https://floonet.grinscan.net/block/251966
Congratulations to its miner Wilke Trei (@Lolliedieb) !
Looking forward to first C32 block on mainnet…
7 Likes
Thanks John 
Quite a lot of work - not to mine the block, because the difficulty on floonet is rather low, but to make it work on GPUs at all ^^ But pays off when you finally see it working 
Will tell more about whats behind on Grincon, see you there 
4 Likes
just saw this awesome post. I also tested my c32 miner. and found few blocks:
https://floonet.grinscan.net/block/253514
FYI:
Titan RTX 1.2gps for C32
May I ask for a refresher on when are we going to see C32 on mainet ? is it an immediate or gradual phaseout? and also will 11 gb cards be able to mine С32
Well it already is on mainnet. In fact grin has 34 proof of work schemes active on mainnet, cuckatoo 31-63 and cuckarood 29. But atm only two are used since the other are too difficult.
The need for C32 rather comes from the fact that C31 will fade out (slowly) from January, so the diff there will increase while the one for C32 will remain stable. This will make C32 to be the better pick from March or April on or so.
I am sure a solution to mine it on 11G will be there when its time
Edit: Even 8G is not only feasible but runs already ^^
3 Likes
Thank you so much, it makes sense!
for 8G and 11G card C32 mining, it depends on it support NVLINK or not and performance different.
base on mygrinminer ( my personal private hobby miner ), I have a dual rtx 2080 with NVLINK, it’s share 8g memory across dual card ,so total 16g for C32 , at 0.5 gps per card and 1 gps for two now ( still under optimization ).
but if it’s not support nvlink, like rtx 2070 or 1080 ti, need implement by smean algo, which will be far slower than nvlink setup.
Thank you, this is helpfull. Probably time to trade my 1080 ti for 2080 ti
Decided to join in on the fun and mine a few blocks using the ePIC-Boost-Miner
https://floonet.grinscan.net/block/254190
It’s always supported C32 but the difficulty on mainnet has always been too high. Nice to know there is an option for those who bought the 16GB cards when grin transitions to C32.
that’s pretty fun now.
3 block found by 3 different gpu and miner. let’s guessing what’s the hardware and graph rate and have fun 
as floonet have no long run c32 miner yet, so diff is very low , let’s assume it will finalize any block if solution found.
1st one : GrinScan [floonet] - Block 251,966
2 min 35 sec,
as lolminer is amd and opencl , 0.44 GPS , it should running on vega VII or vega frontier 16G.
2st one : GrinScan [floonet] - Block 253,514
52 sec
it’s mygrin miner on titan rtx 16G, 1.2 GPS
3rd one : GrinScan [floonet] - Block 254,190
5 min
it’s epic miner, on rx570 16g, 0.2 GPS.
Need to correct you wrt the speed ^^
The first floonet block was in fact a Nvidia Tesla P100, running lolMiner on OpenCL at 0.37 g/s
Wrt speeds of my C32 code (atm):
Radeon VII, 16G: 0.6 g/s
Vega 64, 8G: 0.4 g/s
RX 5700 XT 8G: 0.33 g/s
RX 580 8G: 0.2 g/s
Nvidias are rather slower on my code atm. Well its a code in alpha status not yet tuned to the very max ^^
Well, I have 4x2080Tis connected to MB via x1-x16 PCI-e risers. Will my cards mine C32 efficiently being paired in pairs with two NVLINK bridges?
I never tested through pcie 1-16 adapter, but after you pair two gpu together, run this command:
nvidia-smi nvlink -s
if you see nvlink was successfully established, it means it works.
like this:
GPU 0: GeForce RTX 2080 Ti (UUID: GPU-0)
Link 0: 25.781 GB/s
Link 1: 25.781 GB/s
GPU 1: GeForce RTX 2080 Ti (UUID: GPU-1)
Link 0: 25.781 GB/s
Link 1: 25.781 GB/s
Thanx, now it remains only to purchase this unreasonably expensive bridge…
Depends on how you use your rig. My NVLINK bring me 30-50% performance gain in tensor flow when training my parallel DNN model. it will worth it.
But for mining grin, 100$/999$ = 10%, if there are no commercial miner support this nvlink pairing, then you purchasing make no sense, so my suggestion is hold on still 2020 if you only mining.