Moving from a 2080ti to 3090ti PSU?

[FishD]

Watt meter shows i'm drawing 425W from my 4.9ghz 8700k & RTX2080ti on an antec gold 750psu. (does not change where gpu is loaded or not).

Would this be enough for a 3090ti with the extra 200 watts rated power, bringing my total power to 625 on a 750w psu?

 

[Woomack]

Even though recommended is more, then it should be fine. On the other hand, if you can afford RTX3090Ti then the new PSU shouldn't be an issue
If your current PSU has thin cables then I would replace it. If it's older than ~7 years then I would replace it too. Other than that, if you load the PC and it will randomly shut down then the PSU can't handle it.

If your PC draws 425W without load then something is wrong as the power saving should work (no matter if you have it additionally enabled or not). I count it should be something around 100-150W without load max.

 

[EarthDog]

Agreed. 750W should be enough...but if that psu is old and you start having issues under load, it's where I'd look first.

 

[Blaylock]

Something to consider if/when you do replace the power supply is to look for one that has a dedicated 12VHPWR (12+4-pin) PCIe cable. The +4-Pin section contains sideband signaling that will not be functional with the 3x8-pin to 12-pin adapter. Without getting technical this can, and likely will, ultimately lead to performance loss as the system will limit the power supplied to the GPU.

 

[EarthDog]

The +4-Pin section contains sideband signaling that will not be functional with the 3x8-pin to 12-pin adapter.

FTR, the 3090 Ti Suprim came with a PCIe-->12VHPWR adapter and it can pull over 450W/reaches its full potential. I wouldn't worry so much about an adapter that's included with the card so much as I would buying one aftermarket that has the two sense pins there to get the full 600W out of the cable (for a 3090Ti, otherwise, even the 450W config is good for most existing cards). It just depends on the config of the adapter's +4 and if the grounds are there/where they are located. A quick trip to amazon yielded one 12VHPWR connector that specifically said 600W. So I'd imagine they'll sell these (reputable sellers, lol) with a wattage rating.

This is a great read on it...

PCIe Gen5 "12VHPWR" power connector to feature 150W, 300W, 450W and 600W settings - VideoCardz.com

Intel’s ATX 3.0 specs for PCIe Gen5 power cable It appears that official specs of the PCI-Express 5.0 power cable for next-gen graphics cards have been leaked. Officially going by a name of “12VHPWR” is the upcoming standard for PCIe Gen5 graphics cards. The data provided by Intel, who are...

videocardz.com

But this of course does not mean graphics cards equipped with 12+4-pin power connector will not work with adapters.

 

[Blaylock]

Agreed, but we don't know which RTX 3090ti OP is looking at. Just because the Suprim doesn't throttle with the included adapter does not mean a "lesser" model wouldn't. A new PSU with dedicated connectors is certainly something to consider, though may not be necessary.

 

[EarthDog]

I don't think the model would matter. Any 'lesser' 3090Ti would still come with the proper adapter (read: one that doesn't throttle the card). And if it comes with the 450W adapter, the 'lesser' models would have less chance to throttle since they tend to use a bit less power (with more limited power limits). Now, if they put a 300W adapter in with a 3090 Ti.................. that would leave something on the table..... but why would they do that on purpose?

My point was simply to clarify that the included adapters work/will work just fine.... but you need to get the right one if it isn't included in the box of the video card (they will, for years to come). I agree 100% that if you're buying a PSU today, you should consider it to include the new connector. The only problem with that is the lack of selection at this time. AFAIK, there are only a few models out there with them today. And if I didn't like what was out there that include the adapter, I wouldn't hesitate to buy a new one without.

 

[FishD]

Many thanks for all the replies. My take home from your comments are.

1. First look into why my setup is idling at 450 watts.
2. Get a reputable brand Card, and best get one with the cables included.
3. Test out with existing PSU (4 years old), before jumping to a new one (already stretching budget for the card)
4. If i run into issues, then look into a new one in more depth.

 

[EarthDog]

1. You're IDLING at 450W? That's not right......unless you have some virus causing high CPU/GPU use, your watt meter is clearly wrong.
2. Any 3090 Ti with that connector will include an adapter.
3. I'm not sure how you're going to test your current PSU thoroughly. All you can do is check voltages on the rails with a multi-meter (don't use software), but that doesn't tell you the whole story. How old is that Antec? That model, IIRC, is old but not sure when you bought your PSU.
4. Right, if you run into issues when under load, that would be the first place I look. That said, I ran a 12900K and 3090 Ti (Suprim X) which uses a lot more power than your system using a 750W PSU without issue.

 

[Blaylock]

It looks like his PSU is from 2018 (4 years old). It should be fine.

