Case Studies

Hybrid Cluster Implementation Grants Users Full Control, Saves Costs

Fixstars Solutions desired to run Kubernetes jobs with both CPU and GPU. In order to do so, they used WireGuard and Kubernetes to create a hybrid cluster. This allowed them to utilize cloud service providers and their own own-premise machines to run Kubernetes jobs, and also granted Fixstars more control over their hardware with high security, while saving money without relying on AWS EKS rental equipment. With a Kubernetes Hybrid Cluster, Fixstars has full control over cost, flexibility, security, data governance and hardware for their clusters.

Read Story

Fixstars Creates AI, Large Language Model-Powered SlackBot on Knowledge Base

Discover the benefits of on-premises AI with Fixstars Solutions' purpose-built chatbot. Eliminate the need for cloud-based services while maintaining a high level of data security and control. Our bot uses large language models and FAISS semantic search to provide efficient and accurate responses, all securely hosted on your own infrastructure. It's a smart, secure, and customizable solution for your organization's communication needs.

Read Story

NoRMCorre Acceleration by GPU for Neuroscientists, 2x Faster than Standard CPU

NoRMCorre is a motion correction algorithm commonly used by neuroscientists. However, it is one of the most time-consuming pipelines used for neuroanalysis. We ported and optimized NoRMCorre to GPU, and our accelerated GPU version runs 2x faster than the original CPU version of NoRMCorre, using a rigid mode data set.

In-Situ Sequencing Acceleration by GPU and Flash Storage for Neuroscientists, Shortened a 6 Month Analysis into 5 Days

In-Situ Sequencing is a new sequencing method which, unlike conventional DNA/RNA methods, is able to simultaneously profile sequences of bases and their locations in a cell. Before our In-Situ Sequencing acceleration, our client's software took about 6 months to analyze RNA images captured by microscopes. Our experienced engineers improved the software source code, improving the performance by 35x, which shortened the usual 6 month analysis into 5 days.

Deconvolution Acceleration by GPU for Biologists, New Software 405x Faster than Deconvolution Lab2

Deconvolution is the process of sharpening image quality of microscopic objects, and it is commonly used by biological scientists. Our GPU acceleration completes the deconvolution process 405x faster than the commonly used Deconvolution Lab2, and is 4.4x faster than Huygens GPU.

Geometric Performance Primitives (GPP) for EDA is 25x Faster than Reference CPUs

The performance gain of computational geometry is one of the most challenging topics. Fixstars developed a GPU/CPU multi-core accelerated computational geometry engine, called the Geometric Performance Primitives (GPP). This enabled faster turnaround times of high-precision designs, tests and verifications by efficiently utilizing higher computational resources available. GPP for EDA can run up to 25x faster than reference CPUs with reduced memory footprint. Our client was able to retain perfect accuracy with 25x faster software acceleration.

Learn more on NVIDIA DEVELOPER

Mizuho Securities Utilizing Xeon Phi Accelerators to Accelerate their Derivative Valuation System, Simulation is 30x Faster than Before

Mizuho Securities, an investment bank in Japan, required faster simulation of derivative price running for their verification tests. They struggled to meet their clients' needs since it took too much time using their derivative valuation system. With the help from Fixstars, they were able to take full advantage of Intel Xeon Phi's parallelism and obtained 30x faster simulation.

HPC for US Air Force Research Lab, Reaches 500 TFLOPS

The US Air Force (USAF) requires supercomputers to operate with the most advanced technology to protect our nation from any domestic or international threats. Fixstars configured and deployed the PS3 Cluster "Condor" for the USAF Research Lab. An estimated 500 TFLOPS of performance were developed for their applications such as radar enhancement, satellite image processing, and AI.

Expansion Microscopy Powered through Accelerated GPU Deconvolution, 405x Faster than Deconvolution Lab2

Expansion Microscopy is a method that neuroscientists use to capture images of neurons. Fixstars Solutions developed the Expansion Microscopy (ExM) Studio, designed to accelerate the deconvolution process, packaged with TeraStitcher for Expansion Microscopy. The ExM Studio contains 2 components: GPU Accelerated Deconvolution and GPU TeraStitcher. The Fixstars GPU Accelerated Deconvolution is 405x faster than the standard Deconvolution Lab2, and 4.4x faster than Huygens. Fixstars Solutions ExM Studio is used nationwide, within various universities and neuroscience labs.

Learn more on exm.studio