699ns
Min RTT (TCP/IP)
699ns
Min RTT (TCP/IP)
798ns
90th percentile RTT
8%
Latency gain
585ns
PCIe DMA RTT
Inconsistent execution times, even at the microsecond level, can mean missed fills, adverse slippage, and strategy degradation. Determinism matters as much as raw speed.
Competing firms are continuously investing in FPGA technology, co-location, and overclocked hardware. Standing still means falling behind.
Building a full ULL FPGA stack in-house requires deep expertise spanning FPGA design, networking protocols, PCIe DMA, and software integration, a multi-year investment with high execution risk
Whether it is market data ingestion, order book construction, pre-trade risk checks, or order routing, every component of the trading pipeline must operate at hardware speed.
The Orthogone and Blackcore partnership exists to solve exactly this challenge, delivering a fully integrated, benchmark-proven, production-ready ultra-low latency infrastructure stack.
It оffers соmplete оfflоad оf TCP/IP and UDP/IP prоtоcоls аt thе hardwarе level. As a rеsult, thе nеtwоrk dаta path bypasses the оperating systеm stack еntirеly, which rеmоvеs CPU sсheduling dеlays, reduces kеrnеl оverhеad, and еliminates latеnсy cаused by sоftwаre.
Specifiсally designеd fоr eleсtrоniс trаding, the ICON 3100-RL+ featurеs an Intеl Cоre i9 prосessоr that can reасh spеeds оf up tо 5.9 GHz, while the G3 SPR-M utilizes аn Intеl Xеоn prоcessоr with speeds оf up tо 4.8 GHz. This setup еnsurеs thаt еvery nanоsесоnd sаvеd at the FPGA lаyеr translаtеs effeсtivеly tо thе аppliсаtiоn levеl.
It оperatеs аrоund 8% fastеr than similar standard servers. This is nоt merely a thеоrеtiсal claim; every соmpоnent has bееn rigоrоusly bеnchmаrked tоgethеr, and the full mеthоdоlоgy alоng with rеsults has bеen published.
Don’t take our word for it. The Orthogone × Blackcore ULL FPGA Framework Benchmark Report documents every test in full, methodology, hardware configurations, software stack versions, and complete latency curves.
Together, Orthogone's ULL FPGA IP and Framework and Blackcore's overclocked high-performance servers form a validated, optimized, and commercially available solution for modern electronic trading.
Verified in independent benchmarks
The Orthogone ULL FPGA Framework, running on Blackcore's ICON 3100-RL+ overclocked server, achieves a minimum RTT latency of 699ns for a 64B frame — with a 90th percentile of just 798ns. Approximately 8% faster than a comparable standard-class server.
Across all queues
Consistent latency across all 16 PCIe DMA queues tested. The Blackcore ICON 3100-RL+ achieves a PCIe DMA RTT as low as 585ns for small payloads, a 50ns gain over standard servers, holding consistently up to 512B payload sizes.
Built for the demands of HFT
Blackcore's ICON 3100-RL+ (Intel Core i9-14900KS, up to 8 cores at 5.9 GHz) and G3 SPR-M (Intel Xeon w7-2495X, up to 24 cores at 4.8 GHz) are purpose-engineered for electronic trading. Deployed in over 28 data centers worldwide.
Keeps latency-critical work in the FPGA
The FPGA handles all latency-critical operations, including TCP/IP and UDP/IP protocol offload at the hardware level, while host software manages non-time-sensitive tasks. The networking data path never touches the OS stack.
No guesswork required
Tested and validated end-to-end. The published benchmark report provides full data, TCP/IP loopback, UDP/IP loopback, and PCIe DMA latency, across three server configurations, with minimum and 90th-percentile results at every payload size.
With next-generation platforms
Current results achieved on AMD-Xilinx x3522pv FPGA. ULL IP Cores are also compatible with AMD-Xilinx UL3524 and UL3422, which can reduce latency by an additional ~25ns. The roadmap to PCIe Gen 5 offers further improvements.
Two technology leaders, each best-in-class in their domain, delivering a combined solution that neither could achieve alone.
Whether you are evaluating FPGA acceleration for the first time, looking to upgrade your existing infrastructure, or seeking an engineering partner, Orthogone is ready to engage.
