Risky Business: Latest Benchmarks Show How Financial Industry Can Harness NVIDIA DGX Platform to Better Manage Market Uncertainty

The current NVIDIA DGX systems, with 640GB of GPU memory each, landed eight efficiency records on the financial markets extensively seen STAC-A2 standard of monetary threat designs, including taking leading honors for energy and space efficiency.

Amid increasing market volatility, monetary danger supervisors are trying to find quicker, much better market analytics. Today thats provided by innovative danger algorithms running on the fastest parallel computing systems.

This was likewise the very first time any STAC-A2 solution utilizing containers has been investigated. Red Hat OpenShift is the markets leading enterprise Kubernetes platform. Kubernetes, an open source container orchestration platform, has actually become the de facto requirement for managing containerized workloads, which are instrumental in deploying intricate multi-stage workflows.

Enhancing the state of the art for risk platforms, NVIDIA DGX A100 systems running Red Hat software application can provide monetary services firms performance and operational gains. These systems utilized a fraction of the energy and space of rival servers in recent standard tests.

The STAC-A2 results emphasize the versatility of DGX A100 systems for integrating with the most recent deployment designs and Red Hat as a provider of a Kubernetes environment that can satisfy some of the most demanding business performance requirements.

A few of the biggest companies on Wall Street and the broader global financial industry rely on the STAC-A2 as an essential danger design benchmark to measure compute platform efficiency.

Nearly 15x More Throughput

Huge banks, hedge funds and danger managers across banks stand to gain from not only information throughput advances however likewise enhanced operational effectiveness.

The current NVIDIA DGX A100 systems with 640GB of GPU memory provide 14.8 x higher throughput (variety of choices priced per second) 1 than a recently checked service based on a single, standard CPU server– a dual-socket CPU-based system2– as measured by the STAC-A2 benchmark. It likewise outperforms formerly checked systems based on 10 CPU-only cloud nodes3 and 8x dual-socket CPU-based servers4

The record results from NVIDIA have actually been examined by the Securities Technology Analysis Center (STAC). Members of the STAC Benchmark Council include over 450 of the worlds leading banks, hedge funds and monetary services technology companies, which contribute to the standards makeup. The STAC Report is available here.


Performance for Reduced Operating Expenses.

The most current DGX A100 system with 640GB of GPU memory5 uses lower operating expenditures:.

Minimized energy and square video expenditures for systems in information centers can make a huge difference in operating costs. Thats specifically essential as IT organizations make the case for financial investments to cover brand-new systems.

2.6 x the energy efficiency6.

2x the space performance of a CPU-based cluster system tested in 2018 and 4x the space effectiveness of a just recently checked CPU-based system7.

DGX systems have actually always been about performance at scale, so its not a surprise that it set brand-new records for the optimum number of assets and courses that can be simulated12. The most recent DGX systems can handle more than 2x the properties of the very best non-NVIDIA system tested13 (despite the reality that the work increases quadratically with the number of assets), and 20x more paths than a recently evaluated CPU-based server14.

NVIDIA DGX was developed to address the worlds most tough calculate problems effectively and at scale. These latest STAC-A2 results show its strength in delivering on that vision.

The STAC-A2 market threat standard mimics variations in rate of interest and other security rate elements with time, evaluating their effect on options prices. One important step is the simulation of underlying security rates paths, which is illustrated in figure 1. The STAC standard involves computing the advancement of thousands, if not numerous thousands, of these security price paths grouped by security.

DGX Devours STAC-A2.

Utilizing a Monte Carlo simulation (arbitrarily tasting a likelihood distribution similar to in figure 1) with the Longstaff-Schwartz technique (a backwards model algorithm, which steps back in time from a maturity date) solves for option-price steps with time.

These simulated results are used in level of sensitivity calculations that make up danger scores understood in the financing industry as “the Greeks.”.

The technique enables financial services firms to determine the danger of present holdings in the future and of possible trades.

Understanding STAC-A2: Meet the Greeks.

NVIDIA set new time-to-solution records for the large issue size as well– 2.3 x the speed in the warm large Greeks benchmark compared with the fastest non-NVIDIA system (an eight-node CPU-based cluster) 11.

STAC-A2 mimics these option-price sensitivities (the Greeks) for multiple possessions by using a monetary analytics technique called Longstaff-Schwartz Monte Carlo.

DGX A100 with 640GB of GPU memory8 didnt just move the bar a little bit. Compared with the very best previous numbers from non-NVIDIA sped up systems, it opened a large space: 3x the throughput9, and 2.6 x the speed in the warm standard Greeks benchmark10.

Find out more about NVIDIA DGX systems


β2.HPORTFOLIO.SPEED2 SUT ID 2103153 SUT ID INTC2103314 SUT ID INTC1810125 SUT ID NVDA2109146 STAC-A2. β2.HPORTFOLIO.SPACE _ EFF vs. SkyLake cluster SUT ID INTC181012, Ice Lake SUT ID INTC2103158 SUT ID NVDA2109149 STAC-A2. β2.GREEKS.MAX _ ASSETS vs. the 8 way Skylake cluster SUT ID INTC18101214 STAC-A2.

β2.HPORTFOLIO.SPEED2 SUT ID 2103153 SUT ID INTC2103314 SUT ID INTC1810125 SUT ID NVDA2109146 STAC-A2. β2.HPORTFOLIO.ENERGY _ EFF vs. the Ice Lake based server SUT ID INTC2103157 STAC-A2. β2.HPORTFOLIO.SPACE _ EFF vs. SkyLake cluster SUT ID INTC181012, Ice Lake SUT ID INTC2103158 SUT ID NVDA2109149 STAC-A2. β2.HPORTFOLIO.SPEED vs. the 10 node cloud cluster SUT ID INTC21033110 STAC-A2. β2.GREEKS.MAX _ ASSETS vs. the 8 method Skylake cluster SUT ID INTC18101214 STAC-A2.

Leave a Reply

Your email address will not be published. Required fields are marked *