Got the answer else where. Is it true we need a 4core cpu as a minimum or a dual core can do the job{have Intel cpu g4400 @3,3ghz.will it work with C31 or need to buy other cpu? Thanks… Getting everything ready for testing
Hi Tromp,
Sorry to belabour this point, i have been reading this page for an hour and playing with some numbers and would like to know if I am on the right track to understanding how to do the calculations on where to assign my rtx 2080 ti at any given moment .
Using the example from above ( block height 12,008) for target difficulty, here are my variables I will use :
GPU speed(GPS31): 1.7
weight (C31): 7936
target difficulty: 84490
so i get:
(1.7 / 42) x 7936 / 84490 = 0.0038018587716776800
you called this a reward rate in the original post, does that mean i get approx 0.0038 found blocks per second as my average reward rate at that point in time?
If i haven’t gone off the rails yet would it be fair to say that:
1 / 0.0038 (reward rate per second) = 263.03, meaning i would have had an average wait of 263 seconds for a block to be found (at block height 12,008)?
Thanks in advance…
2 Likes
Yes, all your conclusions are correct.
1 Like
Thanks for looking that over, much appreciated, I am sure you are very busy in these days!
I was hoping you could also confirm that if I were substitute these values below into the formula above
GPU speed(GPS29): 6.3
AR Scale: 1856
(6.3 / 42) x 1856 / 84490 = 0.0032950645046751100
if that is correct, can i compare that rate of blocks found per second (GPS29) directly to the figure in my previous example (GPS31) ?
Or is there something else that needs to be considered like the 90% / 10% split of blocks being issued?
That is correct too.
No; the reward splitting is already reflected in the value of AR Scale, which you used to arrive at your reward rate.
I think i may have overlooked something, i noticed on grinscan, when looking at block # 0, the ar scale was 1856, then it started to drop with every block,
block # 1… 1840
block # 1,000… 1388
block # 12,008… 503
block # 15,000… 874 (it’s gone up?)
if i am not mistaken, it looks like i should have used the figure 503 as the ar_scale when calculating my reward rate for block 12008.
If that is correct what about the weight figure for GPS31? Does 7936 stay constant? You wrote that, over the course of 31 weeks the weight31 will also drop linearly. Does that hold true on the floonet or is it using a different schedule? Can i get this weight 31 figure somewhere?
I have the feeling of not being able to touch the bottom of the lake with my toes right now, can you clarify?
Oops; you’re right. I failed to check your claim that
Indeed that should be 503.
And yes, the weights for Cuckatoo stay fixed until a graph size gets phased out.
Floonet and mainnet behave the same. You can find the weight function source code at https://github.com/mimblewimble/grin/blob/master/core/src/consensus.rs#L164-L177
ah, ok, and for the figure of weight for gps 31 ? Is it still a constant 7936? I am not sure where the floonet is in it’s evolution? As of right now is it still using the 7936 weight(31) ?
Can the figure for current weight(31) be looked up anywhere?
It’s in my very first message
yes i saw that, but later on you said that
i was wondering if the current value is still 31 * 2^8 or has it began dropping on floonet?
If yes, where can the current value be found?
I also wrote
It hasn’t been a year yet…
ok, thanks,
as far as i can tell, the current value is not published on GrinScan nor GrinExplorer.
Seems like a valuable piece of the puzzle, do we really have to work it out manually to have access to this?
I’ve made a little tool for you to calculate whether to mine C29 or C31:
8 Likes
GTX 1080ti here but I’m always getting Status: Errored with Cuckatoo31 using grin-miner v0.5.2.
Jan 15 07:17:13.081 DEBG Mining: Plugin 0 - Device 1 (GeForce GTX 1080 Ti) Has ERRORED!
Reason: Device 1 GPUassert: out of memory /home/grin/grin-miner/cuckoo-miner/src/cuckoo_sys/plugins/cuckoo/src/cuckatoo/mean.cu 429
Tue Jan 15 07:22:57 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 410.78 Driver Version: 410.78 CUDA Version: 10.0 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce GTX 108... On | 00000000:01:00.0 On | N/A |
| 0% 47C P8 17W / 180W | 263MiB / 11175MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 1 GeForce GTX 108... On | 00000000:02:00.0 Off | N/A |
| 0% 44C P8 16W / 180W | 16MiB / 11178MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 2 GeForce GTX 108... On | 00000000:06:00.0 Off | N/A |
| 0% 43C P8 16W / 180W | 16MiB / 11178MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 3 GeForce GTX 108... On | 00000000:0A:00.0 Off | N/A |
| 0% 41C P8 16W / 180W | 16MiB / 11178MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
Try to disable device 0. Your GUI enviromnet consumes 263MiB VRAM but the miner wants all…
You have too much memory in use for the default compile of
build_cuda_target("${AT_MEAN_CUDA_SRC}" cuckatoo_mean_cuda_gtx_31 “-DNEPS_A=130 -DNEPS_B=84 -DPART_BITS=1 -DEDGEBITS=31”)
in
grin-miner/cuckoo-miner/src/cuckoo_sys/plugins/CMakeLists.txt
But if you lower either NEPS_A or NEPS_B by 1 it might fit in the 3 cards that wasted 16MiB.
Is GPU memory speed or GPU core clock speed play a more important role when it comes to mining on C31?
I’m looking to optimise my GPU due to heat issues.
I was trying with Device 1 which was using 16MiB.
@tromp thanks will try that.
Core clock is more important than memory clock for Cuckatoo (and Cuckaroo). But watch your wattage. Therefore will be some crossover point where power consumption goes through the roof. try undervolting as well.
Is the weight for a Cuckatoo32 solution submitted to Cuckatoo31+ the same as a Cuckatoo31 solution? Or does it get a 2x higher weight?