Version 3 introduces new compression levels (for GPU mining only):

Compression Level	C30	C31	C32	C33	C34	C35	C36	C37	C38
Plot size (GiB)	47.51	44.93	42.30	39.65	36.96	34.36	31.71	28.96	26.16
Reward	213%	226%	240%	256%	275%	295%	320%	350%	388%
Capacity per RTX 4090 (PiB)	4.24	3.34	2.33	1.49	0.966	0.572	0.307	0.185	0.106
Capacity per GTX 1080 Ti (PiB)	0.661	0.509	0.359	0.235	0.147	0.0853	-	-	-

Old (v2) compression levels are still supported by the new client (both plotting and mining).

Compression Level	C10	C11	C12	C13	C14	C15
Plot size (GiB)	70.8	64.9	60.8	57.3	53.8	50.3
Reward	143%	156%	167%	177%	188%	201%
Capacity per RTX 4090 (PiB)	73.3	23.5	13.7	7.83	4.06	1.60
Capacity per GTX 1080 Ti (PiB)	24	9	4.4	2.5	1.2	0.5

Embedded GPU miner uses CUDA and works with nVidia GPUs. Minimum required GPU driver version is 440.33 (for Linux) and 441.22 (for Windows). GPU miner has no special requirements for system RAM or PCIe lanes (mining GPU can be connected via x1 (one lane) PCIe, but plotting GPU needs x16 PCIe for best performance).

Compression Level	C30	C31	C32	C33	C34	C35	C36	C37	C38
Min GPU memory (GiB)	1.33	1.88	2.57	3.57	5.04	6.94	10.07	13.90	21.36
Recommended GPU memory (GiB)	4.73	4.80	4.93	5.77	7.62	9.20	11.37	16.40	21.74

Compression Level	C10	C11	C12	C13	C14	C15
Min GPU memory (GiB)	0.13	0.25	0.84	1.31	2.30	4.49
Recommended GPU memory (GiB)	1.02	1.03	1.54	2.59	4.20	6.82

The client tries to use all available GPUs for mining and to allocate recommended amount of video memory in each GPU. Recommended amount depends on GPU model and can differ from values presented in the table above. You can choose particular GPUs for mining, or limit video memory for each GPU using "-g" option. When memory is limited, performance is degraded (20%–40% worse performance for minimum memory compared to recommended).

The client runs benchmark at startup and shows how many plots of each size can be safely used on your PC. By default, C30 plots are generated. You can select different compression level using "-c" option before start plotting.

CPU plotter (works without GPU) is still available for C30–C38.

Minimum required RAM for plotting is 3.7–6.0 GiB (depending on compression level and plotter type), but maximum performance is achieved when the following amount of RAM is available:

Compression Level	C10	C11	C12	C13	C14	C15	C30–C38
Max RAM (GiB)	118.8	114.8	106.7	106.7	102.7	102.7	164.7

Note, that C30–C38 require 164.7GiB of RAM for full-in-ram plotting, therefore a temporary file (321GiB in size) is now needed when plotting new compression levels on machines with 128GiB of RAM, although only ~52GiB is written to the temp file for each plot.

Use "--no-temp" option on machines with lots of RAM to disable temporary files and to instruct the plotter to always allocate maximum amount of memory.

GPU plotter with "--no-temp" option works 1.5–2.5 times faster on Linux than on Windows.

Output	RAM Disk	HDD
GPU plotter	1m	4m
CPU plotter (Intel 13900K)	3m	5m

On machines with less amount of RAM some temporary data will be written to temporary files. Plotting time for such machines directly depends on the number of connected HDD:

Number of HDD	RAM (GiB)	Plot time (275MiB/sec HDD)	Plot time (200MiB/sec HDD)
4	64	7m	9m
	32	12m	17m
	16	20m	28m
	8	26m	36m
3	64	8m	11m
	32	18m	25m
	16	27m	37m
	8	34m	47m
2	64	18m	25m
	32	30m	42m
	16	43m	1h
	8	53m	1h 13m
1	64	1h 5m	1h 30m
	32	1h 40m	2h 20m
	16	2h 5m	2h 55m
	8	2h 25m	3h 20m

GPU plotter additionally requires 3.1–6.6GiB of video memory (depending on compression level). "-x" option can be used to select exact GPU for plotting and, optionally, to set video memory limit for the plotter.
If "-x" is specified, the client allocates video memory for the plotter first, then for the miner (on the same GPU, if this GPU is used for mining too).
If "-x" is not set, the client tries to use first GPU in the system for plotting and allocates memory for the miner first, then for the plotter.

