Resource VideoCardz citing his source, he said that only one model of the NVIDIA CMP HX series of mining cards uses the Ampere graphics architecture and Samsung’s 8nm process technology. We are talking about the senior accelerator CMP 90HX.
According to the source, it uses a PG132 PCB (the same as that of the GeForce RTX 3080) and a GA102-100 graphics chip, but the number of CUDA cores is unknown. The power consumption of this card is 320W. It is equipped with 10 GB of video memory.
The younger models CMP 30HX, CMP 40HX and CMP 50HX use the Turing architecture. Information about the first two was previously confirmed by the latest NVIDIA driver, which contains the IDs of the video cards. As points out VideoCardz The CMP 50HX model is based on the TU102 GPU-100 GPU and PG150 PCB used for the GeForce RTX 2080 Ti. The first two mining accelerators mentioned above will go on sale before the end of the current quarter. The CMP 50HX and CMP 90HX will not arrive until the second quarter of this year.
This is not the first time NVIDIA has released specialized mining cards. About three years ago, the manufacturer released the P104-100 and P106-100 models of the Pascal generation, which were distributed by NVIDIA partners. Mining cards do not have the gaming characteristics of GeForce GPUs.
According to the source’s assumptions, such cards use GPUs with non-working or defective tensor cores, rasterization units, or texture units. At the same time, they are great for mining where complex mathematical calculations are required. In addition, NVIDIA mining cards lack external video interfaces and are often equipped with cheap cooling systems for greater savings in their production.
If you notice an error, select it with the mouse and press CTRL + ENTER.