NVIDIA claims the CMPs will not affect the supply of GeForce cards and now we know why. VideoCardz reports references in the newly released GeForce 461.72 driver confirm that the first CMPs are not based on the new Ampere GPU architecture.
The CMP 30HX and 40HX, the first NVIDIA mining processors to hit the market, are based on the previous-generation Turing architecture. The former is based on the TU116 and the latter is a derivative of the TU106. The chips are made on a 12nm process from TSMC, so at least in terms of foundry capacity, there's no impact on the production of the Ampere GPUs. The newest generation from NVIDIA uses Samsung's 8nm node.
Turing-based mining processors are not ideal as these chips are less profitable than Ampere -- they have a higher power consumption and the hash rate isn't impressive either. It seems like a tough sell as long as better alternatives are available.
Will it help?Whether the CMP will become a success remains to be seen. NVIDIA expects CMPs will pull in only about $15 million in revenue this quarter, which is insignificant versus the current volume sold to miners. The only way the firm can push miners to these chips is by artificially limiting the mining performance of its GeForce GPUs.
The hash rate limiter of the GeForce RTX 3060 is a first step but its success will depend on how easily this protection can be hacked. According to NVIDIA, it's a very secure system, but we wouldn't be surprised if someone comes up with a software or hardware-based solution to circumvent the anti-mining algorithm.
Future NVIDIA GeForce cards are expected to feature similar anti-mining protections. At the moment, it's unknown whether the green team will take measures to make current Ampere-based GeForce cards less attractive to miners. NVIDIA will not limit the mining performance of existing cards -- that wouldn't make sense anyway as miners would simply refuse to install these drivers or firmware. But there are some rumors that NVIDIA could reintroduce the current GeForce lineup under new SKUs.
Making CMPs on the 12nm node may help a bit to get more GeForce cards in the hands of gamers, but we still think both lines are competing for the same resources. Foundry capacity is not the only constraint at the moment, there are also reports about shortages of GDDR6 memory, substrates, and various other components that are used by both products.