In this article we will see by comparing each manufacturer's bet and how they deal with technical categories such as their features and their engineering. Of course, we will compare your performance with the software indiscriminately 3DMark Time Spy, which includes all the latest features and facts of APIs, as well as Asynchronous Compute, remains questionable between both products.
RDNA 1.0 vs Exercise: who's moving a cat into the water?
As we can see, both chips have a perimeter size that is very similar to their PCB, and although this is rarely measured, when done depending on the size of the chip. In this case we are dealing with the NVIDIA TU116-300-A1 against the Navi XLE, the first at 12 nm and the second at 7 nm, both produced by TSMC.
The NVIDIA chip measures 284 mm2, while AMD at 7 nm is smaller at 251 mm2, which in percent gives us a 11.61% prefers the youngest of them.
The lithographic process describes the number of transistors that can be installed per chip, so the calculation goes up to 6,600 million in the case of the NVIDIA chip and 10,300 million in the AMD case, showing what Lisa has been able to put into the 11.61 physical space % a 56.06% more transistors
Both products feature in their small units of 64 shades, so knowing that in the case of the Navi 10 XLE 36 CU is included and in the case of the TU116-300-A1 22 SM is included, it gives us Shaders Prices 2,304 and 1408, an amazing 63.63% more for AMD.
This obviously affects the ROPs and TMUs, in the case of 144 vs 88 and 64 vs 48, a difference at the level we have already seen, because they are perfectly equal.
Finally, and within this category, we have L2 storage memory, where AMD is twice the price on its RX 5600 XT card compared to its NVIDIA competitor, which gets 3072 KB versus 1536 KB.
Frequencies and VRAM: a specific problem
Although the core and muscle of the game are less important when the player in their play and expects that NVIDIA doesn't have a second one, but it goes straight to increasing the frequency, which is exciting to compare the latter logically.
The RX 5600 XT reaches 1560 MHz, while the GTX 1660 SUPER reaches 1785 MHz. This is controversial, because an AMD card can see this value too high for custom models, which almost match the figures of its equivalent. Controversially, NVIDIA and its custom variants never ring other MHz, but as we compare the factory features, it should be said that NVIDIA gets 14.42% more often than not the competitive model.
As for its VRAM, it's another issue in the same way as the increasing frequency has always been. Although both models feature GDDR6 and 6 GB of total capacity, AMD specifies with its reference model the speed of 1500 MHz, that is, 12 Gbps.
NVIDIA, on the other hand, opted to launch at 1750 MHz or 14 Gbps initially. The controversy comes because there are currently more RX 5600 XT models with memory at 1750 MHz than 1500 MHz, because AMD changed the details of many manufacturers, allowing for such an increase in clocks.
In any case, and depending on the reference definitions as usual, NVIDIA will have one line 11.66% of profits when it comes to watches. This contributes to the same percentage when comparing bandwidth, since both cards have a 192-bit bus, so we'll get 288 GB / s for the RX 5600 XT and 336 GB / s for the GTX 1660 SUPER
Performance, utility and price
Theater performance in most cases does not reflect actual performance, but in this world you have to specify prices in order to be able to compare and be reliable. Thus, the FP32 performance of the AMD card is 7,188 TFLOPS, while NVIDIA produces 5,027 TFLOPS (+ 42.98%).
Though related to this, PCIe versions are also as important as these, especially for newer Intel and AMD platforms. Therefore, the RX 5600 XT is equipped with PCIe 4.0 X16, while the GTX 1660 SUPER card features PCIe 3.0 X16.
The difference is 100% performance from 4 to 3, as it doubles the amount of its bandwidth at bus speed.
Both cards on their stock PCBs will have one 8-pin connector, enough for 150 watts of the red model and 125 watts of the green model, which means the NVIDIA GPU 16.6% work well rather than AMD.
Depending on the price, the RX 5600 XT starts at $ 279, while the GTX 1660 SUPER is cheaper and costs $ 229 for its basic or reference models. That is, the AMD model is 21.83% too expensive
Finally, what about actual performance? As we noted, we will use the 3DMark Time Spy test in a comparative and impartial way to list the features we would expect under current and future DX12 games.
Scores are mapped to 7824 points for the RX 5600 XT and 6173 points for the GTX 1660 SUPER, giving us a percentage difference 28.36%
Knowing that the price difference is 21.83% higher than the AMD card, the difference in performance between the two a + 6.53% revenue compared to the NVIDIA graphics card based on performance / value.
