How to use --tune for RTX 30XX (ethash)?

With version v1.8y4rc2 we modified --tune option.  Now you will be able to adjust two parameters for your GPU (not compatible with the previous v1.8y4rc1 version, see below).

The default values are not yet optimal for everyone. We suggest you to play around with i and d starting with small variations around the defaults.

--tune=iN1dN2 where i stands for intensity and d stands for drag. N1and N2 are numerical values.

  • Intensity is similar to the value that you were using to tune with v1.8y4rc1. However, values are no longer compatible. N1 is an integer and can take values in the interval 0-64.
    i default values are: 15 (3060), 15 (3070 Ti), and 10 (3080 Ti), on Linux. And 19 (3060), 15 (3070 Ti), 10 (3080 Ti), on Windows.
    Default value with driver 460.39/461.40 for RTX 3060 is 43.
  • Drag will balance intensity. N2 is a decimal number and values can vary in the interval 0-100.
    d default values are: 55 (3060), 60 (3070 Ti), 59 (3080 Ti), on Linux. And 54 (3060), 57 (3070 Ti), 55.5 (3080 Ti), on Windows.
    Default value with driver 460.39/461.40 for RTX 3060 is 64.

Note: If you increase PL, usually you’ll need to decrease i value.

Usage is as follows:

  • To apply tune to all GPUs in your rig, with i=17 and d=57:
    miniZ.exe [email protected]:3096 --tune i17d57

    or

    miniZ.exe [email protected]:3096 --tune=i17d57
  • To apply tune with i=17 and d=57 to GPU 3:
    miniZ.exe [email protected]:3096 --tune3 i17d57
    
  • You can also tune only one of the parameters, i or d, and apply the values to all or just one GPU. Similarly to the previous examples:
    miniZ.exe [email protected]:3096 --tune i17

    or

    miniZ.exe [email protected]:3096 --tune3 d57

* Tune with v1.8y4rc1 version *

The following is only valid for the v1.8y4rc1 version. Read above for usage for other versions (from v1.8y4rc2 on).

With v1.8y4rc1 version --tune was working with one parameter. Only the RTX 3060 supported this option.

Default value with recent drivers such as 470.63/472.12 is 6. Default value with driver 460.39/461.40 for RTX 3060 is 38.

We were advising you to lower this value when the miner is not able to get rid of the anti-mining lock.

Usage is as follows:

  • To apply tune with value 5 to all GPUs in your rig:
    miniZ.exe [email protected]:3096 --tune 5

    or

    miniZ.exe [email protected]:3096 --tune=5
  • To apply tune with value 5 to GPU 3:
    Note that in this case you cannot write with ‘=’ sign

    miniZ.exe [email protected]:3096 --tune3 5

 

mt options to adjust memory timings (Linux only)

From miniZ version v1.8y3, there are a few available options that will have impact on performance on some GPUs.

*** The --mt options will only apply settings to Pascal devices. ***

These improvements are usually applied to ethash/progpow algos, however equihash algos may also profit from them.

This option is known to work well with GTX 1080, GTX 1080Ti, and other devices with Hynix memory. Other GPUs may profit from this adjustments but to a much smaller extent.

The --mt options will require super user permissions.

  • --mt-auto: use automatic memory timings (sudo required)

If you are unsure about the timing values, use the --mt-dump option first to get a hint. Lower values usually will give a better performance.

  • --mt-dump: dump memory timings in use (sudo required).

To apply list of memory timings per GPU use --mt. Remember when using the --mt option to always specify the timing pair, with the values separated by ‘.’ such as ‘16.5’ .

  • --mt=[timings] [16-32].[3-9]: Ex. --mt=16.5. Always specify the timing pair. (sudo required)
    If you have multiple devices you can specify multiple values like this --mt 16.5,0,21.4. The 0 means no timings should be applied.

Warning: some values will be invalid for your GPU and may cause the driver/GPU to crash. In this case you will need to restart your rig/system/computer. Use with caution.

 

Option inspired on the work by Tiago Shibata.
https://github.com/tiagoshibata/meth-pills
donate: stratum1+tcp://[email protected]:3333

 

How to use --mode?

In some cases you may want to select a specific mode/kernel for miniZ to run in your GPU. This can happen for example after using --ocX1 to tune miniZ for a specific algo+GPU+OC, or just want to test it in another similar GPU.

At the moment, you will be able to run a mode for your rig, or a specific GPU, by using --mode option. This is still a work in progress. We may find a better way to achieve the same task 🙂

  • To select mode 8 for all GPUs in your rig:
    miniZ.exe --url [email protected]ne-zel-euro.equipool.1ds.us:50053 --log --mode 8

    or

    miniZ.exe --url [email protected]:50053 --log --mode=8
  • To select mode 8 for GPU 3:
    Note that in this case you cannot write with ‘=’ sign.

    miniZ.exe --url [email protected]:50053 --log --mode3 8
  • To select mode 5 for GPU 0, and mode 3 for all other GPUs. The order matters. Ex:
    miniZ.exe --url [email protected]:50053 --log --mode 3 --mode0 5
  • To select mode 8 for GPU 3, and mode 1 for GPU 4:
    miniZ.exe --url [email protected]:50053 --log --mode3 8 --mode4 1
  • To select mode 5 for GPU 3 and for GPU 4:
    miniZ.exe --url [email protected]:50053 --log --mode3 5 --mode4 5

Keep in mind that when you apply OCs to a GPU it is possible that some kernels do become unstable and others do not. But, if a mode is dramatically unstable also after you try it with stock settings, we appreciate that you contact us 🙂 Reporting GPU model, miniZ version, mode in question, and algorithm is very useful in this case. Thanks!

