SANTA CLARA, Calif., Could 07, 2024 — As we speak, AMD (NASDAQ: AMD) introduced that Optiver – a market maker buying and selling on 100+ exchanges – is utilizing AMD chips for its new information heart infrastructure. Utilizing AMD EPYC processors, AMD Solarflare Ethernet adapters, AMD Virtex FPGAs and AMD Alveo adaptable accelerators, Optiver is addressing technical challenges in capital markets to construct programs for monetary markets.
“We’re a crew of researchers and engineers pushed by a standard imaginative and prescient: constructing programs that make monetary markets higher,” stated Alexander Itkin, U.S. chief expertise officer, Optiver. “We’re continually pushing the bounds, whether or not it’s difficult the pace of sunshine to attenuate latency or constructing world-leading computational scale. We want computing infrastructure that allows us to navigate the complexities of the capital markets in nanoseconds. For us, that is solely doable with high-performance merchandise from AMD, which assist us efficiently and constantly adapt to monetary markets.”
“AMD is uniquely positioned to fulfill the wants of business prospects with a strong and various portfolio of compute engines that allow spectacular efficiency and vitality effectivity; two vital parts of the fashionable information heart,” stated Archana Vemulapalli, company vice chairman, Business Enterprise, AMD. “Our work with Optiver is a superb instance of how we work along with prospects to push the bounds of what’s doable, whereas demonstrating the capabilities of an AMD business answer and the way it can tackle each present-day and future challenges for our prospects.”
The precise AMD enterprise merchandise that allow Optiver to modernize its infrastructure and unlock pace and effectivity for its enterprise embody:
- AMD EPYC CPUs: Optiver is utilizing 4th Gen AMD EPYC processors to construct out large-scale compute wanted to drive information evaluation and ship core density for the corporate’s information heart modernization and workload consolidation efforts.
- Solarflare Ethernet Adapters: Optiver additionally makes use of the AMD Solarflare X2 collection Ethernet community adapters designed for high-performance digital buying and selling environments. These adapters present low-level management of community interfaces on the utility degree, permitting Optiver to higher design programs to scale and keep aggressive.
- Virtex UltraScale+ FPGAs: The exponential development of information calls for sensible community and information heart options that attain past fixed-function silicon units to ship most throughput, excessive information processing functionality, and the flexibleness to adapt to evolving connectivity requirements. The AMD Virtex UltraScale+ VU23P FPGA helps Optiver speed up its programs whereas remaining endlessly adaptable.
- Alveo Accelerators: Optiver makes use of the AMD Alveo U55C and Alveo UL3524 adaptable accelerator playing cards, the latter of which delivers a 7X latency enchancment over prior technology FPGA expertise1.
“Our final purpose was to get quicker entry to information and knowledge that’s vital to fulfill the distinctive wants and workflows that the monetary business calls for,” stated Kevin Sprague, U.S. {hardware} lead, Optiver. “The high-performance Alveo ultra-low latency playing cards are vital to serving to us transfer extra data, quicker. Our collaboration with AMD helped us develop the merchandise and options our business calls for. Moreover, our joint work has allowed us to develop and optimize our providing by leveraging not solely AMD FPGAs, but additionally AMD EPYC CPUs to construct out high-performance buying and selling programs and infrastructure.”
AMD helps enterprise prospects tackle the challenges of in the present day with a strong portfolio of revolutionary and confirmed options. Optiver illustrates the power of the management AMD compute portfolio, the place particular person units in tandem kind clever options higher than the sum of their components, delivering unparalleled worth to corporations and prospects.
1 Testing performed by AMD Efficiency Labs as of 8/16/23 on the Alveo UL3524 accelerator card, utilizing Vivado Design Suite 2023.1 and operating on Vivado Lab ({Hardware} Supervisor) 2023.1. Primarily based on the GTF Latency Benchmark Design configured to allow GTF transceivers in inside near-end loopback mode. GTF TX and RX clocks function at similar frequency of ~644MHz with a 180 levels section shift. GTF Latency Benchmark Design measures latency in {hardware} by latching worth of a single free operating counter. Latency is measured because the distinction between when TX information is latched on the GTF transceiver and when TX information is latched on the GTF receiver previous to routing again into the FPGA cloth. Latency measurement doesn’t embody protocol overhead, protocol framing, programmable logic (PL) latency, TX PL interface setup time, RX PL interface clock-to-out, bundle flight time, and different sources of latency. Benchmark check was run 1,000 instances with 250 frames per check. Cited measurement result’s primarily based on GTF transceiver “RAW Mode”, the place PCS (bodily medium attachment) of the transceiver passes information ‘as-is’ to FPGA cloth. Latency measurement is constant throughout all check runs for this configuration. System producers could differ configurations, yielding completely different outcomes. ALV-10
