The Chinese website Zhihu published the first results of the MSI Delta 15 laptop with the Radeon RX 6700M. It is built on the GPU Navi 22 with active 2304 stream processors and 10 GB of GDDR6 graphics memory on a 160bit bus.
According to the average of all game results, Radeon achieved 109.1% GeForce performance.
The editorial staff of the VideoCardz website came up with an interesting hypothesis that Radeon is faster, but it is because it has a higher TDP. This is supported by a comparison of the data that the system with the GeForce RTX 3070 was equipped with an 80W version and the Radeon RX 6700M was about 110 watts. VideoCardz came to this conclusion by considering that the Radeon RX 6700M exists in 80W, 110W and 135W versions, while (according to the test author) it is excluded that it is a 135W model. Using the SmartShift function, Radeon consumed up to ~ 115 watts, from which VideoCardz estimates the 110W model.
This reasoning has two cracks. The first is the fact that laptops supporting SmartShift do not have a fixed TDP set, but there is only the total cooling capacity of the laptop (in this case 130 watts) and the GPU and CPU can share without one or the other being limited by a fixed limit. for a given chip. If we subtract 45 watts from the total cooling capacity of 130 watts, which would normally be a TDP processor, we have 85 watts left, which would not be mistaken for the 110 watts that VideoCardz is talking about. The GPU can only take over 85 watts if the CPU “spills” a portion of its own “TDP” that it does not need at the moment.
The second flaw in VideoCardz’s reasoning lies in a similar fact, but with Nvidia. It is a GeForce RTX 3070 in Max-Q and this brand has included a similar TDP overflow as last year as AMD SmartShift.
At last year’s slide, we can see that the 80W GeForce can use a technology called Dynamic Boost in the case of the Max-Q model and transfer part of the processor’s “TDP” to graphics that run above 80 watts (for an illustrative situation in a slide at 95 watts) .
Thus, in the context of these products, we cannot speak of a TDP graphics card, as the power requirements can legitimately exceed or reduce the hypothetical TDP and provide it to the processor when the GPU is not busy. The key in this case is the total consumption limit (for the sum of CPU and GPU), which, however, is not known in the case of the assembly with GeForce.
–