If you have small amount of video memory and struggle to activate the plotter and the miner on the same GPU, set plotter video memory limit to the minimum, as it usually has little impact on the performance.

On 24GiB GPUs like RTX3090 or RTX4090 the client by default allocates 21.75GiB of VRAM for mining, leaving no VRAM for the plotter. Use "-x" or "-g" option to overcome this if you need to use the plotter too.

Although CPU plotter is almost as fast as GPU plotter, the latter is much more energy efficient and should be prefered.

This version of the client recognizes old C1–C5 plots and can mine on them (using CPU), but can't produce them.

Main options

-a <address>	Chia address for receiving reward. Can be omitted to disable mining. -a xch1d4...pao0wa
-d <path>	Directory where plots and temporary files will be stored. -d C:/Plots -d "E:/My Plots" This option can have additional parameters separated by comma: Maximum amount of disk space to use. Valid units are: B, KB, MB, GB, TB (bytes, kibibytes, mebibytes, etc): -d,5TB <path> -d,1600GB <path> Maximum number of plots: -d,100N <path> Allowed file types: -d,t <path> -d,tp <path> where t - temporary files p (or f) - plots Disallowed file types: -d,~t <path> -d,~tp <path> Read-only specifier: -d,r <path> No modifications are made in read-only directories, they are used for mining only Multiple parameters can be specified like: -d,8TB,50N,tp <path> Use "-d" option multiple times to specify more than one directory: -d <path1> -d <path2> Using more than one temporary directory per HDD is strongly discouraged. Plotter always places exactly one temporary file in each directory where temporary files are allowed. The size of each temporary file is 321GiB/N where N is the number of temporary files.
-c <level>	Compression level (10-15 or 30-38) for new plots (default is 30).
-m <size>	Amount of RAM the client should use during plotting. If not specified, the client will try to use as much RAM as possible.
-w <name>	Worker name displayed on the website
-s <path>	Directory for settings and cache files. By default, $APPDATA/NoSSDChiaPool is used. Specify current directory "." to make client portable. You can run client from readonly or nonpersistent file system, but you should always ensure that settings files are located in writable and persistent directory.
-g <number> or <gpu_name>	GPU to use for mining, can be a number or any part of a name or any part of a serial number, optionally includes memory limit: -g,5G 2080 -g,8G 3070 -g 4090 Multiple mining GPU can be specified.
-x <number> or <gpu_name>	GPU to use for plotting, can be a number or any part of a name or any part of a serial number, optionally includes memory limit: -x,5G 2080 -x 4090 Only one GPU can be used for plotting
--no-temp	Disable temporary files and allocate maximum amount of RAM for best plotting performance.
--no-benchmark	Do not run benchmark at startup.
--no-mining	Disable mining. Has the same effect as omitting "-a" option.
--no-gpu-mining	Disable GPU mining.
--no-plotting	Do not make new plots.
--no-gpu-plotting	Disable GPU plotting.
--no-stop	Suspend (don't stop) the plotter when output directory is full, resume when there is enough free space for new plot.
-v	Verbose mode. For verbose level 2 use "-vv" and so on. At verbose level 2 the client prints detailed information about signage point processing, which can be useful for resolving mining issues.

Performance tuning

--r-threads <count>	Number of reading threads. This option switches the client to the special mode for high latency storage (like cloud storage). Best range is 32-256 depending on farm size.
--p-threads <count>	Number of plotting threads. By default equals to the number of logical CPU cores.
--m-threads <count>	Number of mining threads (for C1-C5 plots). By default equals to the number of logical CPU cores.
--no-rescan	Do not rescan directories for new plots.
--rescan-interval <seconds>	Set custom rescan interval in seconds.
--low-priority	Lower mining threads priority (for C1-C5 plots). Use this option if mining disturbs other applications too much. Can increase probability of stale shares.

Network and updates

-u <policy>	Automatic client update policy 0 - turn automatic updates off 1 - critical updates only (protocol changes, etc) 2 - security updates 3 - performance improvements 4 - any new version (default)
-4 -6	Use IPv4 or IPv6 (by default both are used)
--no-ntp	Do not check system clock.

