Nvidia presented new mobile video cards GeForce RTX 2050, MX570 and MX550. We partially know their characteristics, and now it’s time for tests.
But let’s start with the details of the characteristics. The GeForce MX 550 is based on the TU117 GPU with 1024 active CUDA cores. This is even more than the GTX 1650 on the same GPU. But a 64-bit bus cut down to a 64-bit bus and a lower core clock (1320 MHz) will probably make the MX 550 slower than the GTX 1650. But the new product should be faster than the MX 450. It is also worth noting that the card has 2 GB of GDDR6 memory clocked at 12 GHz.
The MX 570 and RTX 2050 are very similar. They are based on the GA107 GPU, have 2048 active CUDA cores, a 64-bit bus and memory with a frequency of 12 GHz, only the younger card has 2 GB of it, while the older one has twice as much. In addition, the first GPU frequency is limited to 1155 MHz, while the second is already 1477 MHz. The mobile RTX 3050 offers generally the same options, but it has a 128-bit bus and even higher frequencies. However, even the MX 570 should provide a very decent performance for a junior adapter. In addition, RT kernels and tensor kernels are preserved here, but the source clarifies that the former are somehow limited. Unfortunately, it is not yet clear what this means.
Now for performance. In 3DMark Time Spy, the RTX 2050 scores 3369 points and the MX 550 scores 2510 points. There is no MX 570 result, but it should be close to the first.
In comparison, the mobile RTX 3050 scores 4800 points, while the mobile GTX 1650 scores about 3400 points. That is, the RTX 2050 and MX 570 are roughly on par with the GTX 1650, but may offer ray tracing and DLSS support. The latter is especially important given the relatively low performance of the cards in general.