Two cables (1x8 + 2x8) for 3x8 GPU

[yoadknux]

Hi guys,

I'd like your advice on something.
I recently upgraded my 2080Ti FTW3 to a 3080 12GB Suprim X (maybe I'll post a comparison topic here if I have time).
My Power supply is a Corsair RM750i. It's about 4 years old. This PSU has two daisy chained PCI-E cables, for a total of 4x8. My previous GPU (375W OC load) required only 2x8 PCI-E so I used two separate cables. This one (436W OC load) requires 3x8 PCI-E, so I had no choice but to use one 1x8 cable and 2x8 daisy chained cable.
So far I have no shutdowns or issues. The RM750i has a digital interface, and it reports that peak gaming draw (Cyberpunk, God of war, Overwatch etc) is about 650W, which is well below the wattage rating of the PSU. Only when I put unusually high load (for example running P95 small FFT in parallel with FireStrike Ultra Stress test) I cross the 700W mark, but no issues.
My only question is, is there a higher risk of overload when using the daisy chained cable? According to HWInfo, Pins 2+3 draw AT MAX 133+115=248W. In addition, does a multi-rail configuration somehow protect a single cable from being overloaded? The RM750i can be switched from single- to multi-rail, and it's currently in its default multi-rail configuration.

Would love to read your responses! Thanks!

 

[freakdiablo]

If the power supply came with the daisy chained cables I doubt there's a risk of overload - Corsair spec'd them out and designed accordingly.

As for the single rail/multi rail question, multi rail doesn't protect a single cable as much as other components if one rail gets overloaded. Or so the theory goes. I know there was a lot of debate back in the day which is better, I've always been of the mindset that as long as you're not overloading any one rail, you're good.

 

[BugFreak]

I'll leave the rail question to some of the more knowledgeable folks out there but I'll chime in on using the daisy chain cables. Unless I'm mistaken some of the earlier problems with3xxx cards dying was related to using those second plugs. It wasn't the primary reason but it was sure something a lot had in common. The 3080 takes a bunch of power so I personally only use the main plug and run three separate cables.

 

[don256us]

I can remember that back in the day, PSU manufacturers were advertising multiple rails with the promise of greater and more stable power. It didn't work and nearly all PSUs have a single rail for each voltage. Whether you use one, two or three cables, they are all connected to the same rail inside the PSU. That's just the way it is.

Further, you want to run around 75% of the rated maximum load for greatest efficiency. If your PSU is older, it will not be as efficient as when it was new.

 

[yoadknux]

Thank you all for your comments. The vibes I'm getting from this post are:
a) Daisy chaining a second cable is fine, but not preferable
b) It's probably not a good idea to put 80-85% continuous load on a 4 years old PSU

Guess I'll be looking for a replacement then.

 

[EarthDog]

A. Yes
B. It's fine too so long as it's a quality unit.

Also, a psu is most efficient right around the 50% mark. At least that's what you see for 80 Plus standards. If you look at reviews that test efficiency, you can see that in practice. Here's a list....

(Uhh copy/paste link isn't working on my phone... go to techpowerup psu reviews and look at the charts)

....go to 3fficiency pages and look at the curve. I attached a an example too. That said, there's no way I'd buy a psu and run it at 50% capacity as that's a waste of money. Efficiency curves are so flat after they ramp up (around the or before the 10% depending on if it's titianium) the difference between running a psu at 20%, 50% or 75% capacity is 2% at most. If you buy a 1.2KW PSU for your 600W machine, you'll never make up the substantial price difference between the units by efficiency. Remember 80 plus tests at 10/20/50/100. Even at 100% 80 plus requirements are 2% different than 50%. I'd call 75% load a sweetspot between price for the unit and efficiency, but it isn't the most efficient part of the Power band.

Worth noting, I happily ran a 12900k and 3080ti on a 750w psu (iirc, one power cable was daisy chained too)...I dont see why you'd get a new power supply...

The concept of a multi rail psu was, imo, a good one. Half that claimed they were multi rail was bologna anyway as there wasn't any ocp on each rail. In other words, one rail could pull all the amperage in some of these units. Those that were, you had to 'balance' the load across the cables sometimes or risk ocp triggering.

 

[yoadknux]

A. Yes
B. It's fine too so long as it's a quality unit.

Also, a psu is most efficient right around the 50% mark. At least that's what you see for 80 Plus standards. If you look at reviews that test efficiency, you can see that in practice. Here's a list....

(Uhh copy/paste link isn't working on my phone... go to techpowerup psu reviews and look at the charts)

....go to 3fficiency pages and look at the curve. I attached a an example too. That said, there's no way I'd buy a psu and run it at 50% capacity as that's a waste of money. Efficiency curves are so flat after they ramp up (around the or before the 10% depending on if it's titianium) the difference between running a psu at 20%, 50% or 75% capacity is 2% at most. If you buy a 1.2KW PSU for your 600W machine, you'll never make up the substantial price difference between the units by efficiency. Remember 80 plus tests at 10/20/50/100. Even at 100% 80 plus requirements are 2% different than 50%. I'd call 75% load a sweetspot between price for the unit and efficiency, but it isn't the most efficient part of the Power band.

Worth noting, I happily ran a 12900k and 3080ti on a 750w psu (iirc, one power cable was daisy chained too)...I dont see why you'd get a new power supply...

The concept of a multi rail psu was, imo, a good one. Half that claimed they were multi rail was bologna anyway as there wasn't any ocp on each rail. In other words, one rail could pull all the amperage in some of these units. Those that were, you had to 'balance' the load across the cables sometimes or risk ocp triggering.

EarthDog, thank you for your feedback.

I followed your advice and found the voltage/efficiency curve of the RM750i (https://cdn.mos.cms.futurecdn.net/Z5GsUT4DYLaSVzZxcxrHTo-970-80.jpg.webp). Indeed the peak efficiency of 92% appears at about 350W, but it only drops to 90% at 750W. So the argument of "it's not very efficient" doesn't make a big difference here.

Your experience is also very helpful, the peak power draw of your system should even be slightly higher than mine.

I think that based on what you said, I will not upgrade my PSU unless I start seeing issues such as shutdowns, crashes or bsods.

The RM750i should be a decent unit according to reviews. I actually remember that a few years ago I flashed a 1.2V unlimited power bios on my 1080Ti with a daisy chained cable, and some power hungry stress test caused a hard reboot, so I at least know this unit has protective mechanisms.

 

[EarthDog]

So the argument of "it's not very efficient" doesn't make a big difference here.

I'm not sure who said "it's not very efficient"... but yes, there isn't much of a difference at all. Literally ~2% difference.

I ended up with a Titanium 850W unit (was the same price as 750W on sale). I'd drop a 3090 Ti in on this system and never worry.