@ElementX
No, they don’t otherwise they’d quit bringing this up.
An RDNA GPU performs better than a GNC 1 – GNC 4 (From Pitcairn up to Polaris / Vega GPU), because the Compute Units has better IPCs and an improved architecture making a single GPU core run let’s say 50% better than the older architecture. For example, if the Series S used 20 GPU cores in this scenario and the Xbox One X used 20 GPU cores and the same clock speed, the Series S would still be 50% faster and more efficient than the Xbox One X even though they would have IDENTICAL TFLOPS.
TFLOPS is a math equation that only works for the same generation of architecture.
TFLOPS = Compute Units * 128 * Clock Speed
So using the above example and a 1.5 GHz clock speed we’d get.
TFLOPS = 20 * 128 * 1.5 GHz
TFLOPS = 3.84 TF
Let’s use the AMD Vega 64 (64 CUs clocked at 1.5 GHz) vs the AMD RDNA 5700 XT (40 CUs clocked at 1.9 GHz).
Vega 64 TFLOPS = 64 * 128 * 1.5
12.288 TFLOPS
5700 XT TFLOPS = 40 * 128 * 1.9
9.728 TFLOPS
On a side note that’s actually pretty close to the PS5 / Series X totals. However, getting on topic, if you go by TFLOPS like you currently are @Raiden, then you’d think the Vega 64 is better, but the reality is the 5700 XT performs on average 20% better in games than the higher TFLOP Vega 64, because it has a new and more efficient RDNA 1.0 architecture. Even though the 5700 XT has less GPU cores (aka 40 CUs vs 64) and a lower TFLOPS (9.7 vs 12.8), it still outperforms the Vega 64 because those fewer cores are around 50% more efficient at doing their job, which theoretically puts the 5700 XT at 14.592 TFLOPS if it were a direct comparison to the Vega 64 card vs 12.8 TFLOPS.
That’s why you can’t compare them directly.
In the case of the Series S it is using RDNA 2.0 compared to the Xbox One X GNC 2.0 (basically an R9 390x moved to 16nm from 28nm which allows for the higher GPU clock speed and lower TDP). RDNA 1 was a 50% improvement of performance per watt improvement over Vega which basically means, you get 50% more performance at the same levels of power and same core count. RDNA 2.0 delivers an additional 50% on top of that.
So let’s say in a magical world we could make a launch PS4 automatically update to RDNA 2.0. With the RDNA 2.0 update, that launch PlayStation 4 would update and be just 20% less powerful than a PlayStation 4 Pro. Same clock speeds, same number of GPU cores.
That’s why you can’t say Series S is less powerful than Xbox One X. At 4 TFLOPS, we can kind of figure out a range for the GPU it’s somewhere between 16 CUs – 20 CUs with a 2GHz – 1.6 GHz clock speed. My guess is it’s 18 or 20 CUs with the same clock speed as the Series X depending on form factor (higher) which puts it at 4.26 TF up to 4.74 TF which should put it right on par with the XBO in performance being in the RX 570 – RX 590 performance range.
