Wiredancer: Hardware Acceleration for High-Performance Blockchains

In the world of high-performance blockchains, milliseconds matter. For validators operating on networks like Solana, the ability to process transactions faster, more reliably, and with minimal latency isn't just a competitive advantage, it's a requirement for survival.

At Jump, we’ve spent years pushing the boundaries of what’s possible in trading systems and distributed infrastructure. Today, we’re applying those same principles to validator performance. One of the key technologies enabling this evolution is the FPGA: the Field Programmable Gate Array. When paired with @DoubleZero’s networking infrastructure, FPGAs help us tackle the growing performance and security demands of modern blockchains.

Here’s how, and why, we implement FPGAs into the infrastructure stack.

What is an FPGA?

An FPGA is essentially a two-dimensional grid of logic gates that can be reprogrammed after manufacturing. Unlike a CPU, where programmability comes at the cost of performance, or a custom ASIC, where performance comes at the cost of rigidity, an FPGA allows us to tailor the hardware for specific tasks and reprogram it as requirements evolve. This makes FPGAs ideal for environments where workloads change frequently or need to be iterated on quickly, like in blockchains.

CPUs are programmable as we can change the content of their instruction memory.  On the other hand, ASICs are efficient because the desired program is laid out as static gates.  FPGAs combine the best of both worlds by making gates out of small memory cells, allowing every gate to be programmable.

FPGA gates are made of Look-Up-Tables (LUTs), connected through a network of programmable switches.  LUTs are small memory cells whose content dictate the functionality of the gate.  Connections between LUTs are determined by the state of programmable switches.  Modern, high density FPGAs, include in the order of a few million LUTs.

FPGA development, just like ASIC, starts with synthesizing the desired program into a circuit, which is a network of gates. The resulting set of gates and connecting wires are then mapped to the resources available on the FPGA, i.e. programmable gates and programmable switches.  This step, often called place-and-route, is a very compute heavy, NP-complete searching problem, often taking hours to complete.  The resulting map is turned into a configuration bitstream that includes the memory content of every gate and the configuration of every routing switch used.

Why FPGAs

1. Parallel Execution

Parallel execution is a first class mechanism in FPGAs.  Every single gate on the chip is performing its desired function at the same time.  In fact, one of the key challenges of describing applications for FPGA execution is to contain the chip’s parallel nature, and perform tasks in sequence.

Many aspects of blockchain technology include parallel execution.  For example when cryptographically signing or verifying transactions, or executing independent transactions, there are numerous execution paths that can be taken at the same time.  Consequently FPGAs are a natural fit for implementing many components of blockchains.

2. Direct Network Connectivity

One of the key components of every blockchain is communication.  FPGAs are extremely efficient and versatile when it comes to data networks.  It is not uncommon to have FPGA boards with direct network cable connectivity to the FPGA chip.  This allows for tight integration with the data network at extremely high throughputs and low latencies.  This is one of the top reasons FPGAs are heavily used in telecommunication and high-frequency-trading applications.

The fact that FPGAs are excellent platforms for communication makes them a great fit for blockchain applications.  The versatile nature of the FPGA allows for every layer of the network OSI stack to be optimized or customized, allowing us to reach very specific metrics suitable for the chain.

3. Longevity and ROI

The FPGAs we showcased at Breakpoint 2024 were originally manufactured in 2017. That’s a seven-year-old piece of hardware still outperforming modern CPUs in specific workloads. This kind of staying power makes FPGAs uniquely valuable in a world where hardware cycles are typically measured in months.

4. Cost Efficiency Through Scale

You can’t build a USB charger for a dollar unless you’re building a million of them. The same logic applies to custom ASICs, they’re fast but only make sense at hyperscale. When it comes to FPGAs, however, because the same FPGA chip can be tailored to multiple applications, we can benefit from shared demand across multiple industries like aerospace and telecom.  This shared demand allows us to tap into the economies of scale without providing the much needed scale in a single application, for example blockchain.

