My Asrock H110 Pro BTC+ caught fire

[sylabis]

For maximum safety, when you have multiple PSU for 1 rig, you have to connect all motherboard connectors, HDD, some GC and ALL risers on first PSU.
Secondary PSU should only be use to power graphic cards. If you split risers between PSU, you still have a risk...

No. A GPU and it’s riser should be on the same PSU.

No, sorry, everything related to motherboard must share the same ground. By plugging risers on different PSU you have a chance to join 2 different ground and make your motherboard break down. In worst case it can start fire. Plugging only graphic cards on a second PSU won't make any problem.
I won't take the risk to see my house burn just for bad plug on a rig...

I agree. The PCI-E specification also clearly states the x16 PCI-E slot can be powered by a different PSU rail than the VGA 6/8-Pin aux power on the card. That's how I setup my triple and dual PSU rigs. The same PSU that powers the motherboard also powers all the risers. The other PSU's are only for the VGA power connectors on the cards. If there is more than one 6/8-PIN power connector on the card, you MUST make sure they are powered by the same PSU rail.


https://image.ibb.co/fVt8hx/PCI_E_Specification.png

All I have read on these forums is power riser+GPU should be on same PSU.

If you are running dual PSUs then PSU(1) that powers the motherboard should also power all the risers. The other PSU(2) should only power GPU 6pin/8pin. This configuration ensures both PSUs are properly grounded.

If you connect an additional power connector eg Molex, from PSU(2) directly to any part of the motherboard then expect things to end badly.

[1714854920]

1714854920

[1714854920]

1714854920

[1714854920]

1714854920

[1714854920]

1714854920

[1714854920]

1714854920

[1714854920]

1714854920

[ggbtctalk000]

There is no telling what is going to happen to which.
- I had one 1200w thermaltake burned when using pcie cables only no sata and still power cable burnt. It caused all GPU-s fans to max and I was scratching my head wtf has happened till i noticed the burn mark.

- another msi z270 sata connector got extremely hot and started smoking.

[ViperGuy]

If you are running dual PSUs then PSU(1) that powers the motherboard should also power all the risers. The other PSU(2) should only power GPU 6pin/8pin. This configuration ensures both PSUs are properly grounded.

You must connect the SAME psu that powers the riser to the same GPU 6/8 pin 12V connector. The purpose of the power to the riser is to give a current boost to the PCIE 12V bus feeding the GPU (up to 54 watts), not to feed power back to the MB.  If you have been doing it the way you described, only luck has been on your side. I know others do it this way as well, but this is actually wrong and is one of the reasons people have stated they can only run 2 risers on the same molex cable bus before it gets hot. It's overheating because the voltage potential between the molex 12v bus on one PSU and the 6/8pin 12V on another PSU is out of phase and causing a balancing current draw. But the part you said about not using the molex on the MB is correct.

Likely the reason why the OP burned up his MB is because he used the molex connector on the MB in the first place. You can't do this while using more than one PSU with the risers or GPUs. The reason is because of voltage phases with switching power supplies. He also may be on to something with one of the GPU PSU's being off and the MB trying to power the GPU alone, but this is a design flaw with all these Molex motherboards. There was never a provision for additional power to be supplied the way they are doing now. In normal circumstances, without the molex power being there, his MB would have simply shut down from an overload and prevented this hazard. but because that molex connector is running 12V on it's own bus, bypassing other circuits and thermal overload protections, it just ran away with the current. His PSU should have also prevented this from happening but the molex bus is allowed to do it's own thing and doesn't have the overload protections that the 24-pin bus has in place.

I'm an electrical engineer, and I have yet to understand how the MB manufacturers have got this so wrong. You can NOT use one PSU to power the risers and another to power the 6/8 pin GPU connectors with switching power supplies. This will create a potential delta of 12 along the same bus if and when one PSU is out of phase with the other on the same circuit. I don't think most people understand how a switching PSU works. It's basically a square wave cleaned up, but under heavy load it still shows itself as a square wave. For argument's sake, think of one PSU at a +12V peak for 60 cycles (an example, but it could be higher) and another at anything less than 12V or even zero, for 60 cycles on the same circuit. Whatever the difference is, the phase will create a voltage potential that shorts the circuit or the power bus in this case.