Maintenance

--emulate-filter <value>	Emulate plot filter for testing purposes. Valid values are: 512, 256, 128, 64, 32. The client still mines correctly with this option, but the load on the HDDs and GPUs will be increased to reflect the emulated filter value.
--check-plots	Test all plots for errors.

Delegated plotting

-p [password]	Password used for delegated plotting. The client will show public key corresponding to this password at startup. Also, if there are untested plots with this public key, they will be tested and private keys will be embedded. If "-p" option is specified without following password, the client will ask for it at startup. Ignore this option if you make plots for yourself and not using delegated plotting.
-k <publickey>	Public key used for delegated plotting. New plots will be generated for this public key. Only who knows the password for this public key will be able to mine on those plots. Ignore this option if you make plots for yourself and not using delegated plotting.
--skip-test	Embed secret keys into plots with public keys without testing. Use this option only if you already know that plots are correct.

The client requires correct system time during mining. If time deviation is more than a minute, the client will be completely unable to communicate with the server.

Chia is a Proof-of-Space coin, therefore it is mined on HDDs. First, a program called "plotter" is used to generate large files called "plots" (this process is called "plotting"). Then, plots are used to mine Chia. More plots you have - more Chia you get. This pool uses custom plot format and custom plotter. All necessary software is embedded in our client program and all required operations are done automatically.

After being downloaded, configured, and launched, our client starts to make plots.

Our plotter is highly optimized and tries to utilize all available resources to produce plots as quickly as possible. Conventional plotters use only one big temporary file during plotting. This file is usually located on SSD because very fast access is required. Our plotter, in contrast, uses multiple temporary files and places them on the same HDDs where plots will reside. Combined speed of those HDDs is enough to completely eliminate need of any SSD. To speed up plotting even further, as much RAM as possible is used to hold temporary data. This allows quick plotting even if only one HDD is used. On systems with 128GiB of RAM or more, all temporary data will fit in RAM. In this case maximum performance is achieved, and number of HDD does not matter.

Number of HDD	RAM (GiB)	Plot time (275MiB/sec HDD)	Plot time (200MiB/sec HDD)
3	64	30m	40m
	32	35m	50m
	16	45m	1h
	8	50m	1h 10m
2	64	40m	50m
	32	45m	1h
	16	1h	1h 20m
	8	1h 10m	1h 40m
1	64	1h 5m	1h 30m
	32	1h 40m	2h 20m
	16	2h 5m	2h 55m
	8	2h 25m	3h 20m

Most HDDs are faster at the begining of the disk and slower at the end. For best performance all temporary files should be located in high speed area of the disk. This will be achieved without any effort in most cases when empty HDDs are used.

Our plotter produces intermediate form of plots first, called "sparse" plot. Sparse plots are faster to generate and suitable for mining too, but 13% larger than final plots will be. After filling all directories with sparse plots, the client starts to convert them into final ones. This is called "finalization" and takes 15–30 minutes for each plot. During finalization some disk space is freed, allowing to generate more sparse plots. This process can repeat multiple times. After everything is done there should be no more than one sparse plot in each directory.

Our plots in their final form are 16–23% smaller than standard 101.3 GiB plots currently used by other pools or for solo mining and generate 20–30% more reward per GiB.

Compression Level	1	2	3	4	5
Sparse plot size (GiB)	96.1	94.1	92.1	90.1	88.1
Final plot size (GiB)	84.5	82.9	81.3	79.7	78.1
Extra reward (%)	20	22	25	27	30

By default, compression level 1 is used and 84.5 GiB plots are generated, but you can choose any compression level you want before start plotting. Be careful, plots with higher compression level require more CPU work during mining for decompression. The client runs benchmark at startup and shows how many plots of each size can be safely used for mining on your PC (assuming 90% CPU load).

Compression Level	1	2	3	4	5
Proof time (ms)	32	65	135	300	750
Number of plots	940000	550000	220000	80000	22000

Using less number of plots will reduce CPU load proportionally. Particular compression level should not be used if corresponding proof time is greater than 4 seconds. If you want to minimize CPU load as much as possible, use lower compression levels.

Mining starts as soon as first sparse plot gets ready. Each plot generates approximately one share in 24 hours. Reward is divided among all participated miners proportionally to the amount of submitted shares. Standard mining fee on this pool is 3.5%.

