Synaptron FAQ

GPU

  1. If I have multiple GPUs, does VRAM add up, or can I only use one GPU?

Software does not currently support a multi-GPU setup with parallel functionality. Each GPU is tied to one NFT, so only one GPU can be used per node / per OS.

  1. If I have multiple GPUs in one rig, what happens? Can I run a node with several GPUs or multiple nodes on one GPU?

Synaptron software can currently only recognise a single GPU and the NFT to GPU relationship is 1 to 1. To take advantage of multiple GPUs on a single machine you would need to virtualise using a hypervisor (with GPU passthrough) AND have an NFT per GPU.

NOTE: This is for advanced users only and beyond the scope of Timpi to support

  1. Which GPUs are supported? Nvidia only?

Currently, only Nvidia GPUs are supported with a minimum of 6.1 Compute Capability. Refer to the CUDA compatibility list for details: https://developer.nvidia.com/cuda-gpus.

  1. Does Lane Speed Matter in GenAI?

Yes, lane speed matters! In simple terms, the "lanes" are like highways that connect your GPU to the rest of the system. More lanes (e.g., 16 instead of 8) mean more data can flow at once, which can improve performance for GenAI models, especially when handling large datasets or complex tasks.

  1. What PCIe throughput will my Synaptron GPU need?

We recommend using a minimum of PCIe Gen3 x8 or above, for sufficient throughput.

  1. Can you share VRAM (GPU Splitting) across multiple VMs if the GPU is big enough?

No, we do not support this configuration natively.

  1. How are nodes categorized into tiers?

The system reports your GPU's VRAM and UUID (GPU unique identifier) to the TAP (Timpi Autonomous Protocol). Which then calculates which Tier your node belongs to.

  • Tier 1: 4-6GB VRAM

  • Tier 2: 8-16GB VRAM

  1. Do nodes with 8GB VRAM get paid the same as those with 16GB?

There are 2 tiers which are decided by VRAM. Initially each node within a tier will receive equal rewards based on an uptime threshold. Over the coming months this will transition into a performance based model where the amount of jobs done will be a factor - with this in mind not only VRAM, but also tensor cores, cuda cores and clock speeds all play a factor. Please refer to the Reward Distribution Section for more details on that.

OS Compatibility

  1. Can this be containerized in Linux?

Yes, containerization is supported. Ubuntu 22.04

  1. Can I run it on Windows 10 or 11?

Yes, both are supported.

  1. Can Synaptron integrate with HiveOS? Many miners use HiveOS.

We currently support Ubuntu 22.04 official release, Windows 10, and 11. HiveOS is outside the officially supported operating systems.

  1. Will it work in a virtual machine?

Yes, VM instructions coming later.

  1. Does this run on Windows Subsystem for Linux (WSL)? Are ports required?

No, it doesn’t run on WSL, and no ports are needed.

  1. Is there a beta version available?

We’re targeting a beta release in late January/early February 2025 - Please watch for announcements.

  1. What are the compute resources required to run a Synaptron node?

  • 4 Core CPU

  • 12GB RAM

  • 250GB SSD/NVMe

  • Nvidia GPU with a minimum of 6.1 Compute Capability. Refer to the CUDA compatibility list for details: https://developer.nvidia.com/cuda-gpus.

Networking and Security

  1. Do I need to set up port forwarding or UPnP?

No, Synaptron nodes don’t require any open ports or forwarding.

  1. Is a static IP required?

No static IP is needed.

  1. Is there a specific internet speed/bandwidth required?

No minimum internet required but it's recommended to be on unlimited internet since nodes will be downloading datasets constantly.

  1. Do I need to worry about cybersecurity? What data is processed?

No need to worry. The system is isolated, handling only public domain data. It operates securely with no open ports.

Workload and Availability

  1. How intensive is the workload?

Expect workloads to fully utilize the GPU at 100%. Ensure adequate power and cooling are available to maintain system stability and performance.

  1. What if the node disconnects during work? Will there be penalties?

This is still to be determined, but there may be penalties for persistent disconnections.

  1. Will there always be a queue of work to do?

Yes. The timpi index is being continually expanded and refreshed by the Collector nodes, providing a continuous stream of text and images to be processed by the Synaptron nodes.

  1. What work will Synaptron nodes perform?

Initially, tasks include image recognition to populate our index. Both tiers will handle these tasks. Over time, more tasks will be added, with high-VRAM tasks exclusive to Tier 2 nodes. Tasks are automatically assigned based on your node’s capabilities.

Reward Distribution

  1. How are total reward pools calculated?

Rewards for both tiers follow a pool approach similar to Collectors or Guardians:

  • For every active Tier 1 node, $1400 NTMPI is added to the monthly reward pool.

  • For every active Tier 2 node, $1600 NTMPI is added to the monthly reward pool.

Inactive node rewards are redistributed proportionally, maintaining Tier 2 nodes earning 14.3% more than Tier 1 nodes.

  1. How are individual rewards determined?

  • Initially, rewards within a tier are equally distributed among active and available nodes.

  • Within six months, a performance-based system will be introduced where nodes completing more work earn higher rewards.

During the initial phase, eligibility is based on node availability. A minimum availability percentage must be met to qualify for monthly rewards. This gives us time to benchmark realistic metrics that can be used for a future performance-based rewards model that is fair.

Example of Reward Distribution If there are 20 Node Access NFTs, and 15 nodes are active (10 Tier 1, 5 Tier 2):

  • Tier 1 pool: $14,000 + $4,667 (inactive contribution) = $18,667 NTMPI

  • Tier 2 pool: $8,000 + $2,667 (inactive contribution) = $10,667 NTMPI

Rewards per active node:

  • Tier 1: $18,667 / 10 = $1,867 NTMPI

  • Tier 2: $10,667 / 5 = $2,133 NTMPI

Once the performance-based phase begins, the total pool remains the same, but rewards are distributed based on individual node performance.

  1. How does it work in practice?

For example:

  • January 2025: 15 active nodes (10 Tier 1, 5 Tier 2)

  • Reward pools:

    • Tier 1: $18,667 NTMPI

    • Tier 2: $10,667 NTMPI

If still in the availability-based phase, all active nodes earn equally. In the performance-based phase, pools remain the same, but rewards are allocated based on individual contributions.

Last updated