Skip to content

FPGA Low Latency Trading Evaluation Options

By myrtle.ai Newsroom on March 31, 2026

myrtle.ai has recently made it possible for anyone to evaluate the performance VOLLO could deliver for FPGA low latency trading, without purchasing hardware, downloading tools, or requiring specialist FPGA knowledge.

VOLLO is a latency-leading machine learning inference accelerator designed to support ultra-low latency machine learning workloads, including modern AI trading tools used in real-time financial systems. To make evaluation simple, myrtle.ai provides multiple ways to test expected performance before deployment.

The VOLLO Sandbox is an online environment that compiles your Python code and simulates how it would run on VOLLO hardware. It allows developers to explore FPGA acceleration for trading strategies and understand the achievable latency before committing to hardware.

This article explains the four available evaluation options and how they support the full journey from early exploration to live deployment.

 

The four evaluation options

  1. 1. VOLLO Sandbox
  2. 2. Downloadable Virtual Machine (VM)
  3. 3. Evaluation licence
  4. 4. Remote evaluation on myrtle.ai servers

 

Key capabilities comparison

Immediate results No download required No charge No FPGA expertise required No FPGA hardware required Evaluation on your own server
VOLLO Sandbox ✔ ✔ ✔ ✔ ✔
Downloadable VM ✔ ✔ ✔ ✔ ✔
Evaluation Licence ✔ ✔ ✔
Remote Evaluation ✔ ✔ ✔

 

Importantly, none of these options require prior FPGA experience. This makes VOLLO accessible to ML engineers and quantitative developers who want to explore FPGA low latency benefits without maintaining in-house FPGA tooling.

Following evaluation, ML engineers and quantitative developers with no FPGA tools or expertise can go on to deploy and even run VOLLO on supported hardware. That said, if FPGA tools and expertise are available, VOLLO is also available as an FPGA netlist for integration into a custom FPGA design.

 

VOLLO Sandbox

The simplest way to begin evaluating FPGA low latency trading performance is the VOLLO Sandbox.

The Sandbox is an online tool that delivers immediate results:

  • No downloads
  • No installation
  • No hardware purchase

It runs a bit-accurate, cycle-level simulation, ensuring latency results closely match real hardware behavior.

 

Using the Sandbox, you can:

  • Measure expected inference latency for your own trading models
  • Test example models if your own are not yet available
  • Evaluate performance on different accelerator cards
  • Explore the trade-off between model complexity and execution latency

This makes the Sandbox ideal for early experimentation with low latency AI inference, particularly when evaluating models intended for real-time market decision systems.

 

Downloadable Virtual Machine (VM)

There are two main reasons to use the downloadable Virtual Machine rather than the online Sandbox.

 

  1. Model security

After initial testing, you may want to continue development without allowing proprietary trading models to leave your internal network.

In this case, you can download the VOLLO SDK, which includes the VM. No licence is required, and it uses the same simulator as the Sandbox, producing identical performance results.

 

  1. Full input/output simulation

The VM enables complete simulation with data input and output, allowing you to back-test trading models under realistic data flows.

This is particularly valuable for teams validating AI trading tools that must operate under strict latency constraints.

 

Evaluation licence

Once simulation results meet your performance requirements, the next step is running VOLLO on real hardware.

 

By requesting a free-of-charge evaluation licence, you can deploy VOLLO directly on your chosen acceleration platform and integrate it into your production network.

 

This option allows you to:

  • Run models on live or replayed market data
  • Measure true end-to-end latency
  • Validate throughput under operational load
  • Confirm performance targets for FPGA acceleration for trading

This stage provides confidence that simulation results translate directly to real-world environments.

 

Remote evaluation

If you want to evaluate VOLLO on hardware but are not ready to acquire acceleration cards yourself, myrtle.ai can provide remote access to an evaluation server.

 

With this option:

  • A server and FPGA accelerator are allocated in a myrtle.ai facility
  • You receive remote access for testing
  • No capital expenditure is required

While exact server and card configurations cannot always be guaranteed, this option allows teams to confirm real hardware performance for ultra-low latency machine learning workloads before investing internally.

 

Summary

VOLLO now offers a complete set of evaluation options supporting the full development lifecycle for FPGA low latency trading:

  • From initial curiosity
  • To detailed simulation
  • To live hardware trials

Whether you are exploring FPGA low latency, validating low latency AI inference, or preparing production deployment of advanced AI trading tools, these options allow you to measure performance with confidence at every stage.

 

To begin your evaluation journey, we encourage you to start with the VOLLO Sandbox today.

 

We look forward to working with you soon.