I too picked up a Z420. Great machine. I tried Ubuntu but as a first-time user, I switched to Windows. But made the mistake of plotting 500 plots in Windows 10 before I realized they were not farming properly. Now I am going back to Ubuntu to see if I can get some compressed plots going. I would love a quick video starting at a fresh Ubuntu install to plotting and copying to a TrueNAS NAS where my main Windows 10 PC is farming on. Great work!

@DigitalSpaceport

9 ай бұрын

That video is soon friend. They JUST released 2.0.1 that fixed Linux GUI plotting (had been broken plots there also) so its possible to do now.

Been struggling with Linux and I need from start to finish to get it working on my Z840. If you have the time please make a Linux install, Chia install for dummies guide!!!

Thank you for all the info you provide to the community. A full guide how to setup standalone bladebit cuda (without a full chia 2.0 package install) on a Ubuntu would be so so so helpful. It is a nightmare to try and do it if your not supper familiar with Linux or a developer. Between trying to compile Bladebit and installing Cuda tools and SDK one can go crazy.

Yes, please do a full setup guide for Linux server!

I would love a thorough guide for bladebit on Ubuntu linux. I am new to Linux and am struggling. I am using the P4 GPUs on a r720xd

The man with the plan....DS is on the scene.

Ive been using 'plow' to move plots easily. What I would like to know is: Minimum specs for VM to run main node and farmer node. In my case, GPU's are located on Harvesters, so I don't expect to need much resources for the fullnode and farmernode. Looking to put them on a HA/FT ESXi VM.

Yes please do a Linux HP Z420 Chia 2.0 GPU Plotting BETA release of the Bladebit GUI software setup video. I'm currently running Widows line command giga-horse setup on the HP Z420 I basicly followed your video for that.

Nice demo thanks! would like to see how lower end gpus do too, tho official chia seem to recommend at least a 3060, the Raspberry Pi guys seem to be falling further and further behind :D ( they are not mentioned so much any more :D but unfortunately less democratised space means less decentralisation ultimately). Interested to hear if you have thoughts about the second order impacts of shifting to GPU plotting e.g. network becoming more enriched in k32. Foolishly I plotted the majority on my farm at higher k33 and k34 to 'future proof' so I would like to see tools to convert existing plots to 'decompressed' format. Keep on chooglin!

@DigitalSpaceport

9 ай бұрын

Good call I have a 1070 I can plop into a machine here for another video. They just released 2.0.1 for linux fixed so might wrapup the tutorial series with that card and hit on a few points not addressed by me so far.

Congratulations for the video. For now with windows I create textures in 4.50 minutes with 128gb ram and rtx 4090 and double NVME (no raid). Do you recommend NVME RAID? Last thing, for moving to the final hard disk, how can I decrease the times? Moving the plot via lan (i have 1gbs)? Or with simple robocopy? Thank you very much

@DigitalSpaceport

9 ай бұрын

NVMe in a raid0 would help stripe writes and provide some endurance cushion, yes. You need 10gbe or bond multiple 1gbe nics into a LACP link.

thanks for the video , i just like using a mouse in windows

which nvidia driver do u have installed ? i cant get lower than 12 minutes per plot ;/ and i cant find out why

Is there any hint on a bladebit ramplot mode working with 128GB RAM coming eventually or is there some hard limit preventing this from happening?

Do you have a video or other resources that explains a little more why swap is bad to have on while plotting? You mention it at around 13:10 in this video.

@DigitalSpaceport

9 ай бұрын

Oh it's in general bad if it starts running as going to disk vs ram adds extra context time and of course speed is drastically lower. It's usually okay as plotting stays off the swap due to size, but if it does spill into swap, times will tank.

Yeah... I'm not sure what the hang up is for Linux. Even if you're not going to make "the switch", you should be using a dedicated machine for Chia anyway. And that is a perfect time to experiment with an os that you're not %100 familiar with. You won't have to worry about killing your Windows system. Anyway, great video. Any inspiration I can get to fuel my desire to tinker is welcome.

@DigitalSpaceport

9 ай бұрын

Thats my thinking also for a great time to learn Linux

You're great, just to like. Excuse me, I am a household computer with 128GB of memory. Do I still need to sacrifice some of my hard drive for caching? What if we only use 128GB of memory without using a hard drive?

@DigitalSpaceport

9 ай бұрын

You HAVE TO use a temp drive for 128GB plotting. Also you need to have it be seperate dedicated drive and high endurance like used enterprise NVMe is good idea.

