The announced limitation for mining in the NVIDIA GeForce RTX 3060 and the possible launch of a revision of the rest of the RTX 3000 range, is something that has caught many by surprise since we did not expect it, but this limitation did not it's a hardware change, which makes us wonder if it is possible to reverse the limitation imposed by NVIDIA.
For some, Jen-Hsun Huang's company is taking advantage of the mining bubble to get more out of the situation, for others, it is a maneuver they are making to protect their market share and core business. Whatever your point of view regarding NVIDIA's decision, there is something we all wonder about. Can the RTX 3060 situation be reversed when it comes to mining?
NVIDIA didn't have time for another solution
While a change in a program is done in a short time, one in the hardware takes a long time to implement, not only due to the fact that it requires changes in the design of certain elements of the GPU, but also requires going through again. the entire period of mass prefabrication in which the physical errors of the chips are corrected, as well as the internal organization, is rethought to achieve higher processor and manufacturing performance.
The recent mining bubble has caught NVIDIA by surprise, as demand for its cards quickly began to be absorbed by cryptocurrency mining farms. This has led NVIDIA to have to find quick solutions to the problem, ruling out the creation of new hardware from the outset and having to pull the software part, specifically the BIOS / Firmware combination and the card drivers.
Because NVIDIA, or any other manufacturer, cannot know how long a bubble like the one that is affecting them lasts, they cannot develop hardware to solve a problem that when the card is released on the market. Although they also cannot predict exactly how long it may last.
Why is RTX 3000 wanted for mining?
The answer to this question is not obtained by observing only the mining ratio, but by observing the mining ratio by the energy consumption of the card. Cryptocurrency mining emulates the search for minerals or oil in the sense that each iteration of the mining algorithm is increasingly difficult to break, requiring more and more power and thus more energy.
A change that NVIDIA made in the GeForce Ampere architecture was the addition of 16 ALUs in FP16 in each sub-core of the SM, which work in a switched way to the 16 ALUs of integers. But, in the most famous Ethereum mining algorithms 32-bit floating-point units are used in SMs all the time, giving it a much higher throughput rate than the previous generation, NVIDIA Turing, which was ignored by miners by hardly increasing Pascal's mining rate per watt.
NVIDIA's solution? Cut the Ethereum mining rate in half, for this the driver detects certain common attributes of the mining of said cryptocurrency and cuts performance. How? We do not know 100%, but everything indicates that they prevent the use of the second set of ALUs in FP32 of the Ampere architecture when mining.
0 Comments