Do NOT use the molex connectors on the ASRock or Gigabyte mining boards. It is not necessary and is actually dangerous when using more than one PSU. It's a design flaw to even offer them. You would only ever need to use the molex connector if you had one single insane PSU run all the GPU cards and didn't use powered risers.  I have been running both the ASRock and Gigabyte mining boards with 12 cards and have never used the molex connector on the MB.

[Vann]

In a multiple PSU setup the reason you want to connect the risers to the the same PSU as the motherboard is because all devices connected to the motherboard MUST share the same power source and ground so the voltage can be regulated according to a single source.

https://bitcointalk.org/index.php?topic=1843586.msg18351224#msg18351224

The problem with powering more than two risers per molex strand is 18 AWG wire is only rated for 10 A max for single core wire, which is 120 W at 12 V and much less for multi stand wire. Using more than two risers per PSU cable strand WILL exceed the rating of 18 AWG wire, causing the wire to heat up and is dangerous.

https://www.engineeringtoolbox.com/wire-gauges-d_419.html

[mindrust]

That's why people advise us to build 1PSU rigs if it is possible.

2-PSU rigs can do alright as long as you dial them right. Like OP, sometimes you don't.

[ViperGuy]

In a multiple PSU setup the reason you want to connect the risers to the the same PSU as the motherboard is because all devices connected to the motherboard MUST share the same power source and ground so the voltage can be regulated according to a single source.

https://bitcointalk.org/index.php?topic=1843586.msg18351224#msg18351224

The problem with powering more than two risers per molex strand is 18 AWG wire is only rated for 10 A max for single core wire, which is 120 W at 12 V and much less for multi stand wire. Using more than two risers per PSU cable strand WILL exceed the rating of 18 AWG wire, causing the wire to heat up and is dangerous.

https://www.engineeringtoolbox.com/wire-gauges-d_419.html

Well, I disagree with this. It is more important to run the same PSU powering the riser and the GPU than it is to worry about MB to riser ground. The PSU, Riser, and GPU ground need to be tied together but the MB does not need to be on the same loop when using a powered riser. The reason is because the MB power is not powering the riser or GPU at this point. The USB cable is transferring data, not delivering 12v power, therefore common ground is less consequential.

On a side note, I have run 4 risers (Titan XPs) on one 18 gauge Molex and the current draw was not even 10A as long as the same PSU powered the GPU 6/8 pin.  I'm not saying it's best practice, but as soon as I separate the PSU that powers the riser from the one that powers the GPU, the current draw goes up and the Molex bus gets hot. This indicates a voltage potential difference on that bus, creating a greater power draw on that Molex bus. If the riser power is on the same bus, most of the power will be supplied via the 6/8 pin connector. While PCIE standards allow up to 75W via 12v and 3.3v combined power, most cards are designed to only draw up to 35 watts via 12V and the rest comes from the 6/8 pin bus.

[sylabis]

If you are running dual PSUs then PSU(1) that powers the motherboard should also power all the risers. The other PSU(2) should only power GPU 6pin/8pin. This configuration ensures both PSUs are properly grounded.

You must connect the SAME psu that powers the riser to the same GPU 6/8 pin 12V connector. The purpose of the power to the riser is to give a current boost to the PCIE 12V bus feeding the GPU (up to 54 watts), not to feed power back to the MB.  If you have been doing it the way you described, only luck has been on your side. I know others do it this way as well, but this is actually wrong and is one of the reasons people have stated they can only run 2 risers on the same molex cable bus before it gets hot. It's overheating because the voltage potential between the molex 12v bus on one PSU and the 6/8pin 12V on another PSU is out of phase and causing a balancing current draw. But the part you said about not using the molex on the MB is correct.