@user-mc8iy9su4t

9 ай бұрын

@@DigitalSpaceport Thank you very much. By the way, what is your opinion on the NOSSD C15 level compression plot。

you recommend us on gtx 1070 back then, whats the gpu recommendation right now?

@DigitalSpaceport

9 ай бұрын

3060

As always very well made video, watched it twice already :) But quick question, does bladebit take any dev fees? If not then I'll be replotting. Thx.

@saurishg

10 ай бұрын

there is no dev fee for bladebit

@winreboot

10 ай бұрын

@@saurishgSweet, I'm replotting my plots now :)

For chia farming on the compressed plot, what cpu and GPU requirements? I am just using a g6400 cpu with 50 tb non compressed plots which is running fine .

@DigitalSpaceport

9 ай бұрын

8GB VRAM Cuda 5.2 (so like a 1070 or greater GPU) and down to 16GB ram (but you need really fast high endurance temp SSD/NVMe the lower the RAM goes)

Here is my test results. Hp z620 on Ubuntu with e5 2620 v2, 128gb 1333, 3070 8g, 2xToshiba 480g mlc sata as raid 0. It’s took 12 minutes for compress 5 plot

@DigitalSpaceport

9 ай бұрын

Yeah the disk IO hits a bit with just 2 in raid0, 4 would be ideal but those Toshiba are great drives all the same.

I have just try the 1660ti with 6gb ram but I got CUDA out of error

why can't the CPU compressed plotting be 256GB of ram, but instead still needs 416GB?

@DigitalSpaceport

10 ай бұрын

Not a question I can answer. Ask it maybe during upcoming AMA on 6th. StorageJM will be there.

Great video thanks alot ...can i plot c7 using 64gb ram using ubuntu?

@bgrngod

10 ай бұрын

I've been attempting to use the BB 3.1.0-beta1 to do C5's on Ubuntu with 32GB RAM and the --disk-16 command with no luck. I get a fatal error about being out of memory.

@saurishg

10 ай бұрын

got it i am waiting for 64 gb cuda plot.i hope bb updates soon for 64 gb ram@@bgrngod

Honestly how far off is the windows 128gb gpu plotting? Not sure I want to redo everything to linux.

Make tutorial for windows 11 sir Please

Hi Mr. Digital Spaceport ,I finally tried Ubuntu with my Ryzen9-5950X with 128 GB ram. I also have 2TB drive Samsung 990 Pro as my temp. The plotting time is around 6 minutes if I copy them to the same nvme ram that I have. But I want to transfer these to my Seagate HDDs from 2 years ago. They are not as fast as before . In best scenario I get 200 MB/S copying speed and slows down also when the drive is closer to be full. I used a bash script that can copy from SSD to HDD when the .plot file is available but the copy speed will be eventually be slower that plot generation speed and as the result my 2TB fills out before the script has a chance to copy all the plot files. Do you have any solution for me? I don't have fast network speed or using Raid with my HDDs.

@KamiMountainMan

10 ай бұрын

btw direct plotting to the HDD takes 12-13 minutes per plot.

@KamiMountainMan

10 ай бұрын

For now I came with this solution. I think it stops plot creation if I am low in space :#!/bin/bash source_folder="/home/Chia" c="" f="" threshold=100000000 # Function to handle Ctrl+C (SIGINT) function handle_ctrl_c { echo "Ctrl+C pressed. Exiting..." exit 1 } # Set up a trap to catch Ctrl+C and call the handle_ctrl_c function trap handle_ctrl_c INT while true; do max_free_space=$threshold free_space=$(df -kP "$source_folder" | awk 'NR==2 {print $4}') if [ "$free_space" -gt "$max_free_space" ]; then sudo ./bladebit_cuda -c "$c" -f "$f" --compress 7 -n 5 cudaplot --disk-128 -t1 "$source_folder" "$source_folder" else echo "Waiting for 10 seconds before checking again..." sleep 10 fi done

@DigitalSpaceport

9 ай бұрын

Stagger your writes off your temp to multiple HDDs at the same time. Create plots into seperate folders and then move scripts for each folder to each HDD and 3-4 at a time should saturate it.

Damn, chia is still trying to be a coin? Why?

@DigitalSpaceport

9 ай бұрын

I feel like you can try better and get a better burn then that. Give it some effort!