MP 1,1-5,1 about PowerLink with cMP/RX5700XT/RX6800 power question

[ddhhddhh2]

Hello everyone, I have a question about cMP + EVGA PowerLink. Yes, I understand, I've done my research and indeed found many recommendations. However, one thing still isn't clear to me. I noticed that many demonstrations simply connect the two mini power supplies from the motherboard to the EVGA PowerLink, and then install the PowerLink on the graphics card, which looks very elegant and simple.

But if I recall correctly, the two mini power supplies on the motherboard can only provide a maximum of 150 watts of power, right? (Please correct me if I'm wrong.) So, if I wish to install an RX5700XT or an RX6800 , it's likely that they won't be able to provide enough power, right?

Or should I really not worry about this? Recently, I noticed some well-priced RX5700XTs (non-mining cards), and thought it might be a good opportunity to replace the RX560 on my cMP. While researching, I noticed the EVGA PowerLink several times, which has really piqued my interest. So I want to confirm again, do I just need to connect the two mini power supplies from the motherboard to the PowerLink? I'm still puzzled about the power figures.

Thank's any responses.

 

[h9826790]

But if I recall correctly, the two mini power supplies on the motherboard can only provide a maximum of 150 watts of power, right?

The mini 6pins are rated up to 75W each. So, total "rated" 150W. But they can actually provide up to about 120W each in real world, which is about 240W in total.

So, if I wish to install an RX5700XT or an RX6800 , it's likely that they won't be able to provide enough power, right?

The card will also draw power (up to 75W) from the PCIe slots. So, in real world, both 5700XT and 6800 can be powered via PowerLink (unless that card is heavily overclocked from factory etc).

Of course, you may downvolt the card etc to further increase the margin. But in general, that's just something "nice to have", but not necessary.

 

[TzunamiOSX]

The mini 6pins are rated up to 75W each. So, total "rated" 150W. But they can actually provide up to about 120W each in real world, which is about 240W in total.


The card will also draw power (up to 75W) from the PCIe slots. So, in real world, both 5700XT and 6800 can be powered via PowerLink (unless that card is heavily overclocked from factory etc).

Of course, you may downvolt the card etc to further increase the margin. But in general, that's just something "nice to have", but not necessary.

Sure? I was of the opinion that at 8 amperes it was over.
P = U x I

8 Ampere x 12 Volt = 98 Watt

 

[ddhhddhh2]

The mini 6pins are rated up to 75W each. So, total "rated" 150W. But they can actually provide up to about 120W each in real world, which is about 240W in total.


The card will also draw power (up to 75W) from the PCIe slots. So, in real world, both 5700XT and 6800 can be powered via PowerLink (unless that card is heavily overclocked from factory etc).

Of course, you may downvolt the card etc to further increase the margin. But in general, that's just something "nice to have", but not necessary.

Ok, I understand now. I will proceed to set it up this way. Thank you.

 

[TzunamiOSX]

Ok, I understand now. I will proceed to set it up this way. Thank you.

A card with under 250 Watt total is possible (Apple say 225 Watt, 75 Watt over PCIe Slot, 2x 75 Watt over PCIe Power connector), all other need the Pixlas mod.

 

[h9826790]

Sure? I was of the opinion that at 8 amperes it was over.
P = U x I

8 Ampere x 12 Volt = 98 Watt

8A is the display limit. The actual power deliver can be more than that. And experiment shows the hard shutdown protection will be activated when any mini 6pin reach about 120W. This is why the PowerLink is so useful on the cMP. It can balance the power draw between mini 6pins. And smooth out the power spike.

 

[tsialex]

A card with under 250 Watt total is possible (Apple say 225 Watt, 75 Watt over PCIe Slot, 2x 75 Watt over PCIe Power connector), all other need the Pixlas mod.

It's not so simple, for basically two reasons.

First, no modern GPU have a power draw of more than around 45W from the PCIe slot. Last card that had a 75W power draw from the PCIe slot was the reference RX 480 and we all know the embarrassment of AMD having to recall the GPUs and send new drivers to stop the motherboards from melting. So, even if the Mac Pro can safely provide 75W from the PCIe slot, no modern GPU will draw 75W from it. You can't account 75W from the PCIe slot, at best 40W, maybe 35W. Some GPUs have almost no PCIe slot power draw, since the PC GPUs have a direct connection to the PSU and that's where GPU power plane is designed to draw power, not the PCIe slot.

