AUTOLYKOS_TUNING.txt

Team Red Miner Autolykos2 (ERGO) Mining
=======================================
This document provides some quick pointers on how to tune the
autolykos2 algo used by ERGO.

v1.1 2022-09-13
v1.0 ?

General background
==================
Autolykos2 is a memory-intensive low/medium power algo. However, with
the small memory accesses involved the algo behaves more like algos
like verthash rather than ethash. Performance is tied to the core clk,
and for max speed (especially for Vegas) core clk needs to be higher
than ethash to support driving the mem controller on the gpu(s).

This algo accesses mem in 32 byte chunks. This means that RDNA
generation gpus (Navi, Big Navi) will not perform well. Their 128 byte
cacheline size means that 128 bytes are read for every 32 byte
request, effectively halving the available memory bandwidth compared
to GCN (which uses 64 byte cachelines).


Pad prebuild
============
Autolykos needs a big pad filled with random data specific for every
block height. This means that every 120 secs (on average) the
algorithm will switch to the next pad. Earlier TRM versions did a
complete stop to rebuild the pad, which lowered the effective hashrate
over time. In v0.10.3, pad prebuild was added. It is enabled by
default for all gpus, and will try to consume as much memory as
possible for a second copy of the pad buffer. The miner will then
prebuild the next pad concurrently with hashing, and in most cases it
will be fully ready when the next switch takes place.

During the prebuild phase, hashrate will drop slightly, and power
consumption will go up. To control this, the prebuild speed can be
controlled using the argument --autolykos_prebuild. The range is 0-9,
when 0 means no prebuild, 9 means max speed. You can pass one argument
for all gpus, or a comma-separated list per gpu. The max speed will be
aggressive with the pad rebuild, raising power temporarily, but the
chance that the pad is fully ready as the next switch takes place is
higher. We believe the default value of 4 is a good choice.


Polaris Tuning
==============
Polaris gpus are simple for autolykos2. We have not spent a lot of
time on tuning, so the examples below should be seen as a starting
point, there might be better combinations of core clk, mem clk, mem
straps to find.

- Quality ethash timings work well.

- Mem clk should be high, existing ethash config is fine.

- Core clk is a big factor for the hashrate.

- In our Polaris tests, a Nitro 470 4GB (Elpida), Nitro+ 570 8GB
  (Samsung) and Nitro+ 580 8GB (Samsung) all displayed identical
  hashrates for the same core clk as long as memory bandwidth was
  sufficient. The 580 will rebuild the table slightly faster though,
  hence produce a slightly better avg hashrate over time.

Polaris tuning examples
-----------------------
Note: sensor power reported, not accurate.

Type           GPU CUs CoreMHz MemMHz TEdge  VDDC   Power
Nitro+ 570 8GB   0 32  1200    2080   42C    875 mV  73 W
Nitro+ 470 4GB   1 32  1235    2000   46C    875 mV  59 W
Nitro+ 580 8GB   2 36  1275    2080   40C    900 mV  80 W
----------------------- GPU Status -------------------------
GPU 0 [42C, fan 44%]       autolykos2: 64.24Mh/s
GPU 1 [46C, fan 44%]       autolykos2: 66.09Mh/s
GPU 2 [40C, fan 44%]       autolykos2: 68.19Mh/s


RX Vega 56/64 Tuning
====================
For historical reference, RX Vegas used to be outperforming other gpus
on autolykos (relatively speaking), being able to reach a full 200
MH/s. This was due to a TRM-specific optimization that only worked
well before the autolykos pad started growing. In today's autolykos,
RX Vegas typically run in the 120-140 MH/s range depending on tuning.

The tuning examples below are the same as when RX Vegas were running
at higher hashrates. They are still applicable, but should be seen as
starting points. There might be better tunings available among the community.

