Topic: Final fan mod to my s-5's (Read 3263 times)
bumped and locked
played with the s-5 some more
photos 2 delta's ,1 nouctua, 1 silverstone, the noutua and the silverstone run off the s-5 controller.
the delta's run off the small dc to dc convert at a constant 10 volts.
the whole mess give me 1331gh with 690-692 watts at 59-60 db about .518 watts per gh that counts all 4 fans.
fullsize
http://imageshack.com/a/img673/6377/zVjeBI.png
http://imageshack.com/a/img913/1238/jdPHaj.jpg
http://imageshack.com/a/img540/2419/OFLEEb.jpg
http://imageshack.com/a/img631/4706/VdhJWm.jpg
http://imageshack.com/a/img633/2356/Z8xbvv.jpg
http://imageshack.com/a/img901/5341/1IK9VV.jpg
http://imageshack.com/a/img673/1138/Vv7Rjl.jpg
http://imageshack.com/a/img540/5448/nTncg9.jpg
photos 2 delta's ,1 nouctua, 1 silverstone, the noutua and the silverstone run off the s-5 controller.
the delta's run off the small dc to dc convert at a constant 10 volts.
the whole mess give me 1331gh with 690-692 watts at 59-60 db about .518 watts per gh that counts all 4 fans.
fullsize
http://imageshack.com/a/img673/6377/zVjeBI.png
http://imageshack.com/a/img913/1238/jdPHaj.jpg
http://imageshack.com/a/img540/2419/OFLEEb.jpg
http://imageshack.com/a/img631/4706/VdhJWm.jpg
http://imageshack.com/a/img633/2356/Z8xbvv.jpg
http://imageshack.com/a/img901/5341/1IK9VV.jpg
http://imageshack.com/a/img673/1138/Vv7Rjl.jpg
http://imageshack.com/a/img540/5448/nTncg9.jpg
Regulators typically aren't designed to sink current, but to source it. The 5V buck regulators on our PSU boards won't work as sinks so getting 7V across the 12-5 on them doesn't work. An actual high-current rail on a PSU is probably more readily able to work with sinking currents, especially if the current is quite a bit lower than what it's capable of handling in output. It kinda depends on the topology of the regulator. I wouldn't trust most things to handle reverse currents near high enough to power an S5, even at 9V loads.
I've thought about doing the 36V chain also, but I don't have any string miners to test with. Four boards on a Prisma would do it, except you'd need level shifters for the serial lines, which could be problematic. S5 there's no trouble stacking, since the only IO is ethernet, which is an isolated line independent of local ground references. You couldn't put two boards of an S5 in series on the same controller without smoking the controller, but four whole machines should work fine.
We're working on a prototype regulator for adjusting voltages for string miners; hopefully we have some good numbers in the next week or two.
I've thought about doing the 36V chain also, but I don't have any string miners to test with. Four boards on a Prisma would do it, except you'd need level shifters for the serial lines, which could be problematic. S5 there's no trouble stacking, since the only IO is ethernet, which is an isolated line independent of local ground references. You couldn't put two boards of an S5 in series on the same controller without smoking the controller, but four whole machines should work fine.
We're working on a prototype regulator for adjusting voltages for string miners; hopefully we have some good numbers in the next week or two.
looking forward to results on that.
BTW this tiny little piece of gear I purchased on amazon is a very nice fan controller.
If I choose to lower freq to 300 and volts to 8> I would have a very quiet s-5 under 50 db running cool.
I figure this small dc to dc converter can run 6 stock s-5 fans.
Just to be clear Phillip, are you only applying the voltage regulator to the single stock S5 fan? This essentially removes the S5's control of fan speed and turns it into a manual setting on the voltage regulator? Any other fan changes involved?
Thanks for your exploration of this area.
yes but that controller can do about 6 fans. feed it 12 volts from your psu. I used the 4 pin molex cable. my molex wires can do 75 to 100 watts on the evga 1300 g2
basically slowing the stock fan to 5-10 volts allows a 1 at a time adjustment. it stays at the volts you pick. if 10 volts is too loud try 9.5 or 9 or 8. there is a small screw to adjust the volts with a flat blade tiny screwdriver..
you need to think of it as a manual fan control. what is nice is it is cheap and can handle a lot of fans.
I am getting bearable sound levels and 383 freq.
if you want really quiet feed it 5 volts and freq 275.
Regulators typically aren't designed to sink current, but to source it. The 5V buck regulators on our PSU boards won't work as sinks so getting 7V across the 12-5 on them doesn't work. An actual high-current rail on a PSU is probably more readily able to work with sinking currents, especially if the current is quite a bit lower than what it's capable of handling in output. It kinda depends on the topology of the regulator. I wouldn't trust most things to handle reverse currents near high enough to power an S5, even at 9V loads.