With our client it is possible to delegate plot generation to another person without giving him the ability to mine on generated plots:

1. Add "-p" option to the client command line and launch it. The client will ask you for a password, come up with some password and enter it. Alternatively, a password can be specified directly in the command line after "-p" option (less secure).
2. The client will show public key based on that password. Give this public key to the person who will make plots for you. Do not tell him the password!
3. That person, then, should specify your public key in his client using "-k" option, and generate required number of plots.
4. After you get your new plots, add new plot directory in the client options if necessary, launch the client with "-p" option again and enter the same password.
5. Look into plot summary. New plots should appear like this:
   
   100 plots (100 untested,CL3) with public key 1a8hmt14cghiw7m2kq...
   
   Check plot count, compression level (CL) and public key (note, that only the beginning of public key is displayed).
   If public key differs from requested one, or no public key is shown, then it means that plots were not generated specially for you!
   New plots will be tested and then private keys will be embedded into them. Testing large number of plots can take a lot of time. To avoid unnecessary delays, untested plots participate in mining too. Do not forget your password until all plots are tested and private keys are embedded or you will not be able to mine on your plots. Always use strong password. Weak passwords like "123" can compromise this scheme in two different ways: First, a person who make plots for you can guess the password and start mining on your plots. Second, if you and someone else use the same password, a person who make plots can notice it, because public keys will be equal too, then he can cheat and give both of you the same set of plots.

-a <address>	Chia address for receiving reward. Can be omitted to disable mining. -a xch1d4...pao0wa
-d <path>	Directory where plots and temporary files will be stored. -d C:/Plots -d "E:/My Plots" This option can have additional parameters separated by comma: Maximum amount of disk space to use. Valid units are: B, KB, MB, GB, TB (bytes, kibibytes, mebibytes, etc): -d,5TB <path> -d,1600GB <path> Maximum number of plots: -d,100N <path> Allowed file types: -d,t <path> -d,tsf <path> where t - temporary files s - sparse plots f - final plots Disallowed file types: -d,~t <path> -d,~tsf <path> Read-only specifier: -d,r <path> No modifications are made in read-only directories, they are used for mining only Multiple parameters can be specified like: -d,8TB,50N,ts <path> Use "-d" option multiple times to specify more than one directory: -d <path1> -d <path2> Using more than one temporary directory per HDD is strongly discouraged. Plotter always places exactly one temporary file in each directory where temporary files are allowed. The size of each temporary file is 225GiB/N where N is the number of temporary files. When there is an uncompleted plot, and the number of temporary directories changes, the plot is deleted and new plotting begins.
-c <level>	Compression level (1-5) for new plots (default is 1).
-m <size>	Amount of RAM the client should use during plotting or finalizing. If not specified, the client will try to use as much RAM as possible.
-w <name>	Worker name displayed on the website
-p [password]	Password used for delegated plotting. The client will show public key corresponding to this password at startup. Also, if there are untested plots with this public key, they will be tested and private keys will be embedded. If "-p" option is specified without following password, the client will ask for it at startup.
-k <publickey>	Public key used for delegated plotting. New plots will be generated for this public key. Only who knows the password for this public key will be able to mine on those plots.
-s <path>	Directory for settings and cache files. By default, $APPDATA/NoSSDChiaPool is used. Specify current directory "." to make client portable. You can run client from readonly or nonpersistent file system, but you should always ensure that settings files are located in writable and persistent directory.
-u <policy>	Automatic client update policy 0 - turn automatic updates off 1 - critical updates only (protocol changes, etc) 2 - security updates 3 - performance improvements 4 - any new version (default)
-4 -6	Use IPv4 or IPv6 (by default both are used)
--p-threads <count>	Number of plotting and finalization threads. By default equals to the number of logical CPU cores.
--m-threads <count>	Number of mining threads. By default equals to the number of logical CPU cores.
--r-threads <count>	Number of reading threads. This option switches the client to the special mode for high latency storage (like cloud storage). Best range is 32-256 depending on farm size.
--no-benchmark	Do not run benchmark at startup.
--no-plotting	Do not make new plots.
--no-finalizing	Do not finalize existing sparse plots.
--skip-test	Embed secret keys into plots with public keys without testing. Use this option only if you already know that plots are correct.
--check-plots	Test all plots for errors.
--low-priority	Lower mining threads priority. Use this option if mining disturbs other applications too much. Can increase probability of stale shares.
--no-rescan	Do not rescan directories for new plots.
--no-ntp	Do not check system clock.

The client requires correct system time during mining. If time deviation is more than a minute, the client will be completely unable to communicate with the server.