To begin to understand these two graphics cards more in depth, we must understand that they provide their GPUs in the first place, which as we well know are the heart of each of them.
Afterwards, the rest of parameters such as VRAM or its speed and capacity will be merely testimonial compared to the power they face, but at the same time on many occasions they are decisive to make a difference, a really strange relationship in this range.
RX 5700 vs RTX 2060 SUPER: war climbs a notch
In these two graphics cards we find the best of both AMD and NVIDIA in terms of architecture, since in the case of the first one you get the new RDNA 1.0, and in the case of the second it maintains Turing as the main asset.
|AMD RX 5700||NVIDIA RTX 2060 SUPER||AMD RX 5700 XT|
|Architecture||RDNA 1.0||Turing||RDNA 1.0|
|Lithographic process||TSMC 7 nm||TSMC 12 nm||TSMC 7 nm|
|GPU||Navi 10 Pro||TU106-410-A1||Navi 10 XT|
|Die size||251 mm2||445 mm2||251 mm2|
|Transistors||10.3 billion||10.8 billion||10.3 billion|
|SM / CU||36 CU||34 SM||40 CU|
|RT Cores||No||3. 4||No|
|L2 cache||4096 KB||4096 KB||4096 KB|
|Base Clock||1465 MHz||1470 MHz||1605 MHz|
|Game Clock||1625 MHz||No||1755 Mhz|
|Boost Clock||1725 MHz||1650 MHz||1905 MHz|
|VRAM Clock||1750 MHz (14 Gbps)||1750 MHz (14 Gbps)||1750 MHz (14 Gbps)|
|Bus||256 bits||256 bits||256 bits|
|VRAM bandwidth||448 GB / s||448 GB / s||448 GB / s|
|FP32 performance||7.95 TFLOPS||7.2 TFLOPS||9.75 TFLOPS|
|PCIe version||PCIe 4.0 16X||PCIe 3.0 16X||PCIe 4.0 16X|
|TDP||180 watts||175 watts||225 watts|
|Feeding||8 + 6 pins||8 pins||8 + 6 pins|
|NVLink / SLI / Crossfire||Crossfire||No||Crossfire|
|Launch price||$ 349||$ 429 ($ 375 currently)||399 dollars|
Where we will not find similarities is in their lithographic processes, since the AMD card obtains the latest and newest node of 7 nm manufactured by TSMC and that gives life to its Navi 10 Pro GPU, while in the case of the NVIDIA card it is also TSMC in charge of providing said TU106-410-A1 chip, but at 12 nm, so AMD is clearly imposed in this section.
The lithographic process logically influences the size and transistors of each GPU. Thus, the RX 5700 achieves a smaller size of 251 mm2 in front of 445 mm2 of the RTX 2060 SUPER for almost the same number of transistors: 10.3 billion for AMD and 10.8 billion for NVIDIA.
This leaves us a similar amount for a difference in size of the 77.29% between them, evidencing the importance of the lithographic leap between them.
Maximum equality in your specifications
Starting with the specifications section itself, we find a count of 36 CU for the RX 5700, which its competitor RTX 2060 SUPER for its part achieves 34 SM.
The number of minimum units yields a total of 2304 shaders vs 2176 shaders, so the AMD card manages to prevail over the paper with 5.88% more cores. The TMU count also favors Lisa Su’s card over Huang’s, due to the higher number of shaders and the ratio these two units must follow.
Thus, the RX 5700 is positioned with 144 TMUs compared to 136 TMUs of the RTX 2060 SUPER. Where, curiously, there is a total equality is in their ROPs, since both carry 64. This is due to certain limitations that RDNA still carries from its previous sister architectures, and it is assumed that RDNA 2 will end this.
Regarding RT Cores and Tensor Cores as units of AI and Ray TracingAMD lacks them on this architecture, but NVIDIA achieves 34 and 272 drives respectively, therefore the reds card cannot perform either of the two tasks used for accelerated DeepLearning or ray tracing.
Another curiosity is that they share the same L2 cache size, 4096 KB, therefore the assignments and information exchange between the GPU, the VRAM and the PCIe will have the same capacity on both chips.
The frequencies will be decisive
Regarding the speeds carried by both GPUs, we must emphasize the fact that AMD has included one more frequency to its cards. So now it has Base, Game and Boostwhile NVIDIA only has the first and last.
The curious thing about this new speed from AMD is that its cards, in practice, fail to maintain any of the three named frequencies, generally staying in a range between Game and Boost.
NVIDIA, on the other hand, achieves the opposite, it always achieves a higher frequency than announced in Boost. This is important to take into account if we specifically look at these frequencies, since the RX 5700 achieves 1465 MHz in base, 1625 MHz in Game and 1725 MHz in Boost, while the RTX 2060 SUPER achieves 1470 MHz and 1650 MHz respectively.
In any case, both cards use GDDR6 with the same capacity (8 GB) and speed (14 Gbps or actual 1750 MHz). They also share data bus (256 bits) and therefore the result is that their bandwidths are exactly the same, 448 GB / s.
The differences are finally beginning to show
Although as we have seen they share many of their characteristics, the reality is that both cards are in different steps if we look at their FP32 performance. The AMD RX 5700 achieves 7.95 TFLOPS, while the RTX 2060 SUPER achieves 7.2 TFLOPS
This performance is not determined by the different versions of PCIe that both cards support, since the AMD GPU is compatible with PCIe 4.0 16X, while NVIDIA’s achieves PCIe 3.0 16X. And it is that if it were really a determining factor, the double performance achieved by version 4.0 compared to 3.0 would leave totally different TFLOPS.
Where, curiously, we will find a difference is in the consumption and power connectors, since the RX 5700, although it has a much more advanced node, achieves a TDP of 180 watts through an 8 + 6-pin system.
The RTX 2060 SUPER is curiously more efficient as it achieves a TDP of 175 watts with only one 8 pin connector.
Finally, we have prices, a determining factor in any current graphics card due to the rise faced by both companies. The AMD card is offered from $ 349while NVIDIA’s was launched at a price of 429 dollars, which has now dropped to an unofficial point among the $ 375 and $ 400 on average.
Performance and value
The differences between the two cards are quite small in terms of performance, but the reality is that currently in global terms, the NVIDIA option is a step above that of AMD in the vast majority of games.
Regarding the ratio performance / dollarAfter the price adjustment of the NVIDIA option, this fight has become quite equal, where it practically depends on the model we choose, it will decline to one side or another.
Therefore, the choice of the card is crucial to be able to give a clear winner due to everything explained, which will also depend on the tastes or priorities of loudness or temperature of each user and team.