Embedded Files
One RNDR Token is mapped to a unit of work called OctaneBench http://octanebench.com, which is measured in seconds and in concurrent OctaneBench power. Users select from preferences for speed, cost, system hardware, node reputation, and security - and the amount of OctaneBench compute work varies based on a user’s preferences as well as aggregate GPU supply and demand on the network. These preferences are organized into three tiers: Tier 1, Tier 2, and Tier 3. The pricing for each tier is stable in order to allow artists to be able to accurately manage rendering costs and for node operators to operate sustainably, above marginal cost. The tier pricing algorithm periodically updates the amount of compute work a user receives in each tier based on changes in GPU performance, the current price of GPU cloud rendering from other providers, electricity costs, and network supply and demand patterns. For example, adjustments to OctaneBench are made with new OctaneRender releases in order to reflect increased rendering performance as well as the efficiencies of new GPU hardware generations.
RNDR Multi-Tier Pricing (MTP) provides a framework to enable GPU supply and demand to self-organize in the most efficient way possible, ensuring that RNDR is working at peak capacity and there is little wasted GPU power[16]. Tier 3 provides users primarily concerned about price - and who currently do not use GPU cloud rendering networks due to high costs - with a highly economical rendering solution. Tier 3 has no queue priority so reductions in cost come at the expense of speed. Tier 2, which provides queue priority and higher hardware and reputation requirements, enables users to access high quality parallel GPU nodes with minimal delay, offering the potential for significantly accelerated rendering speeds when compared to Tier 3. For artists on a deadline, Tier 2 reduces operational delays like rendering bottlenecks by providing ‘first-in-line’ access to decentralized GPU compute supply on the network. Finally, Tier 1 provides levels of security and pricing commensurate with current GPU cloud rendering costs.
MTP works in conjunction with the network’s Reputation, Allocation and Governance systems to achieve an equilibrium. Multiple price tiers enable the network to meet the needs of artists with different preferences (i.e. speed, cost, or security) as well as creating a mechanism for artists to prioritize these preferences. When a user identifies their preferences by selecting a tier, it helps the network’s allocation process operate more efficiently, enabling better matching of variegated GPU compute supply and demand. For Operators, higher tier work can only be processed with sufficient levels of trust and positive reputation history. MTP therefore incentivizes nodes to maintain a good Reputation and current, up-to-date hardware.
MTP is designed to create a virtuous cycle that maximizes overall network utility. Tier 3 is priced close to operating cost, enabling the network to capture marginal demand. An efficient market occurs when pricing is at marginal cost (P = MC), providing Pareto Optimality to producers and consumers. Thus, Tier 3 generates marginal demand from users who would otherwise forgo GPU cloud rendering due to price distortions. Capturing marginal demand is critical for the network because it increases overall network utilization, enabling node operators to generate positive reputation history needed to process higher Tier 2 work. As more marginal demand is added to the network, increased congestion incentivizes users with preferences for speed or premium hardware to upgrade to Tier 2 with queue priority and demonstrated node reliability.
Because node operators earn more tokens per compute cycle when processing Tier 2 work, the network creates an incentive for node operators to successfully process work at Tier 3 and build a positive reputation history. The process of organizing nodes into tiers based on rendering activity provides the foundations for the network’s Proof-of-Render Governance system. Thus, by capturing marginal demand with Tier 3 (in which price is close to marginal cost) network utilization is increased, and the network creates node history needed for the creation of additional services like priority and trusted rendering.
Multi-tier pricing operates using a multiplier attached to the OctaneBench algorithm. Tier 1 work has an OctaneBench multiplier of 1, thus 1 RNDR = 25 x OB4 * HR x a multiplier of 1. Tier 1 is the basis for RNDR compute pricing. Tier 2 has an OctaneBench Multiplier between 2x and 4x, providing users 2 to 4 times the amount of cumulative OctaneBench compute work per token. Similarly, Tier 3 has an OctaneBench multiplier between 8 and 16. The OctaneBench multiplier for different RNDR tiers may be fixed, or algorithmically adjusted based on real-time supply and demand conditions on the network - for example, a multiplier starting at the high end of the OB range when demand is low, and moving towards the lower bound when congestion increases. Below is a description of how much OctaneBench compute a user receives in each tier.
The Pricing of a RNDR Tok
[Figure 1 — MTP Table with General Parameters for Tiers]
Tier 1: Highest Security and Priority Tier
Tier 1 Render Services provide the equivalent of 256 OctaneBench x 256 seconds adjusted for OctaneBench4 and normalized to one US dollar. $1 worth of RNDR = 100xOB4/Hr. This pricing reflects the current costs of GPU cloud rendering on centralized services like Amazon Web Services (AWS)
As the network develops, and decentralized nodes are added to Tier 1, there will be additional audits and/or security procedures that Tier 1 nodes need to follow.
Tier 2: Highest Priority, Parallelized Peer-to-Peer Rendering
Tier 2 Render services provide 2x to 4X the total OctaneBench work as a Tier 1 RNDR Token on the network providing 200- 400% more compute power than Tier 1.
Tier 2 rendering work has Queue Priority over Tier 3, providing accelerated first in line parallelized rendering services on the network when compared to Tier 3 rendering
Tier 3: Low Priority
Tier 3 Services provide 8-16x the OctaneBench work as a Tier 1 RNDR Token. As a result, 1 RNDR Token will provide 800 - 1600% more compute power than Tier 1.
Tier 3 services have the lowest queue priority, so they are not recommended for time sensitive rendering jobs.
Nodes are tiered based on factors like Total System OB, Node History (Successful % of jobs processed), Minimums for VRAM and RAM which construct a Composite node Score. Users selecting a specific tier of work are matched to nodes from the selected tier on RNDR.
Figure 2 — MTP Table with Tiers and Select Rolling Factors
With Multi-Tier Pricing (MTP), the network is able to efficiently sort decentralized supply and demand, reducing deadweight loss when supply and demand don’t intersect at equilibrium[17]. The network pools GPU compute resources, enabling users with different preferences to meet on a peer-to-peer network. Multi-Tier Pricing organizes these preferences and resources according to price elasticity. As a result, both price sensitive and inelastic users derive utility from using the network, while node operators are rewarded commensurately for their resources and trustworthiness. Multi-Tier Pricing helps eliminate some trade-offs between speed, quality and resolution that artists have previously face when using current GPU cloud rendering platforms. RNDR enables artists to scale to the cloud with reduced cost or time pressures, providing more users with access to GPU cloud rendering tools needed to create boundary pushing cinematic work. Meanwhile, node operators have a platform to generate productivity from idle compute power.
Report abuse