EDIT: agreed on the idle issue. 450W idle is wrong or there's something else causing it to not idle (virus).

 

[EarthDog]

Yeah, I saw that too but wasn't sure if he bought the PSU with that build or if it was a carryover. It's an old model that's been around for several years, even prior to the 2018 signature.

 

[FishD]

Yup, I was totally wrong with the 450 watts. was measuring something else.

Here's the correct results.
150-200w idle
220w CPU loaded (prime)
500w GPU loaded (82c),
Hitting 600w! with GPU and CPU loaded, and crashing.

(So stability issue with the current GPU I need to look into, or maybe reaching the limits of the old PSU!)

And this PSU will not cut it for a 3090ti.

Might also need a climate control system in the room !!!

Once again thanks for the advice,

 

[EarthDog]

I'll ask again...how old is the PSU???

And this PSU will not cut it for a 3090ti.

???

When working properly, it should be enough. Again, I ran a 12900K (at stock uses more power than your CPU) and 3090 Ti Suprim X with a 750W unit. Worth noting, 600W is the power taken from the WALL, not what the PSU is outputting. The PSU itself, assuming 90% efficiency, is actually using ~540W which is well under the limit.

Also worth noting, when are those loads realistic? When are you using 100% CPU like in a stress test while gaming? Also, what test did you run to load the GPU? Please not Furmark, lol. Unless your use case is running the CPU at 100% like a stress test (think rendering, maybe F@H) and the GPU at 100% you're fine. I'd bet your gaming wattage is much closer to the 500W (~450W actual) value than it is 600W...

Once again thanks for the advice,

Happy to help, but it feels like you're walking away with the wrong idea (that PSU isn't enough) ATM. Again, if it's working right, it should be fine... at least hold you over until you can save up for a new unit if the 3090 Ti kills the bank.

 

[WrkBoot]

Yeah...I would say, buy a new PSU. That way you are sure you're getting the power you need and the correct cabling.

 

[DaveB]

I

Also worth noting, when are those loads realistic? When are you using 100% CPU like in a stress test while gaming? Also, what test did you run to load the GPU? Please not Furmark, lol. Unless your use case is running the CPU at 100% like a stress test (think rendering, maybe F@H) and the GPU at 100% you're fine. I'd bet your gaming wattage is much closer to the 500W (~450W actual) value than it is 600W...

Important point. I got a lot of uninformed sarcastic remarks about using a 450W EVGA PSU in my i7-12700F/3060 Ti rig. But my usage profile NEVER has both the CPU and GPU at 100% load. The CPU usage profile is 175W PL1/241W PS2 with the GPU under 100W. The GPU profile is 200W with the CPU under 100W. So, maybe 325W max in either profile. The PSU is commonly over-spec'd by many DIY builders.

 

[Blaylock]

If you're crashing at 600W, I would question your overclocks first, then your thermal solution. If those are all in check then maybe the PSU but I would do some tests before concluding it's the PSU. At 600W you are well below that PSU's limit and up till this point, you were unaware of your 'instability' leading me to believe the stress test is not a real-life scenario for your use.

Stress test the CPU by itself. Stress test the GPU by itself. Then test together watching your CPU, GPU, and case temps. Naturally do this with your fans pegged at max.

 

[FishD]

Again good points guys.
Yes its been a while since it OC'ed this rig, so need to get back to understanding the instability issues and if real or not, and will post in appropriate channels as needed. Have not had issues with games I play or rendering in DR. So all good ATM. And yes (EarthDog) crashes were with Prime and Furmark, as I have been using them forever for stressing, so comparing like with like. I'll trawl the threads to get up to date views on stressing.

 

[EarthDog]

Stress testing is a personal thing, bud. You ask each person here what they use, how, and for how long, you'll get as many answers.

Me, I prefer using AIDA64 and it's System Stability test. I use CPU/FPU/Cache for the CPU, then add System Memory when testing that. I find that testing for a few hours at the end of all the tweaking yields a stable system for me and my uses.

If your issues only happened in an unrealistic stress test load, then I wouldn't sweat it.

 

[Blaylock]

Like Earthdog, I generally use AIDA64 Engineer (free) for stress testing the CPU, though I still occasionally drag out Prime95. For GPU stress testing I stick with 3DMark. They have a very nice suite for different GPUs. I use Firestrike, Firestrike Extreme, or Firestrike Ultra depending on the GPU but also will use Time Spy and Port Royal once in a while.

 

[FishD]

Just to close the loop on this thread.
Finally jumped for an RTX 3090ti (see sig), but put in an 850w PSU. System pulling 600w on unigine bench, though I guess gaming will be less.

Initial results show pretty much a straight 40% improvement in performance from the 2080ti