I've thought about doing the 36V chain also, but I don't have any string miners to test with. Four boards on a Prisma would do it, except you'd need level shifters for the serial lines, which could be problematic. S5 there's no trouble stacking, since the only IO is ethernet, which is an isolated line independent of local ground references. You couldn't put two boards of an S5 in series on the same controller without smoking the controller, but four whole machines should work fine.
We're working on a prototype regulator for adjusting voltages for string miners; hopefully we have some good numbers in the next week or two.
I've thought about doing the 36V chain also, but I don't have any string miners to test with. Four boards on a Prisma would do it, except you'd need level shifters for the serial lines, which could be problematic. S5 there's no trouble stacking, since the only IO is ethernet, which is an isolated line independent of local ground references. You couldn't put two boards of an S5 in series on the same controller without smoking the controller, but four whole machines should work fine.
We're working on a prototype regulator for adjusting voltages for string miners; hopefully we have some good numbers in the next week or two.
looking forward to results on that.
BTW this tiny little piece of gear I purchased on amazon is a very nice fan controller.
If I choose to lower freq to 300 and volts to 8> I would have a very quiet s-5 under 50 db running cool.
I figure this small dc to dc converter can run 6 stock s-5 fans.
Just to be clear Phillip, are you only applying the voltage regulator to the single stock S5 fan? This essentially removes the S5's control of fan speed and turns it into a manual setting on the voltage regulator? Any other fan changes involved?
Thanks for your exploration of this area.
Nice Mod! I will try this on my second S5.
As for now, I run 2x Silverstone SST-FHP141, both switched to performance mode running 1680rpm on each. Temps stay around 60c @ 343.75 frequency.
As for now, I run 2x Silverstone SST-FHP141, both switched to performance mode running 1680rpm on each. Temps stay around 60c @ 343.75 frequency.
Regulators typically aren't designed to sink current, but to source it. The 5V buck regulators on our PSU boards won't work as sinks so getting 7V across the 12-5 on them doesn't work. An actual high-current rail on a PSU is probably more readily able to work with sinking currents, especially if the current is quite a bit lower than what it's capable of handling in output. It kinda depends on the topology of the regulator. I wouldn't trust most things to handle reverse currents near high enough to power an S5, even at 9V loads.
I've thought about doing the 36V chain also, but I don't have any string miners to test with. Four boards on a Prisma would do it, except you'd need level shifters for the serial lines, which could be problematic. S5 there's no trouble stacking, since the only IO is ethernet, which is an isolated line independent of local ground references. You couldn't put two boards of an S5 in series on the same controller without smoking the controller, but four whole machines should work fine.
We're working on a prototype regulator for adjusting voltages for string miners; hopefully we have some good numbers in the next week or two.
I've thought about doing the 36V chain also, but I don't have any string miners to test with. Four boards on a Prisma would do it, except you'd need level shifters for the serial lines, which could be problematic. S5 there's no trouble stacking, since the only IO is ethernet, which is an isolated line independent of local ground references. You couldn't put two boards of an S5 in series on the same controller without smoking the controller, but four whole machines should work fine.
We're working on a prototype regulator for adjusting voltages for string miners; hopefully we have some good numbers in the next week or two.
looking forward to results on that.
BTW this tiny little piece of gear I purchased on amazon is a very nice fan controller.
If I choose to lower freq to 300 and volts to 8> I would have a very quiet s-5 under 50 db running cool.
I figure this small dc to dc converter can run 6 stock s-5 fans.
Regulators typically aren't designed to sink current, but to source it. The 5V buck regulators on our PSU boards won't work as sinks so getting 7V across the 12-5 on them doesn't work. An actual high-current rail on a PSU is probably more readily able to work with sinking currents, especially if the current is quite a bit lower than what it's capable of handling in output. It kinda depends on the topology of the regulator. I wouldn't trust most things to handle reverse currents near high enough to power an S5, even at 9V loads.
I've thought about doing the 36V chain also, but I don't have any string miners to test with. Four boards on a Prisma would do it, except you'd need level shifters for the serial lines, which could be problematic. S5 there's no trouble stacking, since the only IO is ethernet, which is an isolated line independent of local ground references. You couldn't put two boards of an S5 in series on the same controller without smoking the controller, but four whole machines should work fine.
We're working on a prototype regulator for adjusting voltages for string miners; hopefully we have some good numbers in the next week or two.
I've thought about doing the 36V chain also, but I don't have any string miners to test with. Four boards on a Prisma would do it, except you'd need level shifters for the serial lines, which could be problematic. S5 there's no trouble stacking, since the only IO is ethernet, which is an isolated line independent of local ground references. You couldn't put two boards of an S5 in series on the same controller without smoking the controller, but four whole machines should work fine.