Likely the reason why the OP burned up his MB is because he used the molex connector on the MB in the first place. You can't do this while using more than one PSU with the risers or GPUs. The reason is because of voltage phases with switching power supplies. He also may be on to something with one of the GPU PSU's being off and the MB trying to power the GPU alone, but this is a design flaw with all these Molex motherboards. There was never a provision for additional power to be supplied the way they are doing now. In normal circumstances, without the molex power being there, his MB would have simply shut down from an overload and prevented this hazard. but because that molex connector is running 12V on it's own bus, bypassing other circuits and thermal overload protections, it just ran away with the current. His PSU should have also prevented this from happening but the molex bus is allowed to do it's own thing and doesn't have the overload protections that the 24-pin bus has in place.

I'm an electrical engineer, and I have yet to understand how the MB manufacturers have got this so wrong. You can NOT use one PSU to power the risers and another to power the 6/8 pin GPU connectors with switching power supplies. This will create a potential delta of 12 along the same bus if and when one PSU is out of phase with the other on the same circuit. I don't think most people understand how a switching PSU works. It's basically a square wave cleaned up, but under heavy load it still shows itself as a square wave. For argument's sake, think of one PSU at a +12V peak for 60 cycles (an example, but it could be higher) and another at anything less than 12V or even zero, for 60 cycles on the same circuit. Whatever the difference is, the phase will create a voltage potential that shorts the circuit or the power bus in this case.

Do NOT use the molex connectors on the ASRock or Gigabyte mining boards. It is not necessary and is actually dangerous when using more than one PSU. It's a design flaw to even offer them. You would only ever need to use the molex connector if you had one single insane PSU run all the GPU cards and didn't use powered risers.  I have been running both the ASRock and Gigabyte mining boards with 12 cards and have never used the molex connector on the MB.

I have read a popular post before from an "electrical engineer" making the same claims as you. The flaw is that most risers still draw power from the MB, if they didn't then you would be correct. Have you ever tested this out? If OP had the PSU powering the MB connected to all the risers, and the MB molex coming from the same PSU that powered the MB 24pin/12v.  then he never would have ran into a issue.  

I don't think MB manufacturers have got this wrong, I think you've got it wrong. Some boards will not post with more than 3-4x GPUs without a molex plugged in.  It's only a problem if the molex you are using is coming from the PSU that's not plugged into the motherboards 24pin/12v connector.  

[ViperGuy]

I have read a popular post before from an "electrical engineer" making the same claims as you. The flaw is that most risers still draw power from the MB, if they didn't then you would be correct. Have you ever tested this out? If OP had the PSU powering the MB connected to all the risers, and the MB molex coming from the same PSU that powered the MB 24pin/12v.  then he never would have ran into a issue.  

I don't think MB manufacturers have got this wrong, I think you've got it wrong. Some boards will not post with more than 3-4x GPUs without a molex plugged in.  It's only a problem if the molex you are using is coming from the PSU that's not plugged into the motherboards 24pin/12v connector.  

I've said it before and I'll say it again, you do NOT need to use the Molex connector on any motherboard when powered risers are used. Doing so is dangerous as seen in this case. Built in safety nets are bypassed when using that out-of-spec setup. There is very little power draw through the USB riser cable. It's primary function is to pass data, not power. If there is more than a couple watts passing through, then the powered riser is not doing it's job and is junk. The molex connector design on MB's was only put in place to facilitate multi GPU's directly connected in the slots for Crossfire/SLI configurations on Gaming MB's that had two or more cards running in the full length slots. It is an unnecessary design to have it on a mining board with 1x slot when powered risers are used. Molex connectors weren't even designed to exceed 36 watts so it's pointless to even have one molex boost 12 PCIE slots, it would theoretically only supply 3 watts per slot anyway. It is a design flaw that the Chinese just winged at and completely ignored the power limits and design standards of the components used. This fire hazard could have been avoided if the molex connector simply had not been used. The MB would have likely just shut down in the OP's situation instead of running current through an unmonitored bus. I would be more worried about having a separate switching PSU power the GPU PCIE 12V bus than the 6/8 pin 12V PSU. There's two high current differentials waiting to create a disaster right within the GPU itself.

[Set Ready Go]

Copied from different forum, nonetheless thoughtful reading.



