Skip to main content

AMD RDNA Mining GPUs Could Be on the Horizon

Radeon RX 5700 XT Reference Design
(Image credit: Future)

Patch notes from a set of fixes for the AMDGPU Linux kernel driver have revealed the possibility of AMD working on mining-specific graphics cards based on the RDNA1 architecture, as reported by Phoronix

More specifically, the kernel driver patch notes specify changes to the NV12 SKU. This SKU in particular is very interesting as it's a headless graphics card using the Navi 12 core that does not support Video Core Next (VCN). In layman's terms, this means AMD's NV12 graphics card would  not support video capabilities at the driver level.

This isn't the first time we've heard of a headless AMD graphics card. Late last year, the same AMDGPU Linux kernel had data on a Navi 10 part with its video and display outputs disabled at the driver level. 

Both of these headless graphics cards point to the possibility of AMD building a lineup of graphics cards dedicated to mining cryptocurrency. Nvidia recently announced Turing-based mining graphics cards of its own. 

AMD's GPU architectures in general have been known to provide impressive hashing rates compared to Nvidia's older counterparts, so a product like this could be huge for AMD. While Nvidia's Ampere GPUs tend to be at the top of the stack for mining, RDNA1 and RDNA2 are no slouch.

Going back to the older RDNA1 architecture could save time, resources, and development costs compared to an RDNA2 solution, especially in the midst of today's graphics card shortage. Besides which, cryptomining doesn't generally benefit from the Infinity Cache used in RDNA2, and definitely doesn't need the ray tracing hardware.

It'll be interesting to see what AMD plans to do around the current crypto craze. If AMD does re-enter the 'mining GPU' market, there could be a new level of competition in the GPU space. Or, it could end up like the Polaris GPUs in 2018, where large shipments arrived right after the bottom fell out of the mining market, leading to a glut of GPUs.