Second, some GPU power circuits have crazy power draw spikes when the card changes from the normal run to full power. Spikes are the reason that a lot of people with GPUs that have a mere 210W power draw, like the common non-OC VEGA 56, have frequent SMC emergency shutdowns with a MacPro5,1. Even some RX 580 that is just a 185W nominal power draw models can have SMC emergency shutdowns.

 

[TzunamiOSX]

It's not so simple, for basically two reasons.

First, no modern GPU have a power draw of more than around 45W from the PCIe slot. Last card that had a 75W power draw from the PCIe slot was the reference RX 480 and we all know the embarrassment of AMD having to recall the GPUs and send new drivers to stop the motherboards from melting. So, even if the Mac Pro can safely provide 75W from the PCIe slot, no modern GPU will draw 75W from it. You can't account 75W from the PCIe slot, at best 40W, maybe 35W. Some GPUs have almost no PCIe slot power draw, since the PC GPUs have a direct connection to the PSU and that's where GPU power plane is designed to draw power, not the PCIe slot.

Second, some GPU power circuits have crazy power draw spikes when the card changes from the normal run to full power. Spikes are the reason that a lot of people with GPUs that have a mere 210W power draw, like the common non-OC VEGA 56, have frequent SMC emergency shutdowns with a MacPro5,1. Even some RX 580 that is just a 185W nominal power draw models can have SMC emergency shutdowns.

My explanation came from the theory side

300 Watt : 4 Slot = 75 Watt each.

In reality, you are totally right. Some newer cards really have hard power spikes.
But a Vega 56 has two 8-Pin connectors who show that you must calculate with a power draw of up to 300 Watt over the two connectors.

If the card never need more than 250 watts through the connectors, then it would have one 6 and one 8-pin connector.

The safest way is to connect the connectors according to the specifications.
1 x 8-Pin connector (150 Watt) = 2 x 6 Pin connector (2 x 75 Watt =150 Watt)

Never had problems with my RX 580 and shut-downs (furmark test).

@ddhhddhh2

I would never use a RX5700XT/RX6800 without Pixlas Mod.
I would say a 5600 XT with 1x8-Pin looks like a better option.

 

[BeatCrazy]

I'm using a 5,1 cMP + EVGA PowerLink with my AMD RX5700 XT Anniversary Edition. No issues at all, but I can't say I've stressed the GPU or drawn close to max power.

 

[TzunamiOSX]

I'm using a 5,1 cMP + EVGA PowerLink with my AMD RX5700 XT Anniversary Edition. No issues at all, but I can't say I've stressed the GPU or drawn close to max power.
View attachment 2230469

I'm a little bit confused.
You are installing a powerful card, but you don't need the power?

You are connecting this powerful card, whose performance you do not use, with an energy source that is not designed for the card and risk an emergency shutdown or a main board damage. But why?

Why not use a Version lower like the 5600 XT or 5500 XT.

For example: I have a Vega Frontier because sometimes I need the card for Video AI and sometimes for gaming. But I have made the Pixlas Mod and use a power table to undervolt (less consumption with the same performance) the card.

I know, you need "MORE POWER" ^^

 

[BeatCrazy]

I'm a little bit confused.
You are installing a powerful card, but you don't need the power?

You are connecting this powerful card, whose performance you do not use, with an energy source that is not designed for the card and risk an emergency shutdown or a main board damage. But why?

Why not use a Version lower like the 5600 XT or 5500 XT.

For example: I have a Vega Frontier because sometimes I need the card for Video AI and sometimes for gaming. But I have made the Pixlas Mod and use a power table to undervolt (less consumption with the same performance) the card.

I know, you need "MORE POWER" ^^
View attachment 2230494

I use a 5700XT because that’s the card I had leftover from a previous Windows build. Until someone shows me a cMP that was ruined from using a similar card, I’ll take the risk. I’ve got <$300 invested in this machine. If it blows up tomorrow, I’ll be OK.