"am very concerned about incorrect and misleading information I am seeing posted on forums and the Internet in general about electrical specification for GPU mining. There are a lot of new miners building rigs and the AMD shortage is pushing people to use 1080 ti's and other Nvidia cards (up to 250 watt cards stock). You need to accept that there are individuals on these forums that make a profit off of your inexperience, and want to sell you stuff. That is not a bad thing by itself, but if facts are distorted and people make incorrect decisions, bad things can and will happen.

I am an electrical engineer with 30 years experience and I will simply present the facts in this post. I sell no mining equipment and have no motives other than to keep people safe in their mining endeavors. You need to make up your own mind what you think is safe for you, your rigs, and your home or facility.

1) New risers have NO power connection to the mobo, only data via the USB cable. They make their own 3.3V on the riser from the 12V rail. You NEED to test your risers to confirm this. Ultimately you are responsible for building your rig in a safe and reliable way. Not me, or any other person on this forum. Having said that, each GPU should be powered from one PSU (both riser and 8 or 6 pin connector). IF you power the riser from the mobo PSU and the 6 or 8 pin connector from a different PSU you can potentially tie together the 12V rail from two PSU's. They will fight to regulate the 12V rail if they are just a little bit off. No one, especially you, wins that battle. This can cause melted cables and fires. Others on this forum insist you must power the riser from the same PSU as the mobo. This is not correct and ties back to very early Bitcoin mining when the risers had power connections from the mobo. Just because you wire it wrong and it works does not mean you are OK. Most GPU vendors have circuitry to isolate the PCIe 12V rail from the 6 and 8 pin 12V rail.... if you have it wrong, and you find one that does not, you will not be happy.

2) SATA connectors and therefore SATA risers are rated for 4.5 amps on the 12V rail. The PCIe 3 specification allows a GPU to draw 5.5 amps from the PCIe 12V bus. A 1080 ti GPU draws 4.4 amps at 80% power (link below), and more than 4.5 amps at 100% power. Therefore a 1080 ti at full power draws more than 4.5 amps and you are drawing too much current through a SATA connector. Anyone who says that its fine, and perfectly safe to use SATA cables on risers with more than 4.5amps draw, has no idea what they are talking about, and absolutely no electrical engineering experience.



The older 4 pin molex connectors are rated at 11 amps for the 12V rail. Therefore, do NOT use the molex to SATA adapters. Run a 4 pin molex cable from your PSU directly to your riser get rid of the SATA adapter. Even if you power two 1080 ti's, and they draw 5.5 amps each (the maximum), you are still within the rating of the 4 pin molex cable and connector. Its simply not worth the risk to overpower a connector when it is so easy to use the right connector.

3) In the US an 8 GPU 1080 ti rig will draw about 90% load on a 20 amp circuit. In the US 80% maximum continuous load is code, and can deal with inrush current. Circuit breakers have one purpose: to keep the wires in the wall from catching fire from too much current. They could care less what is plugged into the line, if its on fire, or not, as long as it does not draw more current that the wires can handle. So circuit breakers do NOT protect your rig or your house from a fire, they are not designed to do so. Keep total current draw on a rig to 80% of the breaker. If you want to build big rigs with 250 watt cards run 20 or 30 amp dedicated lines. Don't try to plug different PSU's into separate residential 15 amp lines. Other countries with 220V mains line voltage is different.

A 7 GPU 1080 ti rig at full power draws 79% continuous load, perfect for a 20 amp line.

4) plugging multiple PSU's from the same rig into different circuits / breakers is a bad idea. If one breaker blows and its not the master PSU, you have half a rig powered. While not normally an issue, it could damage your cards, depending on how you have your rigs wired.

5) Molded SATA connectors (like the ones sold with risers) are known to have the internal crimp connections that can move in the molding process and can be close enough to each other to eventually short (even at normal SATA loads, not over rated). Heat will help to facilitate this. While this can take months or years, once you get a short, the SATA connector WILL burst into flames. Now, you have a fire. Don't use them, especially on 1080 ti's, just use the 4 pin molex connectors and cables from your PSU, or buy non molded SATA connectors for use with GPU's that draw 4.5amps or less.