We're working on a prototype regulator for adjusting voltages for string miners; hopefully we have some good numbers in the next week or two.
The regulator he's linked is a switching regulator that takes in 12VDC and only 12VDC, then uses a high-current PWM pulse and lowpass LC filter to get an adjustable DC voltage lower than 12V out of it at relatively high efficiency. It does not put any load on 5V, 3.3V or anything else.
Additionally, if you connect a load between 12V and 3V then draw 20A, the 3V will see a -20A not -80A.
With a 4-wire fan, you can also get speed control by pulsing the PWM line (usually the blue wire). It takes an open-collector (or open-drain if using FETs) pulse driver, nominally around 24KHz, and the duty cycle determines the speed using the fan's internal circuitry. Something like that can be wired up with a 555 timer or some comparators in not very many minutes. There's no real load on the PWM line (milliamps), so you can control basically as big a fans as you want.
Additionally, if you connect a load between 12V and 3V then draw 20A, the 3V will see a -20A not -80A.
With a 4-wire fan, you can also get speed control by pulsing the PWM line (usually the blue wire). It takes an open-collector (or open-drain if using FETs) pulse driver, nominally around 24KHz, and the duty cycle determines the speed using the fan's internal circuitry. Something like that can be wired up with a 555 timer or some comparators in not very many minutes. There's no real load on the PWM line (milliamps), so you can control basically as big a fans as you want.
Thanks for pointing out my error. I had this question because I guess that you could just use 3V as ground and get 9V input for S5 to reach super high efficiency less than 0.3W/GH, is that possible with such high load on 3V rail?
Another way is to chain 3x12V output together to reach 36V, and then divided them into 4 chains of ASICs on S5, in this way you will also get 9V for the rig
Unfortunately I have no S5 at hand to test with
Yeah I bumped the volts to 9.6 and the freq to 381
it runs just a little too hot, but the room is 86f.
1
30
381.25
67---------------Temp 67
2
30
381.25
61---------------Temp 61
Only 7 HW errors in 8 + hours
hash rate is 1261 at 655 watts.
I am going to open the window in that spare bedroom to drop temps a bit.
it runs just a little too hot, but the room is 86f.
1
30
381.25
67---------------Temp 67
2
30
381.25
61---------------Temp 61
Only 7 HW errors in 8 + hours
hash rate is 1261 at 655 watts.
I am going to open the window in that spare bedroom to drop temps a bit.
any more testing?
The regulator he's linked is a switching regulator that takes in 12VDC and only 12VDC, then uses a high-current PWM pulse and lowpass LC filter to get an adjustable DC voltage lower than 12V out of it at relatively high efficiency. It does not put any load on 5V, 3.3V or anything else.
Additionally, if you connect a load between 12V and 3V then draw 20A, the 3V will see a -20A not -80A.
With a 4-wire fan, you can also get speed control by pulsing the PWM line (usually the blue wire). It takes an open-collector (or open-drain if using FETs) pulse driver, nominally around 24KHz, and the duty cycle determines the speed using the fan's internal circuitry. Something like that can be wired up with a 555 timer or some comparators in not very many minutes. There's no real load on the PWM line (milliamps), so you can control basically as big a fans as you want.
Additionally, if you connect a load between 12V and 3V then draw 20A, the 3V will see a -20A not -80A.
With a 4-wire fan, you can also get speed control by pulsing the PWM line (usually the blue wire). It takes an open-collector (or open-drain if using FETs) pulse driver, nominally around 24KHz, and the duty cycle determines the speed using the fan's internal circuitry. Something like that can be wired up with a 555 timer or some comparators in not very many minutes. There's no real load on the PWM line (milliamps), so you can control basically as big a fans as you want.
this.
I also could have used the cpu power cable and fed in 2x 12 volt wires allowing for more then 100 watts of power.
I am taking photos in a few minutes.
four pin molex with 12 volt power in this wire may be 16 ga not 18 ga but 18 ga can run 4 delta fans or 4 stock fans.
that is the output end of the device . I will show the dc out on the next photo
dc out is 9.28 volts
here is dc in measure point
dc in is 12.16 volts
note room temp is 84 f and I have a low noise fan plugged into the oem fan power I can keep the dc converter very cool this way
this is what I use to test a lot of fans at one time. the power device goes from 1 to 15 watts. and I use two grounding bars to hook up 10 fans. the little part from amazon for 13 bucks can do the same as that power supply that cost 70 bucks. and since the amazon dc converter feeds off the psu it is 95% of a 90% psu or 85.5%
that 70 dollar power supply is 70-80 % with 6 deltas or 6 oem s-5 fans you will save 10 watts
The regulator he's linked is a switching regulator that takes in 12VDC and only 12VDC, then uses a high-current PWM pulse and lowpass LC filter to get an adjustable DC voltage lower than 12V out of it at relatively high efficiency. It does not put any load on 5V, 3.3V or anything else.
