• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

How are people with 3000 series cards faring with PSU usage?

Associate
Joined
24 Apr 2019
Posts
52
I've seen reports of spiking power draw and my psu manufacturer advised me to go 850W for a 3080. How are people finding the cards, especially if you're on a 650W or 750W PSU?
 
Will let you know. I've got a Palit 3090 Gamerock OC arriving today to pair with my Superflower 750W PSU. It does support 62A on the 12V rail and 2 separate PCI-E cables. My 1080 + i9 pull 360W from the wall during AC: Valhalla. So going to check out the same area with the 3090 and see what the difference is.

Yeah that will be v interesting. Mine is multi rail and can take 30A on the GPU rail, but transient spiking up to 450-550 could trip that. I was wondering what the recommended single rail current is
 
Running 700w SFX-L prior (Silverstone SFX) but now 750w (Corsair SF) SFX unit with 8700k and 3090 as its going into a HTPC. No issues. even with both parts OC'd, not missed a beat.

Usually calculators and recommended specs are increased to account for mediocre units and given the manufacturer headroom.

This is promising. Funny enough calculators and my wattage calculations suggest I can use 650 still but my current calculations and the manufacturer have come to 850W so the 40A GPU rail isn't at risk of OCP tripping

It's a top rated PSU as well
 
Any properly engineered PSU designed at 500W+ will likely be fine with transient spikes. LOL wattage kW PSUs aren't required in the last majority of cases just to cover off the chance.

Normally yes but when it's spiking up to 600W on just the GPU that's a lot more than say a 1080ti. My psu would have been fine but the GPU rail only handles 30A which is why I'm worried. Any old GPU that was plenty. This is one of the highest rated multi rail GPUs
 
Let's assume gpu spikes to 600W and CPU running around 250W. 850/12V = 70A so in theory and single rail with 62.5A aren't enough right?
 
Lets assume that the stated wattage is for continuous load, and transients can go higher as long as the voltage meets ATX spec.

Is there any literature to support this? Would be nice to know exactly what level of current spike it can handle without ocp kicking in. My manufacturer advised 850W version which can handle 40A on GPU rail
I have no idea if there are spikes. It hasn't turned itself off or anything and my 3DMark scores seem about the expected level. The 3080 is undervolted and restricted to 1900Mhz. Doing that actually gave me higher scores than leaving it at stock and letting it reach 2050Mhz (which it only did intermittently, whereas undervolted it stays at 1905Mhz). My 1660 Super is never under heavy load. It is purely there for additional displays in xplane. The 3080 is doing all rendering.
Very interesting. I wonder if the driver update plus your undervolt and restriction help it there.
 
Only had a few hours with my 3090 but so far, so good with my Corsair 650w PSU. I've run quite a few games, synthetic tests and 30 mins of Furmark and no problems at all. It's still at stock and I have a 3700x which is just using PBO so I guess I'm being careful not to overload on the power.

I think if you have a good quality 650/750w PSU you should be fine, unless you go over the top with other components drawing a lot of power.

Promising results. Which PSU model if I may ask?
 
I've just taken current readings from the mains input to my system (GTX1070 + R5 3600)

Win10 Desktop 126w

Prime95 169w

Heaven Benchmark 259w

Furmark stress test 363w

These are the highest readings I saw after a few minutes of each test.

GTX1070 is rated at ~150w. The elusive RTX3080 is ~320w. If I assume the 3080 uses 200w more than my 1070, my highest power load would be around 560w. I think my 650w PSU would cope fairly well.

In some cases it's the current I'm more worried about than the wattage. The transients go pretty high but it seems they might not be that long duration according to the Igor article I linked
 
So many people who are clueless giving their "expert" advice. The idea that you're saving money by spending £150 on a new 1000W PSU to avoid running outside of your "efficiency sweet spot" is absolute nonsense. Looking at the efficiency curve for my particular unit, the difference between 300W and the full 750W is 2%. 2%. And it's still 90% efficient under those less than "ideal" conditions. I'm sure that's really going to add up on the electricity bill. Maybe if I use it maxed out 24/7 for the next few decades I might creep up towards what a new PSU would cost.

efficiencylpk4x.jpg


This particular unit is also rated for its full 750W draw on the 12V rail, as a sustained load at 50 degrees celsius. A temperature that it won't be getting anywhere near in actual real-world use. PSU reviewers struggle to get anywhere near that even in torture testing with hot boxes designed specifically to get the PSU as hot as possible. Looking at the JG review, 44 degrees was the hottest he ever managed to get it even under those conditions. And of course that's running the full 750W as a sustained load, which certainly won't be happening with the average system and these Ampere cards. Nowhere close in fact, given the power limit on most of them is in the 320-370W range. Sure, if you want to buy an AIB card and then flash it with a 480W special BIOS and then overclock the snot out of it for the sake of a few percent more performance, buy yourself a nice PSU to go with it. But unless you're running a 10900K or HEDT chip overclocked to the hilt and performing a Blender render in the background while you game, you'd be lucky to see even 500W sustained while gaming. And it's not even sustained load which is causing problems for certain units with these cards for that matter. They're shutting down due to transient spikes tripping poorly-set (or covering for poorly-built platforms) OCP or OPP.

Yes exactly this
 
Back
Top Bottom