Metroid, my apologies for the slight jab in my previous post.

I think it's great that Sapphire took the plunge to release a unique product for Grin while others are running away from GPU's.

Nvidia seems to be getting all the attention nowadays and it may get worse in ETH if and when they change to Progpow. I certainly hope that Navi puts AMD and Sapphire back in the race.

Hey VoskCoin, seems you are wrong about the lower rewards; it has been really well the same as before the fork.

While you are correct that the rewards went from 3 to 2 coins per block, remember that the block time has also changed from 20 seconds down to 12 seconds. So unless you are mining just for one block and then turning off your gpu, the returns are the same. ie. 2/3 rewards x 20/12 time is approximately unchanged.

Yes, I think I am one of the thousands of fools who got suckered in by SQRL and their FPGA hype. The only use I get out of my acorn is as an expensive paperweight.

Takes a long time to understand Grin as by design, it tries to address many of the deficiencies of today's cryptocurrencies.

To answer your question of why mine c31 on a gpu, it's very simple ... ARBITRAGE!  You point your GPU resources where it has the  highest leverage or the highest potential returns ... which is exactly what GPU miners do jumping from algo to algo or coin to coin.

c31 has 2 operating modes, mean (big memory for GPU) and lean (smaller memory for ASIC's). The return on c31 (basic-friendly algo) is much higher than on c29 (asic resistant algo) because of the network hash rate. I remember it was 15x more hashing power on c29 than on c31 meaning there is less competition on c31 and therefore higher rewards.

What is particularly interesting is that the returns on c31 increase about 1% a week and the returns on c29 fall 1% a week. My conclusion after doing the math is that c29 will be unprofitable soon whereas c31 becomes more profitable.

The other point to note is that all the grin return calculators seem to be incorrect as they do not model the return dynamics over time and are simply ports of existing calculations using todays static return.

Don't confuse age with ability. I have an old mustang what will destroy my 911 on the track (with some supercharger and suspension enhancements).

Grin mining for Cuckatoo31 or c31 on GPU's favors large frame buffers. It's been designed to run in 12 or 14 gb memory which is why the sapphire board trying to target. Otherwise you have deploy a memory paging or solver overflow routine which decreases fidelity (real life performance).

From what I've seen in the labs, the sapphire board outperforms the 1080ti mining C31 undermining pool conditions even though 1080ti reports 10% better performance.

Fwiw Metroid, this time you might be the troll.

Grin has been designed to support GPU with a phase-in of ASIC over the span of 2 years. I have several GPU's switched over to mining Grin and also one of the Sapphire cards.

ROI is 40 weeks or so as I running in in C31 which is the ASIC friendly algorithm. It's very ironic that a GPU is making money using the Grin ASIC branch.

Here's a breakdown of the algorithms and phased rewards. Yes it's in geek talk but these who understand will figure out how to make money.
https://www.grin-forum.org/t/on-dual-pow-graph-rates-gps-and-difficulty/2144