Nvidia Quadro RTX 4000 Mining: Hashrate | Ethereum | Benchmark | Profit (Review)

Gaming Performance parameters of NVIDIA Quadro RTX graphics cards: TLOPS

All Pascal video cards, with the exception of the GP100, suffer from problems with half (FP16) and double (FP64) computational precision: the performance here does not at all correspond to the level required for tasks based on these types of computations. Turing cards, even of the highest class, also have limited processing power in terms of double precision, but in terms of FP16 they have reached a qualitatively new level: 14.2 TFLOPS RTX 4000 cards already give a noticeable advantage in the corresponding tasks.

AMD Vega graphics cards have been offering high performance half-precision computing for two years now, but in deep learning tasks, they have a hard time competing with RTX cards due to the lack of Tensor cores or any equivalent. AMD says that similar technologies will also be implemented in their future GPUs, so we are waiting for the answer from the “red” team.

Currently, NVIDIA RTX technology supports a number of software packages that we are just getting started with. This review includes benchmarks showing the advantages of the Turing architecture itself, but as for the capabilities of the Tensor and RT cores themselves, they will be analyzed later in our next reviews.

Test PC configuration and test suite

Next, we will look at the results of testing video cards. Our test suite includes a wide variety of scenarios – from rendering to computational tasks – and contains both synthetic benchmarks and tests with real applications, for example, from Adobe and Autodesk.

We have selected seven graphics cards for this review, but the main focus will be on comparing Quadro cards of two neighboring generations: P4000 and RTX 4000. It will also be interesting to compare them with the AMD Radeon Pro WX 8200 graphics card, which went on sale last fall at a comparable price ($ 999) …