RX Vegas are always slightly complex to tune well. Mining distros can
help greatly here. We discuss three different tunings. Background
info:

- Mem timings should be used, ethash timings of some sort are good
  choices. Other timings for Equihash, Cuckoo or CN can produce good
  results as well. You can check the TRM discord for more tuning
  examples.

- Mem clock does not need to be high unless you're aiming for the
  highest hashrates.

- The most efficient setups makes sure the soc clk stays at a lower
  level, and maximizes the mem clk for that level, i.e. sets it to the
  soc clk frequency. 

- Your _effective_ core clk will decide your hashrate. Vegas are
  notorious for not running at the configured frequency when AVFS
  p-states are used.


RX Vega Simple Tuning
---------------------
This is for people who don't care about soc clk level and just want to
start hashing at a decent level around 120-135 MH/s. 

1. Set ethash mem timings (see our ethash guide for examples).
2. Set core clk to 1250 MHz
3. Start with mem clk at 960 MHz (Vega 64) or 847 MHz (Vega 56).
4. Set voltage to 875mV.
5. Run the miner. Check the hashrate.
6. Increase core clk until you hit your target hashrate. If you hit a
   bottleneck where increased core clk doesn't boost the hashrate,
   increase mem clk a little more. Repeat from 4.
  
7. If you crash, bump voltage a little more. Repeat from 4.
8. If you run stable for a while, lower voltage.


RX Vega Efficient Tuning
------------------------
This tuning is a more efficient approach than the simple tuning above,
maxing out the potential hashrate at 960 MHz (Vega 64) or 847 MHz
(Vega 56) mem clk. For Vega 64, flashing a Vega 56 bios will likely be
the best choice, but it isn't as critical as for ethash mining. The
goal is to stay at soc clk 847 MHz for Vega 56 (or Vega 64 with
flashed 56 bios), and soc clk 960 MHz for Vega 64s. You might need to
lock p-state levels using OverdriveNTool (Windows), mining distro
helpers, or sysfs controls (Linux).

Note 1: Vega 56 Hynix can follow the same guide below, but can end up
at a slightly lower hashrate. You can then switch up to 960 MHz soc
clk level, following the Vega 64 guide below instead. You can keep the
mem clk lower than 960 MHz though, depending on what hashrate you'd
like to target.

Note 2: if none of the above doesn't make sense to you, the critical
piece of information here is that RX Vegas can't use a mem clk higher
than the current soc clk. However, a higher soc clk means a more power
hungry gpu, meaning we can't lower voltage as much as we'd like or the
gpu will crash. Finding the sweet spot soc clk level, and maximizing
the use of it by setting mem clk equal to soc clk is important when
optimizing for efficiency.

 1. Configure ethash timings.
 2. Vega 56: Use core p-state 2: set to 1225 MHz.
    Vega 64: Use core p-state 3: set to 1225 MHz.
 3. Vega 56: Use mem p-state 2: set to 847 MHz.
    Vega 64: Use mem p-state 3: set to 960 MHz.
 4. Set voltage to 850mV as a start.
 5. Lock core and mem p-states.
 6. Run the miner. Press 's' and verify that the soc clk is at 847 MHz
    (Vega 56) or 960 MHz (Vega 64).
 7. Hopefully you'll reach your target hashrate (120-135 MH/s) and we're done.
 8. If not, increase core clk slightly. Repeat from 6.
 9. If you crash, increase voltage.
10. If you've run stable for a longer period, try lowering voltage.


RX Vega Max Performance Tuning
------------------------------
This tuning targets 155-160 MH/s. Power draw will be higher.  For Vega 64,
flashing a Vega 56 bios will be the best choice here as well, but it
isn't critical. For this tuning, we just go with the highest p-states.

For Vega 56 with Samsung mem, if you have applied timings that can
reach 53-54 MH/s when mining ethash, then keep them.

Note: for Vega 56 Hynix, the guide below can still be followed, but
the target hashrate for us had to be lowered slightly.

1.  Configure ethash timings.
2.  Use core p-state 7: set to 1550 MHz as a start.
3.  Vega 56: Use mem p-state 3: set to 990 MHz if you can run ethash at 52-54 MH/s.
    Vega 56: Use mem p-state 3: set to 950 MHz if you can run ethash at 50 MH/s.
    Vega 64: Use mem p-state 3: set to 1107 MHz.

    NOTE: if your gpu can't take the high mem clk values suggested
             above, set it to the level you can mine ethash at.
            
4.  Set voltage to 925mV as a start.
5.  Lock core and mem p-states.
6.  Run the miner. Check the hashrate.
7.  As long as you're underperforming the hashrate target, keep raising
    the core clk. Under plain amdgpu-pro on linux, the scaling is
    absurd and you might have to increase up to 1650-1700 MHz before your
    true effective clock is around 1400 MHz. Windows does not scale as
    aggressively.
8.  If you crash, increase voltage.
9.  If you continue to crash even with 950mV, you might need to give up
    and settle for a lower hashrate target with a lowered mem clk.
10. If you've run stable for a longer period, start lowering voltage as
    much as possible, in small steps.


Radeon VII Tuning 
=================
Radeon VIIs perform reasonably well, but are limited by core clock and
the resulting high power usage.  Typical VIIs can expect to hit between
210-240 MH/s on air cooling, and up to 270MH/s on liquid cooling.  To
reach the highest hashrate and efficiency you will need to run in linux
using the same setup procedure to run ethash C mode as described in 
the ETHASH_TUNING_GUIDE.txt (linux kernel params + running as root).
Using a mining distro that already supports the changes needed for TRM
ethash C mode will be the easiest option. We briefly discuss some high
level VII tuning concepts below:

- Mem timings are NOT important.  VIIs will very much be bottlenecked
  on core clock, and memory tuning does not need to be pushed.

- Mem clock can be significantly lowered to save power and keep the 
  HBM2 cool.  Even at high hashrates, memory clk can usually be dropped
  to around 750MHz.

- The limiting factor in hashrate will be core clk.  This in turn will
  be limited by the cooling of the card.

- The TRM 'VII Boost' enabled by the ethash C mode procedure described 
  above will increase hashrate by around 10% at the same core clock.

Radeon VII tuning examples
--------------------------
These are rough examples that should serve as a good starting point for
tuning.
Note: sensor power reported, not accurate.

Setup   CoreMHz SocMHz MemMHz  VDDC    Power   Peak Hashrate
Linux*  1500    971    801     850 mV  127 W     175.5Mh/s
Linux*  1700    971    801     935 mV  144 W     200.0Mh/s
Windows 1500    971    801     850 mV   -        <test not performed>
* - Linux tests performed with kernel params set as described for
      ethash C-mode (or R-mode).

Navi GPUs
=========
As stated above, Navis simply won't do well on autolykos2 due to
architectural changes that don't work well with the smaller mem
accesses. Therefore, we don't expect RDNA gpus to run this algo. This
guide will be updated if any serious improvements for RDNA/RDNA2 gpus
are implemented.

For tuning, you can use an existing configuration for ethash as a
starting point, then lower the core clk about -10%. 

Example tunings:
Type    GPU CUs CoreMHz SocMHz MemMHz TEdge TMem  VDDC   Power
5700XT  0   40  1150    1085   912    41C   70C   737 mV  80 W
5600XT  1   36   950    1266   910    40C   70C   800 mV  93 W
------------------------ GPU Status ---------------------------
GPU 0 [41C, fan  0%]       autolykos2: 106.2Mh/s
GPU 1 [40C, fan 49%]       autolykos2: 81.50Mh/s

Type    GPU CUs CoreMHz SocMHz MemMHz TEdge TMem  VDDC   Power
RX6800  0   60  1075    685    1049   52C   76C   787 mV 116 W (voltage not tuned) 
------------------------ GPU Status ---------------------------
GPU 0 [52C, fan 28%]       autolykos2: 118.8Mh/s