1: Do not use/apply oc1/oc2/ocX (or --mode)at the same time on the same GPU. This could cause confusion in understanding miniZ behavior. One of the options will prevail though, usually the last one appearing in the command line.

For other command line arguments check Usage page.

Happy mining!

 

Argument options for automatic tuning performance and efficiency: --ocX

*** Do not use --ocX when mining with RTX 30XX with anti-mining lock. ***

From v1.5t3 version you can use --ocX for miniZ to automatically choose the best kernel for your GPU.
Available for 125,4 and 150,5,3 algorithms in v1.5t3. Available in all algorithms from v1.5u version.

--ocX  is a new option that runs a few available miniZ kernels and chooses the one that performs best.1 It starts with the one set by default, the one we chose to be the best in stock settings. For a few GPUs the defaults are not optimised because we did not have access to that GPU model for testing. Generally, from each algo table performance you can infer which GPUs we tested.

    • To run ocX you only have to add it to your command line. Ex.:
miniZ.exe --url [email protected]:50053 --log --ocX

If you have more that one GPU, it will start optimisation for all GPUs, asynchronously. It starts mining with default kernels, we called it mode in the messages:

After stabilizing a bit (temperature, performance..) miniZ will start switching modes:

After testing all available modes for that GPU it will stop switching.

Some GPUs may finish later. This will be OK, since the ones that tested everything first will start mining with the mode(s) that performed best. There is no need to stop and restart miniZ.

After the tune is finished, take note of the mode in case you want to use it later (we may improve this in future versions). You can run a mode for your rig, or a specific GPU, by using --mode. This is still a work in progress but we have a separate FAQ for it here.

  • To run ocX on a specific GPU. Ex. apply ocX to GPU 3:
miniZ.exe --url [email protected]:50053 --log --ocX=3

In this case, optimisation will only start for GPU 3, similarly to what is shown above. The other GPUs will just start mining as usual.

  • To run ocX on specific GPUs. Ex. apply ocX to GPU 0 and GPU 3:
miniZ.exe --url [email protected]:50053 --log --ocX=0,3

In this case, optimisation will only start for GPU 0 and GPU 3. The other GPUs will just start mining as usual.

Keep in mind that when you apply OCs to a GPU it is possible that some kernels do become unstable and others do not. But, if a mode is dramatically unstable also after you try it with stock settings, we appreciate that you contact us 🙂 Reporting GPU model, miniZ version, mode in question, and algorithm is very useful in this case. Thanks!

1 --ocX is different from --oc1/--oc2.
--oc1 or --oc2 will set a specific kernel mode for the GPUs. --ocX will select and run a few kernels and let you know which performed best. Do not use/apply oc1/oc2/ocX (or --mode)at the same time on the same GPU. This could cause confusion in understanding miniZ behavior. One of the options will prevail though, usually the last one appearing in the command line.

For other command line arguments check Usage page.

Happy mining!

Argument options to adjust performance and efficiency: --oc1 and --oc2

miniZ includes a command line argument for improving performance at distinct overclock (OC) values.

When you’re adjusting your power limit (PL) and OC setting values, you can use --oc1 or --oc2 to change the default kernel mode for your GPU. The following table (Table 1.) shows oc1 and oc2 support for each combination of GPU and algorithm. --oc0 is default mining mode that each GPU.

--oc1 or --oc2 will set a specific kernel mode for the GPUs. These refer to kernels we noticed that performed best in some OC settings, and for some GPUs. This is why not all GPUs have oc1 or oc2. Sometimes the default (optimised for stock settings) was the best we found.

Recently, from v1.5t31,2 version, you can use --ocX for miniZ to automatically choose the best kernel for your GPU. For more information check ocX related FAQ here.

Usually, at lower PL/OC values --oc1 produces better performance. When you push your GPU to higher PL/OCvalues, --oc2 is likely to be produce higher hashrate. It all depends on your OC settings. 

The best is to try both and check which performs best for your settings and GPUs. For stock settings it should not be necessary to use this option.

When you add --oc1 or --oc2 to your command line this setting is applied to all GPUs. You can also apply the setting to a specific GPU.

Examples for a rig with 6 GPUs, mining Zel on equipool:

1. To apply --oc1 to all GPUs

miniZ.exe --url [email protected]:50053 --log --oc1

2. To apply --oc1 only to GPU#2 and GPU#3

miniZ.exe --url [email protected]:50053 --log --oc1=2,3

3. To apply --oc2 to all GPUs except for GPU#1 and GPU#4. Note that order matters (try it out!).

miniZ.exe --url [email protected]:50053 --log --oc2 --oc0=1,4

4. To apply --oc1 to most GPUs, --oc2 to GPU#0, and default mode to GPU#3 and GPU#4.

miniZ.exe --url [email protected]:50053 --log --oc1 --oc2=0 --oc0=3,4

 

1 Available for 125,4 and 150,5,3 algorithms in v1.5t3. Available in all algorithms from v1.5u version.

2: Do not use/apply oc1/oc2/ocX (or --mode) at the same time on the same GPU. This could cause confusion in understanding miniZ behavior. One of the options will prevail though, usually the last one appearing in the command line.

 

* NOTE: Adding --oc1/2 to the command line does not apply OCs to the GPU, it just makes the GPU run with different kernel functions. When you adjust the OCs in Afterburner (or other program) and run miniZ without --oc1/2 the miner will run with the default kernels (usually optimized for stock settings). The --oc command line option just change the way miniZ runs not the OC settings of the GPU. *

For other command line arguments check Usage page.

Happy mining!