–
If you were hoping that AMD would add some protection against cryptocurrency to graphics cards, we have bad news. According to the company, it must be enough that the optimization of the RDNA 2 architecture for the game use of the new graphics makes it less suitable for crypt mining.
Before and after the release of GeForce RTX 3060, it was addressed many times in the media and on forums performance limitation in cryptocurrency mining, which Nvidia implemented (but then it turned out that it only works against the “coin” of Ethereum and also did not monitor the distribution of drivers and released a version where the limiter was more or less deactivated). In this context, questions arose as to whether AMD would deploy something similar to Radeon graphics to protect the gaming market from hungry cryptographers. Looks like we have an answer now – negative.
PCGamer asked Radeon Product Manager Nishe Neelalojanan about the possibility that the Radeons would also get a mining performance lock during a conference call to grafikám Radeon RX 6700 XT. However, he found that this card did not have limited cryptocurrency performance. And it won’t be, AMD reportedly doesn’t plan anything like that.
According to the employee, AMD will not block mining, but also no other burden. Unfortunately, we can’t expect any action that could alleviate the demand from the miners a bit (although the question is how effective it would be at all, given how it turned out for the RTX 3060).
–
AMD Navi 22 GPU (right) compared to Navi 21 chip (left) Zdroj: AMD
–
Gallery: AMD Radeon RX 6700 XT: revelations and official benchmarks
RDNA 2 is less suitable for mining than Ampere
Neelalojanan, however, also remarked on the matter that a certain small deterrence may already be that Radeon RX 6000 graphics with RDNA 2 architecture they are much more optimized for games than for computing loads.
There is a grain of truth to this, because despite past experience, it is between RDNA 2 and Ampere architecture from Nvidia in some things really looks a bit like the situation in, say, the 2014 between GCN and Nvidia Maxwell, but vice versa. One thing that is particularly important for cryptocurrencies such as Ethereum is memory bus and throughput. And now it’s Nvidia who has chosen a strategy of a wider bus, faster memories (in the case of GDDR6X) and thus higher throughput.
In contrast, AMD uses a narrow memory bandwidth for RDNA 2 chips with no high throughput (384 GB / s RTX 6700 XT, 512 GB / s RX 6800 to RX 6900 XT), but this is compensated by the large Infinity Cache in the GPU. But the key is that this cache works for games, while cryptocurrencies like Ethereum are deliberately (a design decision to prevent the development of specialized mining ASICs) dependent on large memory capacity and throughput. As a result, Radeons RX 6000 have a relatively lower extraction performance than GeForce graphics of the same category (Navi 21, for example, reaches just over 60 MH / s, while Ampere GA102 in the full RTX 3090 to 120 MH / s configuration). Navi 22 should again be worse than the GA104 in the RTX 3060 Ti and RTX 3070 graphics, which the Radeon RX 6700 XT competes with, making it less attractive to miners. Of course, when Amperes are unavailable, they will buy worse alternatives, which probably also applies to the RTX 3060.
–
Infinity Cache in RDNA GPU 2 04 Zdroj: AMD, via Hardwareluxx
–
The Radeon RX 6700 XT is even tested to have lower performance in Etherea mining than the older generation Radeon RX 5700 XT – this is because it has 192-bit memory with a throughput of 384 GB / s, while the RX 5700 XT with 256-bit bus gave a throughput of 448 GB /with. And mining more or less scales with gross physical permeability. Unlike games where Infinity Cache works and significantly reduces memory requirements.
In the end, this may be an effective sub-measure against the interests of miners, as it is a physical limitation that cannot be “hacked” or bypassed by mining the mining algorithm. It must be said that although AMD presents this as a form of protection against miners instead of targeted measures, it is certainly not a ready-made strategy, but only a by-product of the RDNA 2 GPU architectural solution, which today has this positive aspect.
Ideally, AMD would support this natural constraint with some “DRM” Nvidia-style mining, provided it is more successful in real enforcement. However, the question is whether certain real obstacles do not prevent this. When the limiter announced to Nvidia, we wrote that an RTX 3060-like approach that would force a signed driver might not be used by AMD because it would break the operation of Radeon GPU drivers for Linux. It is not clear whether this has contributed to the company’s current position, or whether AMD would be able to find a solution. But this complication could probably be one of the factors.
Infinity Cache in RDNA 2 GPU
Source: PCGamer
