I could not find a satisfactory answer for my question by searching the forum.
I am farming 230 TB bladebit c7 plots (actually 3.000 compressed plots. My box has intel i7 10700F 128 GB ram and rtx 3060 12 GB. The syatem is running without any issue with windows 11 pro.
I want to add 3.000 c7 plots (about total 230 TB) more. Is it possible ? Can my pc handle 460 TB. 6.000 C7 plots?
I appreciate any help.
Maybe somebody has a similar farm. Please let me know how big is your Bladebit C7 farm if your system has intel i7. I know that GPU is the primary workhorse.
I have zero experience with BB.
However, one way to estimate where your farm is right now and by how much it can be expanded is to look at your proof times. If you want to double the size, I would multiply your current avg proof time by 2x or maybe 2.5x to get a guestimate where you may end up. As long as that average is below 8 secs, you should be good. Although, some (most?) pools accept proofs with more than 8 secs delays, so that may not be a hard stop.
You could also check your GPU utilization. It should be some sort of square wave form. As long as your peaks are shorter than your valleys, you should have the room to upgrade. If your current peaks are close or more than 50%, you will saturate your GPU.
Lastly, if you are tight with your calculations, you could drop some portion of your plots to c6 level to give your GPU some break. So, just monitor your farmer / harvester every time you add few HDs.
2 Likes
There is also a proof of space tool that you can use to test your system and see how much it can farm.
Proof of Space might be the name for the Gigahorse tool, but Bladebit also has a tool to check with. Don’t forget to take into account the filter halving.
1 Like
You will get no problem farming 6000 k32 c7 bb plots with rtx 3060. (I’m under debian but that should be similar).
1 Like
Bladebit has a built in simulator, no ide why the command is not there on the github page but I once wrote it all down here:
It’s old though…not sure anything changed
| Basic Example (using CPU) | ||
|---|---|---|
| bladebit_cuda -t 2 simulate --no-cuda -n 1000 -p 8 /mnt/ssd/plot-k32-c07-2023-03-06-17-07-5599dc157e404e3c2246fd0464b742012933e9a3d52bb7698e51b75c64739be5.plot | ||
| Basic Example (using GPU) | ||
| bladebit_cuda simulate -n 1000 /mnt/ssd/plot-k32-c07-2023-03-06-17-07-5599dc157e404e3c2246fd0464b742012933e9a3d52bb7698e51b75c64739be5.plot | ||
| -t (threads) | No. of threads per context | p * t <= total system threads. So with a 2 core 4 thread cpu you have the options for -t and -p: 1:4, 2:2 , 4:2 More contexts increases memory usage, more threads can improves max latency. Note that -t must come before simulate everything else behind it. |
| -p (parallel contexts) | No. of parallel contexts | |
| -n (number of iterations) | -n 1000 is recommended | |
| -f (filter setting) | default is 512 (current filter), -f 256 would simulate the result after the first filter reduction described in CHIP-12 | |
| – power (duration in seconds) | These two are used together to simulate a fictitious farm of x size for y seconds. e.g., adding --power 120 --size 1PB to the command, simulates a farm of 1PB for a duretion of 120 seconds. |
|
| – size (farm size to simulate) | ||
| Full example (using CPU) | ||
| bladebit_cuda -t 2 simulate --no-cuda -n 1000 -p 8 --power 360 --size 400TB -f 512 /mnt/ssd/plot-k32-c07-2023-03-06-17-07-5599dc157e404e3c2246fd0464b742012933e9a3d52bb7698e51b75c64739be5.plot | ||
| Full example (using GPU) | ||
| bladebit_cuda simulate -n 1000 --power 360 --size 400TB -f 512 /mnt/ssd/plot-k32-c07-2023-03-06-17-07-5599dc157e404e3c2246fd0464b742012933e9a3d52bb7698e51b75c64739be5.plot |
2 Likes