6) Just because something works on one GPU from one vendor does not mean in will work on another. I do use SATA cables to power my Nitro Sapphire RX480 rigs (they draw less than 4.5 amps on the 12V PCIe bus, one SATA cable / GPU, non molded adapters). However, I do NOT use SATA on the XFX RX480 cards. The XFX cards draw more power from the PCIe 12V bus and will get warm SATA connectors. So each vendor has different power draws and you need to be very careful. If you do not know how to measure, just use the 11amp molex cable and don't worry. I have had a SATA connector get very hot on an XFX card (you could smell it).

Ok so that is all I have for now. I expect some backlash from this post, but if I can help one person avoid a dangerous situation I will be more than happy."

[qqqua]

there is no need to connect aux power if you are using powered usb riser, i have this mobo on my rigs

[Vann]

There IS an electrical connection from the motherboard to the riser, period. An 'electrical engineer' that says otherwise is an idiot. As I said, if you measure the outer pins of the USB connector coming from the motherboard PCI-E slot connector with a multimeter while the motherboard is on you will see there is 3.3V coming from the motherboard PCI-E slot through the USB cable to the riser. If you plug in the USB cable coming from the x1 riser connector to the riser and measure the voltage on the two pins on the bottom of the riser circled in yellow, you will see this 3.3V IS passed to the riser. If you look at the traces on the bottom of a Version 006/008 riser you will see the traces from the supplemental power connector and the traces from the USB connector are connected, which is why you should only power the risers with the same PSU connected to the motherboard, so the voltages can be regulated according to a single source. As the PCI-E specification clearly states, the power source to the x16 PCI-E slot DOES NOT have to be on the same rail as the power source to the 6/8 pins connector on the card.



How much power is drawn from the PCI-E slot depends on how the card is configured by the manufacturer, not how you set up your PSU's. A card mining WILL pull between 35-55 W through the riser, depending on how it's configured and how much power is needed. Dual mining also causes more power to be pulled from the riser.

[ViperGuy]

Copied from different forum, nonetheless thoughtful reading.

I agree with most of what that engineer said. You can confirm his findings with a voltmeter or by simply following the tracings on a powered riser.

There IS an electrical connection from the motherboard to the riser, period...

How much power is drawn from the PCI-E slot depends on how the card is configured by the manufacturer, not how you set up your PSU's. A card mining WILL pull between 35-55 W through the riser, depending on how it's configured and how much power is needed. Dual mining also causes more power to be pulled from the riser.

You are partially correct. Yes, the GPU uses power through the RISER, but not through the MB to the riser. I am not disputing that the GPU uses 35w or even 55w through the PCIE slot on the riser, I'm stating that the electrical connection between the MB PCIE slot to the riser is not 35w or even 10w. The 3.3v connection is a low current bus and isn't supplying much power to the riser. The argument here is with the 12V line, not the 3.3v. The 3.3v is isolated from the PSU through multi-phase regulators so this can be shared across PSUs. The MB is not powering the riser, proof of this is if you disconnect the 12V PCIE power to a quality riser, you lose the card connection. The USB cable is a shielded data cable with 24-22awg wires and even if it were a bigger gauge the specs only allow for a 2.4 amp maximum draw at 5V (12 Watt for 5v, 8w max for 3.3v) without attenuation issues to the data lines.  If the MB were powering the risers, you would have 420 - 660 Watts running through the MB in a 12 GPU setup.  My MB's are drawing less than 90 watts total so I know this is not the case. If your MB PCIE bus is getting pulled for power, you have a problem with inferior risers.  Good risers electrically isolate the 12V from the MB, and ties the PCIE 12V power to the 16X slot using only the 12v PCIE connector on the riser.  They also use a voltage regulator from the 12V PCIE power to convert to 3.3v. Your pic doesn't show the top side of the riser where this is happening.