Additionally, if you connect a load between 12V and 3V then draw 20A, the 3V will see a -20A not -80A.
With a 4-wire fan, you can also get speed control by pulsing the PWM line (usually the blue wire). It takes an open-collector (or open-drain if using FETs) pulse driver, nominally around 24KHz, and the duty cycle determines the speed using the fan's internal circuitry. Something like that can be wired up with a 555 timer or some comparators in not very many minutes. There's no real load on the PWM line (milliamps), so you can control basically as big a fans as you want.
Additionally, if you connect a load between 12V and 3V then draw 20A, the 3V will see a -20A not -80A.
With a 4-wire fan, you can also get speed control by pulsing the PWM line (usually the blue wire). It takes an open-collector (or open-drain if using FETs) pulse driver, nominally around 24KHz, and the duty cycle determines the speed using the fan's internal circuitry. Something like that can be wired up with a 555 timer or some comparators in not very many minutes. There's no real load on the PWM line (milliamps), so you can control basically as big a fans as you want.
Are you sure if this kind of wiring would not disturb the output of the whole 12V rail since it is connected to a 5V rail? I think 5V rail and 12V rail have very different current
the molex four pin connectors on my evga can do 5, 7 or 12 volts.
If I use the 7 volt the two fans cool well up to freq 343.
If I use the 12 volt wires from the molex and use the amazon converter to drop power from 12v to 9.5v the 2 fans cool well up to freq 381.
These converters can take 12 volts from your psu and drop the volts from 10 to 5 and anything in between. Far better then a fan controller in that they will handle lot of fans.
http://www.amazon.com/gp/product/B00OT58TL4/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1
I will photo the hookup of it on sat. I need some batteries for the camera.
Would it negatively affect the PSU since now it is under a strange load on different rails? Maybe for a fan it is ok, but will that introduce extra load for the rail of lower voltage?
Normally the current flows from 12V/5V/3V rail to the ground, but now the current flows from 12V back into 3V rail. If you connect some heavy load between 12V and 3V, and 12V is loaded by 20A, then 3V rail will have a negative current of 80A, is that safe? Usually PSU provide very little current for 3V rail, but that is for positive load
Are you sure if this kind of wiring would not disturb the output of the whole 12V rail since it is connected to a 5V rail? I think 5V rail and 12V rail have very different current
the molex four pin connectors on my evga can do 5, 7 or 12 volts.
If I use the 7 volt the two fans cool well up to freq 343.
If I use the 12 volt wires from the molex and use the amazon converter to drop power from 12v to 9.5v the 2 fans cool well up to freq 381.
These converters can take 12 volts from your psu and drop the volts from 10 to 5 and anything in between. Far better then a fan controller in that they will handle lot of fans.
http://www.amazon.com/gp/product/B00OT58TL4/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1
I will photo the hookup of it on sat. I need some batteries for the camera.
Are you sure if this kind of wiring would not disturb the output of the whole 12V rail since it is connected to a 5V rail? I think 5V rail and 12V rail have very different current
Okay I had a lot of people ask me about a fan controller for the s-5
I came up with a great cheap fan controller able to run 4 stock monster loud fans.
this handles 100 watts. even more with a fan on it.
I feed this with a four pin molex cable directly from my evga psu. that molex will feed 12 volts to this and the cable can handle 75-90 watts.
item is only 12 bucks if you are willing to cut the wires on the stock fan it will run it quietly.
http://www.amazon.com/gp/product/B00OT58TL4/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1
I dialed down the power to 9.5 volts and I am running 2 delta fans in push pull.
gear is clocked to freq 381
I came up with a great cheap fan controller able to run 4 stock monster loud fans.
this handles 100 watts. even more with a fan on it.
I feed this with a four pin molex cable directly from my evga psu. that molex will feed 12 volts to this and the cable can handle 75-90 watts.
item is only 12 bucks if you are willing to cut the wires on the stock fan it will run it quietly.
http://www.amazon.com/gp/product/B00OT58TL4/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1
I dialed down the power to 9.5 volts and I am running 2 delta fans in push pull.
gear is clocked to freq 381
after some more time I bumped to freq 343
pretty good just a little bit warm---------------- if you ran this in a cooler room you would get better temps. --------still db of fans are 53 and not a nasty pitch mod was under 20 bucks.
pretty good just a little bit warm---------------- if you ran this in a cooler room you would get better temps. --------still db of fans are 53 and not a nasty pitch mod was under 20 bucks.
space----------
So I get about 53db I run the gear at freq 331 hash 1100
So I get about 53db I run the gear at freq 331 hash 1100
space
more space
connect to fans
connect to fans
Jump to: