Compute Client Node Reward Mechanism Update

Render Network
Render Network
Published in
6 min readNov 17, 2023
Update on RENDER Compute Client Node Rewards

This post outlines a mechanism for distributing the the 300,000 RENDER SPL Compute Client node rewards that are reserved in RNP-006 for the first Render Network Compute Client partner, io.net — and are set to be begin distribution through December, 31, 2023. This mechanism will be added to an updated RNP-006 and submitted to the Render Network community for final vote.

Objective

This proposal outlines a system for distributing RENDER tokens to GPUs sourced from the Render Network for use by Compute Clients (“Compute Clients”) as described in RNP-004: Compute Clients.

Qualifying Nodes

Onboarded Nodes that meet a minimum threshold of performance and bandwidth will qualify for rewards in a given Epoch.

To qualify the Nodes, GPUs need to:

  1. Be on the approved list of GPU models in section 6 below,
  2. Have onboarded via www.renderfoundation.com/waitlist/ which specifically includes providing the Render Network with a valid Solana wallet for Reward deposit,
  3. Be live on a Compute Client within the cohort, for example having successfully installed io.net’s client and connected it, and for io.net having correctly identified yourself as a Render Network Node, and
  4. Operate with an average download bandwidth speed of greater than 100Mbps per second and upload speed of greater than 75Mbps.

Please note all 4 requirements need to be met for a Node to qualify for rewards.

Rewards pool

A total of 1,140,852 RENDER has been reserved for allocation as Compute Client Rewards.

For November and December 2023, rewards will be exclusively earned by Nodes available on io.net, thereafter all Qualifying Nodes will be eligible to participate.

The initial Epochs and Reward pools will be as follows:

  • 100K RENDER — 00:00 GMT Monday, November 20 — 23:59 Thursday, November 30
  • 200K RENDER — 00:00 GMT Friday, December 1 — 23:59 Sunday December 31

From 1/1/24, Compute Client rewards will continue on monthly Epochs until 10/31/24.

The Reward Pool for these monthly Epochs will be calculated as the total remaining Rewards divided by the number of months left up to 10/31/24.

By way of example, should 840,852 Rewards remain unallocated as of 1/1/24, the January Rewards pool would be 840,852 divided by 10, for a total of 84,085.2 RENDER.

Note, to avoid edge cases, the available rewards per Epoch will be the lower of the above Rewards, and the product of the number of Qualified Nodes available during the Epoch and 75 RENDER.

By way of example if there are 1,000 qualified nodes in the November Epoch, the Rewards pool would be limited to 75,000 RENDER.

1. Factors for Reward Calculation:

For a given node i, the score is calculated over the following epochs and with the following set of tokens:

  • A (Quantity of Compute — FE): USD earned running Node
  • B (Bandwidth — BW): Data upload and download capability of the node, in Mbps.
  • C (GPU Model — GM): Score indicating GPU’s performance level.
  • D (Uptime — UT): Ratio of active time to total possible active time in Epoch.

2. Weight Assignments:

Specified weights for each factor:

  • Weight for A (FE): W_A = 0.25
  • Weight for B (BW): W_B = 0.35
  • Weight for C (GM): W_C = 0.20
  • Weight for D (UT): W_D = 0.20

3. Formula

The formula for calculating the reward for a node i is:

4. Share of Compute Work done:

This will be calculated as a ratio of the total amount earned running the node compared to the earnings by all nodes over the given Epoch.

The formula for calculating Ai is given by:

where ri is the calculated reward for node i, and rj represents the set of all calculated rewards.

The term ri / max(rj) is a normalization step. It scales the individual reward ri based on the range of all calculated rewards. Here’s how it works:

The numerator ri is the relative position within the range of rewards, and the denominator max(rj) is the maximum reward among all nodes and normalizes this position to a scale of 0 to 1.

Multiplying this normalized value by 0.9 and adding 0.01 shifts the range to start from 0.01 instead of 0, providing a floor value. Ai therefore spans the range [0.01, 1].

This choice of normalization allows Ai to represent the relative standing of a node’s reward within the overall distribution, making the metric comparable across different contexts and avoiding biases introduced by extreme values.

5. Bandwidth Scoring System:

The following bands will be used to classify Nodes:

Should a node, on average during the epoch, exceed both the minimum download and upload requirements for a tier, it will achieve classification within that tier.

6. GPU Model Scoring System:

5. Uptime

Uptime will be measured as the ratio of active time to total possible active time in Epoch. This will factor in partial periods, so a Node joining half way through the Epoch and achieving 100% uptime, would be scored at 50% uptime for the Epoch.

Example

Scenario:

  • Network Size (n): 1,500 nodes.
  • Total Reward Pool for the epoch (November): 100,000 RENDER tokens.
  • Total earnings in USD for the epoch (November): $125k
  • Maximum reward among all nodes max(rj): $2,500
  • Weights: W_A = 0.25, W_B = 0.35, W_C = 0.20, W_D = 0.20.

Node i Characteristics:

  • A (FE): $100 earned by the node
  • B (Bandwidth): Medium speed (score 0.4).
  • C (GPU Model): Entry level RTX 3060 GPU (score 0.6).
  • D (Uptime): 95% (score 0.95).

Reward Calculation for Node

Share of compute work Ai = 0.1 + 0.9 × (100) / (2500) = 0.136

Distribution and Communication

The Render Network will facilitate payments and will make best efforts to do so during the epoch following the current epoch.

Details on how to qualify for Rewards, Rewards earned, and Rewards support will be provided by the Render Network, and shared via official channels.

Compute Client requirements

Compute Clients will integrate with the Render Network via its APIs.

Compute Clients will provide second level Rewards support and are responsible for all technical support.

Participating in the io.net Compute Client

Nodes that are interested in participating in the io.net Compute Client rewards program should register for the Render Network waitlist and sign up to be a node on io.net — following these steps:

During the io.net signup process, nodes will be asked if they are a Render Network node, and they should respond, yes. Only nodes on both the Render Network and the io.net waitlist will be eligible for Compute Client node rewards.

Please see the Qualifying Nodes section earlier in this post for detailed requirements to be eligible for RENDER Compute Client node rewards.

Please follow the Render Network’s official social channels to stay up-to-date on the latest Render Network proposal updates and news:

Official Website: https://renderfoundation.com
Twitter: https://t.me/rendernetwork
Discord: https://discord.gg/rendernetwork

--

--

Render Network
Render Network

https://render.x.io Render Network is the first blockchain GPU rendering network & 3D marketplace.