Additionally, unlike 12V power which comes direct from the switching PSU, the 3.3v line is isolated from the PSU because it is created from multi-phase voltage regulators on the MB. It is safe to have the 3.3v come from two different sources because the switching part has already been isolated and converted in the circuit. You don't need to worry about voltage potential differences because they would only be in the 0.1v or less range. It would be unlikely that large power draw through the 3.3v would occur unless it was a complete short at the riser from a bad circuit design.

[Vann]

I didn't say the riser is pulling 12V from the motherboard, that's not the point. As I showed, it's the fact that the riser IS sharing an electrical connection with the motherboard and all electrical connections to the motherboard must come from the same power source to avoid conflicts in regulating the voltages from having different power sources. PSU's also supply 3.3V to the motherboard directly. There is no need for the motherboard to create it.

[ViperGuy]

I didn't say the riser is pulling 12V from the motherboard, that's not the point. As I showed, it's the fact that the riser IS sharing an electrical connection with the motherboard and all electrical connections to the motherboard must come from the same power source to avoid conflicts in regulating the voltages from having different power sources.

Ok, fair enough. But the 3.3v is created from the 12V through a regulator on the top side of the v008 board. Even if this were electrically shared with the 3.3v bus on the MB, it is regulated and isolated from the Riser PSU via the 3.3v regulator on the riser board. The riser does not use 3.3 from the PSU, in fact it converts switching 12V to regulated 3.3v. This isolates the riser PSU from the MB 3.3v and PSU, therefore there is not harm in having a separate PSU on the riser from the MB. The problem is not with sharing 3.3v power from two different components on two different PSU's, that is actually safe because of the regulators and low current parameters, and the power is no longer switching. Small potential differences of 0.1v is not going to short out or ground the 3.3v circuit. The real problem is from sharing direct 12V Switching Power and this is where the danger and the fires start.

You absolutely 100%, without a doubt, want to power your Riser with the same PSU powering that GPU 6/8 pin 12V, this is a fact. If you don't, this is where the danger arises. Internally the GPU is drawing 12v Switched power from two sources if your riser is on a different PSU, and that's a big red flag.  What's the point in worrying about the isolated, regulated, low current 3.3v MB to Riser connection when you are mixing pure high current 12V switched power right at the GPU? If one switched PSU happens to go 180 degrees out of phase from the other switched PSU on the same 12V circuit, bye-bye. This is likely what happened with the OP. We don't really know how he was setup, likely two PSU's on the same MB for all we know, but had he simply left the Molex disconnected on the MB, he would not have fried it.

I don't know who decided this was the rule to use the MB and risers on the same PSU, but electrically it's wrong to mix PSU's with the Riser/GPU combo, and it's electrically safe to have the MB powered by it's own PSU separate from powered risers when switched 12V is not involved.  

Friends don't let friends put Risers and GPUs on different power supplies, remember this!

Part of my past applied studies were in voltage regulated circuit design, I'm not speaking out of my arse here, I know what I'm talking about.

[Vann]

The voltage regulator on a V006/008 riser is there to stabilize the line and prevent spikes, not generate 3.3V. Why would the traces of the USB connector where the 3.3V is passed from the motherboard on the V006/008 risers be connected to the pins leading to the PCI-E slot on the riser if the riser created it's own 3.3V?

https://en.wikipedia.org/wiki/PCI_Express

Powering the riser with a different PSU as the motherboard causes the 12V line to come from one power source and the 3.3V line and the grounds on the riser x16 PCI-E slot to come from mixed power sources, which is the problem. Sorry, but anyone that says it's safe to power four risers off of one molex strand, clearly doesn't know what they are talking about when it comes to mining.

[BitBustah]

that you are connected auxilliary  power from a different psu

My first concern too!

[teedeexyz]

Copied from different forum, nonetheless thoughtful reading.



"am very concerned about incorrect and misleading information I am seeing posted on forums and the Internet in general about electrical specification for GPU mining. There are a lot of new miners building rigs and the AMD shortage is pushing people to use 1080 ti's and other Nvidia cards (up to 250 watt cards stock). You need to accept that there are individuals on these forums that make a profit off of your inexperience, and want to sell you stuff. That is not a bad thing by itself, but if facts are distorted and people make incorrect decisions, bad things can and will happen.