5. Speed of Iteration

Even though FPGAs are hardware solutions, their upgrade cycle is close to software scale.  If the blockchain changes or a new attack vector emerges, we can synthesize a new hardware solution in a matter of days.  In addition, the process of pushing an FPGA bitstream to the device is in the order of 30-60 seconds. That’s a fraction of the time it would take to refactor and redeploy production-grade hardware.

6. The Right Compromise

In today’s world, solutions focus on three main factors: programmability, performance and power.  While FPGAs may not top the list in all categories, they do hit the sweet spot when all three categories are equally desired.  For example, when targeting blockchain, we don’t need nanosecond-scale networking similar to high-frequency-trading applications.  Similarly we don’t need a sub-watt power profile, nor do we need a Rust-programmable platform.  What we need is reasonable flexibility, efficiency, and scalability and FPGAs hit that trifecta better than any other platform today.

DoubleZero + Wiredancer: Deploying FPGAs in Production

FPGAs alone are powerful, but when combined with DoubleZero’s high-performance networking stack, they become a force multiplier for validator performance.

Through a system called Wiredancer, we deploy daisy-chained FPGAs at the edge of the validator, between the public internet and the validator’s internal stack. These cards are PCIe-based, passively cooled, and equipped with QSFP ports for direct fiber connectivity.

What do they do?

• First FPGA: Deduplicates incoming transactions.
• Second FPGA: Verifies ED25519 signatures in hardware at line rate.
• Output: Clean, verified, low-jitter traffic directly into the validator box.

The result? Validators spend less time filtering spam and more time packing blocks.

Real-World Results

In a test cluster, we pushed roughly 1 million TPS through a Solana mini-network. Under DDoS conditions, flooding the cluster with incorrectly signed transactions, throughput dropped by over 50%.

With Wiredancer deployed, the same attack barely registered. The FPGAs filtered duplicates and invalid signatures before they even touched the validator, restoring full throughput without any need for software modification.

Rethinking DDoS: Make It Not Your Problem

Blockchains are inherently transparent. Everyone knows your IP. That makes validators attractive targets for DDoS attacks.

You can try to mitigate these attacks in software, but that only scales so far. Our approach is different: make DDoS attacks irrelevant to us by out-scaling the attacker. The first step in a DDoS attack is to reach the provider.  That means any attack must traverse the vast network of ISPs leading to us, the service provider.  With FPGAs, we can make sure we can handle so much throughput that the attacker would run out of delivery bandwidth before our platform runs out of processing bandwidth.  In effect, the attacker needs to overwhelm the entire network of ISPs reaching us just to make a dent.

And if that happens? It’s no longer our problem, it’s theirs.

The Road Ahead for FPGAs

As Moore’s Law slows, we’re entering the age of specialization. Throwing more generic CPU cores at a problem won’t solve the bandwidth, latency, or efficiency challenges of tomorrow’s blockchains.  We believe FPGAs, and their evolution toward more programmable, specialized architectures, will continue to play a major role in the scaling of decentralized systems.

With solutions like Wiredancer, the barrier to entry is lower than it’s ever been.  Wiredancer’s daisy-chain architecture of deploying FPGAs streamlines future expansions should new tasks be found suitable for offloading.  For example, a key component of the Solana blockchain is Packing.  One could implement a new packing algorithm that could be partially offloaded to FPGAs, providing the validator with leverage for solving this intense task.

Conclusion

At Jump, we build for the frontier. And when it comes to blockchain infrastructure, we believe the frontier is hardware-accelerated, validator-optimized, and network-aware.

FPGAs help us get there. Not by replacing software, but by freeing it from the bottlenecks of the public internet.

With DoubleZero providing the foundation and Wiredancer augmenting the edge, we’re excited to be supporting the next era of performant, resilient blockchain infrastructure.

Watch Kaveh on FPGA’s on DoubleZero’s X page.