Bandwidth, an important factor when mining Ethereum
The reason why the AMD RX 5700 XT is superior to the RX 6700 XT in mining is because of the VRAM bandwidth. We cannot forget that while the AMD RX 5700 XT uses a 256-bit GDDR6 bus, in the case of the RX 6700 XT its bus is 25% narrower with only 192 bits under the same type of memory.
That is why at the level of mining Ethereum , the most powerful graphics card by AMD is not precisely an RDNA, but the almost unprecedented Radeon VII that with its bus of more than 1 TB/s of bandwidth via HBM2 memory exceeds far in VRAM bandwidth to even the RX 6900 XT.
So the advantage is not architecture but memory bandwidth. Although this begs the question: what about the Infinity Cache? Theoretically it gives RDNA 2 based GPUs an advantage in bandwidth and especially in consumption, which in theory should be beneficial for mining.
The Infinity Cache is completely useless to miners, being a Victim Cache it does not participate in the VRAM data capture process , where the process is the same in both CPU and GPU. The work of the Infinity Cache is to adopt the cache lines and therefore the data discarded from the second level cache of the GPU itself, so that when looking for them the GPU does not have to access the GDDR6, which which allows access using less power and thereby raising the clock speed margin.
So not only does the RX 6700 XT not provide any advantage when it comes to mining Ethereum, but the same can also be said for the rest of the AMD RX 6000 range.
Consider the capabilities of the RX 6700 XT in cryptocurrency mining using different algorithms, the optimal overclocking performance and video card settings for different algorithms, profitability and consumption.
Characteristics of the RX 6700 XT graphics card
| Specifications | RX 6700 XT |
|---|---|
| Graphics Card | |
| Release Date: | Mar 3rd, 2021 |
| Generation: | Navi II (RX 6000) |
| Graphics Processor | |
| GPU Name: | Ships 22 XT |
| Architecture: | RDNA 2.0 |
| Render Config | |
| Shading Units | 2560 |
| TMUs | 160 |
| ROPs | 64 |
| Compute Units | 40 |
| SM Count | — |
| Tensor Colors | — |
| RT Cores | 40 |
| L0 Cache | 32 KB per WGP |
| L1 Cache | 128 KB per Array |
| L2 Cache | 3 MB |
| L3 Cache | 96 MB |
| Clock Speeds | |
| Base Clock | 2321 MHz |
| Game Clock | 2424 MHz |
| Boost Clock | 2581 MHz |
| Memory Clock | 2000 MHz 16 Gbps effective |
| Memory | |
| Memory Size | 12 GB |
| Memory Type | GDDR6 |
| Memory Bus | 192 bit |
| Bandwidth | 384.0 GB/s |
| Board Design | |
| TDP | 230 W |
| Suggested PSU | 550 W |
| Power Connectors | 1x 6-pin + 1x 8-pin |