I am an electrical engineer with 30 years experience and I will simply present the facts in this post. I sell no mining equipment and have no motives other than to keep people safe in their mining endeavors. You need to make up your own mind what you think is safe for you, your rigs, and your home or facility.

1) New risers have NO power connection to the mobo, only data via the USB cable. They make their own 3.3V on the riser from the 12V rail. You NEED to test your risers to confirm this. Ultimately you are responsible for building your rig in a safe and reliable way. Not me, or any other person on this forum. Having said that, each GPU should be powered from one PSU (both riser and 8 or 6 pin connector). IF you power the riser from the mobo PSU and the 6 or 8 pin connector from a different PSU you can potentially tie together the 12V rail from two PSU's. They will fight to regulate the 12V rail if they are just a little bit off. No one, especially you, wins that battle. This can cause melted cables and fires. Others on this forum insist you must power the riser from the same PSU as the mobo. This is not correct and ties back to very early Bitcoin mining when the risers had power connections from the mobo. Just because you wire it wrong and it works does not mean you are OK. Most GPU vendors have circuitry to isolate the PCIe 12V rail from the 6 and 8 pin 12V rail.... if you have it wrong, and you find one that does not, you will not be happy.

2) SATA connectors and therefore SATA risers are rated for 4.5 amps on the 12V rail. The PCIe 3 specification allows a GPU to draw 5.5 amps from the PCIe 12V bus. A 1080 ti GPU draws 4.4 amps at 80% power (link below), and more than 4.5 amps at 100% power. Therefore a 1080 ti at full power draws more than 4.5 amps and you are drawing too much current through a SATA connector. Anyone who says that its fine, and perfectly safe to use SATA cables on risers with more than 4.5amps draw, has no idea what they are talking about, and absolutely no electrical engineering experience.



The older 4 pin molex connectors are rated at 11 amps for the 12V rail. Therefore, do NOT use the molex to SATA adapters. Run a 4 pin molex cable from your PSU directly to your riser get rid of the SATA adapter. Even if you power two 1080 ti's, and they draw 5.5 amps each (the maximum), you are still within the rating of the 4 pin molex cable and connector. Its simply not worth the risk to overpower a connector when it is so easy to use the right connector.

3) In the US an 8 GPU 1080 ti rig will draw about 90% load on a 20 amp circuit. In the US 80% maximum continuous load is code, and can deal with inrush current. Circuit breakers have one purpose: to keep the wires in the wall from catching fire from too much current. They could care less what is plugged into the line, if its on fire, or not, as long as it does not draw more current that the wires can handle. So circuit breakers do NOT protect your rig or your house from a fire, they are not designed to do so. Keep total current draw on a rig to 80% of the breaker. If you want to build big rigs with 250 watt cards run 20 or 30 amp dedicated lines. Don't try to plug different PSU's into separate residential 15 amp lines. Other countries with 220V mains line voltage is different.

A 7 GPU 1080 ti rig at full power draws 79% continuous load, perfect for a 20 amp line.

4) plugging multiple PSU's from the same rig into different circuits / breakers is a bad idea. If one breaker blows and its not the master PSU, you have half a rig powered. While not normally an issue, it could damage your cards, depending on how you have your rigs wired.

5) Molded SATA connectors (like the ones sold with risers) are known to have the internal crimp connections that can move in the molding process and can be close enough to each other to eventually short (even at normal SATA loads, not over rated). Heat will help to facilitate this. While this can take months or years, once you get a short, the SATA connector WILL burst into flames. Now, you have a fire. Don't use them, especially on 1080 ti's, just use the 4 pin molex connectors and cables from your PSU, or buy non molded SATA connectors for use with GPU's that draw 4.5amps or less.



6) Just because something works on one GPU from one vendor does not mean in will work on another. I do use SATA cables to power my Nitro Sapphire RX480 rigs (they draw less than 4.5 amps on the 12V PCIe bus, one SATA cable / GPU, non molded adapters). However, I do NOT use SATA on the XFX RX480 cards. The XFX cards draw more power from the PCIe 12V bus and will get warm SATA connectors. So each vendor has different power draws and you need to be very careful. If you do not know how to measure, just use the 11amp molex cable and don't worry. I have had a SATA connector get very hot on an XFX card (you could smell it).

