Radeon RX 7600M XT is the most powerful mobile graphics card built on the core Navi 33. Compared to higher Navi 31 a Navi 33 it does not stand on the 5nm process, but only 6nm, which is TSMC’s 7nm technology enriched with EUV. The big currency of the chips with the RDNA 3 architecture was supposed to be high clock frequencies, which did not turn out as expected. Tests have shown that the chips handle them stably, but only with computational load. With graphics, the stably achievable clocks are significantly lower and the consumption is incomparably higher, so AMD was forced to set clock frequencies for this generation that do not differ much from the previous one.
This technological background suggests that neither performance Navi 33 it will not be as the company planned, but due to the lower clocks (also) lower. The first Radeon RX 7600M XT test shows that the card scores 10,451 points in 3DMark Time Spy and 30,393 points in 3DMark Fire Strike. This places it 4.7% below the mobile GeForce RTX 4060 in Time Spy (10,969 points) and 12.9% above it in Fire Strike (26,931 points).
The original plan for Navi 33 however, it was more of a competition to the “number larger” GA106 core. On the other hand, the Radeon performs surprisingly well when compared to the GA107 / GeForce RTX 4060. One would expect that with handicaps such as Radeon’s 6nm process versus GeForce’s 4nm process and significantly reduced clocks, the card would struggle with efficiency in terms of both power-to-watt and power-to-transistor ratios. However, from these results, it does not appear that the significantly older process somehow Radeon / Navi 33 discriminated against. GeForce RTX 4060 has a TDP of 115 watts, Radeon RX 7600M XT adjustable in the range of 75-120 watts. Even if the above numbers were from the 115W GeForce and 120W Radeon configuration, it would be a very good result.
Radeon Core (Navi 33) is made up of 13.3 billion transistors. The transistor count of the GeForce (GA107) has not yet been disclosed, but higher models such as the GA102 and GA103 achieve an identical density of 0.121 billion transistors per square millimeter, so if we go by the website’s claims TechPowerUp with a GA107 area of 146 mm², the number of transistors would correspond to roughly 17.7 billion.
This data, combined with the performance results, creates a strange picture according to which it would Navi 33 / Radeon RX 7600M XT with lower transistor count, 6nm process and low clocks handicap performed slightly better than GA107 / GeForce RTX 4060 with higher transistor budget, 4nm process and higher clocks.
The conditional is used here on purpose, because it seems unlikely that – considering how GeForce RTX 4000 and Radeon RX 7000 stand in the high-end – the situation would be so diametrically different in the lower price segment. The position of the GeForce RTX 4060 compared to the Radeon RX 7600M XT is simply not what it should be, and from the little information that is known and certain so far, it is not possible to reliably deduce why this is so.
