Showdown: GeForce RTX 3090, RTX 3080 and RTX 3070 including Intel Core i9-10900K with and without OC against 11x be quiet!

Three GeForce RTX and a thick Intel CPU against the power supply, that’s the plan and that’s how I did it power supply by power supply. The gaming load generated the higher load peaks on the graphics card, but the protective circuits of all tested power supplies have a tripping delay of 20 ms. I have tested this and even short spikes with up to over 580 watts on the RTX 3090 OC have never been the trigger for shutdowns. However, the time of 20 ms is still sufficiently short to ensure a safe shutdown in the event of other types of damage. So it fits.

The maximum test with Furmark (1920 x 1080 pixels, extreme burn-in, PostFX) and Prime 95 (AVX, 16 threads, small FFTs) ran for at least 30 minutes on each system. A job that can be done elegantly in the same room and which also saves you the heating in the dull November. For gaming I let good Lara run through the benchmark three times in Shadow of the Tomb Raider. I’m sure she’ll forgive me.

Entry level: Pure Power 11 and System Power 9

Here I always test the model with 700 watts, everything else would not make sense with this test setup, especially since the efficiency in the full load range is no longer the best. Here one should already apply a little reason and insight into what is still reasonably feasible for the price. With the GeForce RTX 3070 it doesn’t matter whether you use two Y-cables with one 6+2 pin connector each or a single cable and then both connectors. They are multi-rail power supplies, but they put both variants away without complaint.  With the RTX 3080 I would however, if at all, use a model with only 2x 8-pin and then take both cables. Due to the price range, I didn’t officially do these tests, but the GeForce RTX 3080 FE, which was later retested out of pure curiosity, ran well, but only with the Pure Power 11 700 Watt.

Lower middle class: Straight Power 11 80 Plus Gold

Now I let go of all three cards, let’s see what works (or not). With the GeForce RTX 3080 and 3090 I need 3x 8-pin, so with the smaller power supplies and the two power cables I used both 6+2-pin connectors of the Y-cable in one of the feeds.

Update from 05.11.2020
As a counter-test I changed the wiring of the 750 Watt power supply again and connected the Y-ends differently to the 3-pin sockets of the MSI GeForce RTX 3090 Gaming X Trio. The individually connected rail is now the one that is plugged into the socket further away from the MB connector. Despite a total of three runs with 30 minutes in maximum Torture OC each, the power supply ran without shutdown. If, contrary to expectations, you have problems here, just change the cable sequence, for whatever reason.

Upper middle class: Straight Power 11 80 Plus Platinum

Efficiency increases, and so does stability. It’s just fun that way. At least as long as the motherboard plays along. The same applies to the smaller models up to 750 watts with the Y-cables.

Upper class: Dark Power Pro 12 1200 Watt

I can save myself the table, because this is a trump card that is beyond all doubt. No matter what, how high overclocked and how long, the hairstyle fits. Even the manufacturer could not find smaller variants in the current test period, but you can make up for it when the opportunity arises. However, I counter-tested an older Dark Power Pro 10 with 850 watts from the water cooling test system with maximum OC and the RTX 3090, and nothing switched off.

Summary and conclusion

A good power supply delivers exactly what it is supposed to, even with Ampere cards. Sometimes even a little more, but it is better never to rely on that alone. You don’t have to overdo it in terms of dimensioning, but you can be a little more generous, also with regard to aging and ambient temperatures. Also in terms of efficiency. The power supply calculators on the relevant pages are or at least often seem to be somewhat (arguably) oversized, but if you take into account the power consumption of the peripherals including the motherboard and the components still installed there, it becomes more plausible in many cases.

The entry-level models should be operated with adequate hardware, you will certainly not risk any stunts. After all, those who buy cheaply always buy twice in the end. Here, therefore, expediency and reason should always take absolute priority over avarice and madness. The latter has also nothing to do with courage and the trust in God belongs pleasantly into the church, but not into the PC. That’s why I only recommend the GeForce RTX 3070 (but with OC on all components) for the two smaller models tested today (Pure Power and System Power), even though the GeForce RTX 3080 (at least not the 3-connector models) would probably do the same. At least with the new power supply (not the matured), but I had already written something about that.

One can also assume that the current Ryzen are running, I have counter-tested this with an overclocked Ryzen 9 3900XT and the Extreme-OC (and also with the upcoming new models). Interestingly, the setups in which the Intel Core i9-10900K performed a shutdown or reboot with the current and the comparable upcoming Ryzen ran stable. I wonder if this could have something to do with efficiency? But we will certainly have to wait a few more days. And it should never tempt you to save money in the wrong place, regardless of CPU efficiency.

In case other manufacturers have read about the challenge, who have not yet been contacted or who have not yet considered it to be effective, it is also important from the consumer’s point of view to concentrate on the real demands of gaming and to be able to rely on the energy source. Anyone who wants to: call or write to us. Who does not, also ok. The power supplies I have stored here will of course be tested later with the AMD cards. Even single units can be combined to a collective test, if there are no wider product ranges. The readers would surely be interested. However, I reserve the right to reject single models if they seem to be unsafe, also with regard to the expensive test hardware.