Ok so that is all I have for now. I expect some backlash from this post, but if I can help one person avoid a dangerous situation I will be more than happy."

I'm using RX570s and 1080tis. Each power line coming from the PSU to the risers is connected to 3 risers using a SATA power cable. Some of the risers use molex connections (using SATA => molex adapter), some (moulded) SATA connections. Does this mean I'm also overloading these lines? O.o
Also: I'm European, so 230V here.

Grounds are connected through Add2PSU btw.

[BitBustah]

I'm using RX570s and 1080tis. Each power line coming from the PSU to the risers is connected to 3 risers using a SATA power cable. Some of the risers use molex connections (using SATA => molex adapter), some (moulded) SATA connections. Does this mean I'm also overloading these lines? O.o
Also: I'm European, so 230V here.

Grounds are connected through Add2PSU btw.

Same way here. I tried to avoid that, but ended up doing it too. I must say I kept feeling those connectors at least 3 times a day to see if they would not heat up.

[ViperGuy]

The voltage regulator on a V006/008 riser is there to stabilize the line and prevent spikes, not generate 3.3V. Why would the traces of the USB connector where the 3.3V is passed from the motherboard on the V006/008 risers be connected to the pins leading to the PCI-E slot on the riser if the riser created it's own 3.3V?

https://en.wikipedia.org/wiki/PCI_Express

Powering the riser with a different PSU as the motherboard causes the 12V line to come from one power source and the 3.3V line and the grounds on the riser x16 PCI-E slot to come from mixed power sources, which is the problem. Sorry, but anyone that says it's safe to power four risers off of one molex strand, clearly doesn't know what they are talking about when it comes to mining.

I said those that cannot run more than two Molex connectors on the same bus without getting hot are experiencing a switching PSU phase short. I did not say I recommend it. You send me a link to a generic wiki page, as if the Chinese even bother to follow PCIE protocols when designing risers.

You sir clearly don't understand how a switching PSUs operate, nor do you understand how the riser operates. The 3.3v is generated from the 12V source. You can clearly following the tracings from the 12v source and they end up at the 3.3v regulator, coil, and capacitor circuit. If the 3.3v was not produced from the 12v then what is the coil there for??? It's clearly a 12v to 3.3v regulated circuit. Heck, the number 3.3 is clearly stamped on the regulator on v006 boards.

The 3.3v bus isn't of consequence, the 12v bus is. The 3.3v is regulated and no longer has a switched phase, therefore it's isolated from the switched PSU and doesn't get affected by the MB PSU. You are focusing on a fly and ignoring the elephant. The 12v switched riser will eventually go 180 out of phase with the GPU PSU and will fry something. You have two engineers tell you otherwise and you still argue something you CLEARLY know little about.  Go ahead and burn your house down.

[Vann]

Did some more research on the right way to configure a dual PSU setup. This Biostar 12 GPU dual PSU mining setup guide states:

http://www.biostar.com.tw/app/kr/event/crypto_mining/page4_2.htm

**The power cable for riser card and power cable for graphics card must come from same power supply uint.

Given Biostar makes mining motherboards, USB riser kits, GPU's and they also have a financial incentive not to warranty thiose products, I believe they have an objective view on how to setup dual power supplies correctly. So I accept powering the riser and GPU with the same PSU is the correct way.

I just rewired the risers and cards on my rigs so that the riser and VGA power connector on the cards are on the same PSU. The power consumptin from both type of setups is the same. The same Biostar mining guide also says you should only use a max of 2 risers per PSU power cable.

*Each power cable from power supply is limited to be connected to 2 graphics cards or 2 PCI-Ex16 riser cards.

On a side note I want to say that I ran my dual and triple PSU rigs with all the risers connected to PSU that powers the motherboard and the secondary PSU only powering the VGA 6/8 pin power under a heavy dual mining load 24/7 for almost a year and nothing bad came from it and my rigs are very stable.