+-db)

Would be interesting what hashspeed is reachable with these settings.

If the heatsinks are the same as the SP30/35, they've got a 4-5mm copper base do they not? Its inset from the footprint of the main alu heatsink so isn't visible from above. I wasn't able to remove my SP20 heatsinks as they're aggressively glued to the PCBs and would have likely caused damage.
full member
Activity: 122
Merit: 100
taxes are 19% plus 10 EUR handling fee for DHL.

Running sp20 in automode reaching as low as 10% with stock fan at 1,2 TH/s means 56 db(A).

if you replace the fan with a smaller PWM (means getting the right connector; use search or PM me) you can get it as quiet as slightly above 40 db within a cool room, but you have to underclock.
sr. member
Activity: 361
Merit: 250
I've been mining since the ATI 5870 days and the SP20s are far, far louder than any GPU fan IMO.

If you read on the unofficial Spondoolies thread (and I just checked my 5 out) and running around 1.2Th/s will draw around 650-700 from the wall. I think you can probably squeeze 2 units off one PSU, but I played it safe and ran 3 off 2 PSUs to allow for some headroom. When my new house is ready in a few months these units will go in the garage where noise/heat won't be a problem so I plan to crank them up to about 1.5Th/s
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
the s-5 is less polished then the sp20 I have been able to get them down to  58 db and mine at  freq 387 giving me 1271 gh .

or at 62 db and freq 412   1331gh.  but they are  a bit messy if you want to push them to overclock.
Pages:
Jump to: