Topic: PhoenixMiner 6.2c: fastest Ethereum/Ethash miner with lowest devfee (Win/Linux) - page 327. (Read 784857 times)

  Go away, troll. We have kernels for Navi - RDNA isn't that different from GCN. The problem is that the GPUs are "locked down" for ethash mining by either limiting the number of TLB entries, or lowering the page size. This may be unintentional but we suspect that it is by design, probably to prevent miners from buying out all Navi GPUs and leaving the gamers with high prices and low availability. RX5700 is the first very competitive AMD gaming GPU since Polaris, so they probably want to win some market share as far as gaming market goes. We hope that this is not hardware limitation and a future driver will remove it but only time will tell.

   With the current no-so-great mining profitability, they really didn't have to do this but it's absolutely their choice, and with the current prices of Polaris, it is still one of the best cards for ethash mining. If the Navi cards weren't locked down, they would achieve hashrate in the 50 MH/s ballpark. We are still working on some kind of workaround but it doesn't look good at this moment.

  We are considering it but you can achieve the same with AMD Memory Tweak -ref option.


  We didn't notice any changes in power consumption in our testing. Please let us know what GPUs you are using and what is the change in power consumption.


  There are no changes in AMD kernels for this version. However, there are two possible reasons for this behavior:

• The hardware control for AMD cards is now more robust. It wasn't working that well with many driver versions because AMD have changed the API many times in the last year or so, and even removed the driver/VBIOS based target temperature controlled fan option entirely. So we implemented more robust control of the fans and temperatures. Most probably you cards were "overshooting" the target temperature or the maximum fan speed with the old version but 4.5b will enforce these limits more strictly. If you have -tt 63 and -tmax 65, as soon as GPU reach 64-65 C it will start to throttle down to avoid overheating. You need to either increase the target temperature (-tt), the maximum temperature (-tmax), or increase the max fan speed (-fanmax) to avoid throttling. A good way to check which of your cards are throttling is to add the new -hstats 2 option and check the core clocks - if they are lower than the ones you have specified via -cclock, you need to either increase -fanmax, -tt, or -tmax
• The new -mt option is nowhere near that good as good modded VBIOS (or good memory timing applied by AMD Memory Tweak). So, if you have modded VBIOS or are using another way to set memory timings, avoid the -mt option. It is intended only as a last resort, quick and dirty method of increasing performance on non-modded cards. Also, avoid changing the -mt option while the miner is running (by changing config.txt and then reloading it with the c key in the miner console) because this frequently leads to crashes.

  Please let us know the GPUs you are using and what is the difference of the hashrate. Is it only on some GPUs or it is the same across all GPUs?


 Yes, we are currently working on it (for AMD cards) but it may take a few weeks and be released as part of 4.6 (it will be out relatively quickly, as we are moving to faster release cycle to follow the frequent driver releases from AMD).