Largest Swiss Retailer Digitec to Receive a Grand Total of 35 AMD RX 6900 XT Graphics Cards for Launch

[nguyen]

That is very interesting. It would be nice to see reviews investigating it - also the reasons why that is.

Because at stock settings, both the 3080 and 3070 are operating at the peak of performance/power curve

By reducing the power consumption of 3080 from 320W --> 220W, performance reduction is around 10-15% at most, meaning the 3080 at 220W still perform better than stock 3070. You can do it to the 3070 too but the efficiency gain by lowering the power limit is not as much as with the 3080.

If you look at the 2080 Super Max-Q reviews, you will see that at the same TDP of 80W, the 2080 Super Max-Q outperform all slower GPUs, even though the desktop 2080 Super has terrible perf/watt.

 

[AusWolf]

Because at stock settings, both the 3080 and 3070 are operating at the peak of performance/power curve

By reducing the power consumption of 3080 from 320W --> 220W, performance reduction is around 10-15% at most, meaning the 3080 at 220W still perform better than stock 3070. You can do it to the 3070 too but the efficiency gain by lowering the power limit is not as much as with the 3080.

If you look at the 2080 Super Max-Q reviews, you will see that at the same TDP of 80W, the 2080 Super Max-Q outperform all slower GPUs, even though the desktop 2080 Super has terrible perf/watt.

I've just had a little play with Wattman. With the power slider moved to -25%, I lost only 7% in performance. I'm not saying that this is something I'd want to do 24/7, but it's very interesting to note.

The downside is that with this setting, the GPU fan curve is much more relaxed, so I only managed to reduce the total system noise, but GPU and VRAM temperatures stayed the same. I guess I'll have to play with the fan curve as well.

 

[nguyen]

I've just had a little play with Wattman. With the power slider moved to -25%, I lost only 7% in performance. I'm not saying that this is something I'd want to do 24/7, but it's very interesting to note.

The downside is that with this setting, the GPU fan curve is much more relaxed, so I only managed to reduce the total system noise, but GPU and VRAM temperatures stayed the same. I guess I'll have to play with the fan curve as well.

That's why I said for games that already running at high FPS, for example 140fps, there is no point running full power limit, you can't notice a 10% perf loss there but you will notice the lower fan noise.
For graphics intensive games that every fps count (~60fps), of course you will want to max out the power limit.
So many possibilities when you have high end GPU, not so much for midrange .

 

[mtcn77]

I'm curious what that would do in terms of efficiency. The 6000 series (and nvidia 3000 series) GPUs are sure fast, but shoving a 300+ W room heater into my PC? Hmm... nope! I'm having enough trouble with the heat output of my 5700 XT as it is.

I respect your preference switching speed that races with these chips, however I have to ask when has distributed computation have performed worse than singular computation devices? The answer is never.
Be at ease and don't struggle against the progress in the industry.