< High-Performance Computing >

Can machines think?

We think, the answer to this question is solely based on the ability to calculate information. A machine which has about the compute power of the human brain can think – if the right algorithms are implemented.

QMware develops both, the powerful hardware and novel quantum algorithms which bring machines to life!

Machine Learning System for AI applications

The QUANTON-HGX2 Machine Learning System for Artificial Intelligence applications includes more than 864 billion transistors, equivalent to about 1 / 100ths of all synapses in the human brain. With its industry-leading computing power density, the system has a peak performance of approximately 80 peta FLOPS – that is, it can calculate 80,000,000,000,000,000 floating point operations (additions or multiplications) per second.

QMware offers this most powerful matrix processing system representing the world’s leading price/performance ratio of about 160 giga FLOPS per €.

QMware sets standards in High-Performance-Computing (HPC) with highest integrated circuits at 7nm scale with AMD & Nvidia.

#1 Best connection possibilities for add-on cards market-wide

#2 The most flexible server solution meets best I/O performance

#3 Amazing NVME capacity

#4 Highest security solution industry-wide

Modular Computing Architecture

All our components have the latest innovative integrated circuits on board and at the same time QMware standardized interfaces which provide compatibility for many years. With this outstanding rack architecture, called Disaggregated Rack Blade (DRB) System QMware’s HPCs morph with adapted workloads.


EPYC processors strike the perfect balance of cores/threads, memory, I/O bandwidth and security features to deliver excellent performance for many High Performance Computing (HPC) workloads.

With its industry-leading core count, EPYC processors can enable more VMs and more robustly configured VMs per server than previously possible in 1 or 2-socket solutions.

EPYC processors natively support up to 32 NVMe or SATA devices in both 1 socket and 2 socket designs, enabling streamlined Software Defined and Direct Attached Storage solutions.

EPYC processors provide a solid foundation for Machine Learning applications such as proactive maintenance, fraud detection, network security monitoring, real-time trading and self-driving vehicles.

EPYC processors can help accelerate leading Big Data and In-Memory Database applications, that require large memory sizes, high memory bandwidth and high-bandwidth I/O, to manage and analyze massive amounts of unstructured data.

Industry-leading core count coupled with the highest memory capacity and bandwidth enables optimal virtual desktop density and performance. Additionally AMD graphics processors support SR-IOV for superior performance.

Memory is key to leverage High Performance Computing. Thus, QMware sets standards regarding memory efficiency and I/O performance. Our Storage Solutions reach from IOPS in TERABYTES to scale-out in EXABYTES.

Storage Virtualization
The PlatinStor® SC is based on an integrated storage hypervisor which excellently combines the enormous benefits of storage virtualization with the requirements of storing the data in the same system. The resulting flexibility and security allows for easy management and separates the storage hardware from the virtual volumes which can be seen by the servers. An integration of existing storage systems into virtualization layers enables centralized management and usage of the advantages of the modern PlatinStor® systems.

Cascade Cache Technology
Multilevel Cache Technology: CISC Cache, RISC Cache and SSD Cache provide unsurpassed performance.

Advanced Comprehensive Tiering
Up to fifteen tiering classes ensure optimum matching of the requirements for the performance of PlatinStor® SC.

Unified Storage
A platform for all storage-related applications, presentation of volumes over iSCSI and FC* and full NAS functionality via CIFS, SMB and NFS with native AD integration.

Native Active/Active
All grid nodes located in the storage can be active and written to simultaneously for one and the same virtual volume and establish maximum availability, even in the event of failure of one side for SAN and NAS services.

Multi-Tier Investment Protection
Existing hardware, even from different manufacturers can continue to be used. Current PlatinStor® SC platforms can also be used in PlatinStor® SC next generation systems. A modular design allows a simple expansion at any given time.


* FC models

Integrated Capacity Optimization
By Native Thin Provisioning, integrated data flow analysis and optimization, almost 100% of physical memory is available for user data. Data Deduplication is also being supported by default.

Enterprise Security and Reliability
Unique features such as: Continuous Data Protection, Advanced Snapshots, Enterprise Storage Virtualization, WSAN, Asynchronous Mirroring, Advanced Site Recovery, MPIO and special service levels make PlatinStor® SC one of the safest and most reliable storage systems in the world.

NVMe Tiering
PlatinStor® SC systems incorporate industries fastest non-volatile tiering which is perfectly fit for databases, high performance clusters and other power hungry services and applications.

True Synchronous Grid
Each block that is confirmed by PlatinStor® SC systems is guaranteed to exist on both nodes, data loss due to failure of one side is thus avoided optimally.

Convergent Connectivity
All popular media and protocols are supported. iSCSI over 1/10/25/40/50/100/200GBit, copper and optical (RJ-45 / SFP +/ SFP28 / QSFP+ / QSFP28 / QSFP56), FC 8GBit, 16GBit and 32GBit*.

Easy Maintenance
The PlatinStor® SC Management Console provides centralized control of the entire SAN grid, also email-notification as well as IPMI enable continuous monitoring, analysis and reporting tools yield a more efficient operation of PlatinStor® SC. In addition to the graphical user interface, there is a command line available.

QMware’s hybrid computing approach includes the network. DPU’s (Data Processing Units) are built into our compute nodes to offload data streams from the other processing units like CPU’s, GPU’s and QPU’s (Quantum Processing Units) as well.

QMware develop software for those DPUs to build in-network storage systems. In this manner, QMware is able to build data centers as one single hybrid quantum high performance computer (HQ-HPC).

The core of QMware is its processor agnostic architecture, which leads to a completely new programming paradigm, called Hybrid Quantum Algorithm (HQA). Once an engineer develops a software for a QMware machine, it will be able to run infinitely with higher and higher speed on new hardware platforms as they become available. The hybrid processor architecture supports any future upgrade.

Today we have CPUs, GPUs and vQPUs (Quantum Simulators) implemented, and this list will continue to grow steadily.

Data center operators need more and more energy to cool the running systems. The conventional methods of cooling by air flows are already reaching their limits. New cooling technologies are needed to cool the systems efficiently. A fast, reliable and sealed method to reduce energy costs and OPEX while reusing heat is Direct Contact Liquid Cooling.

Direct contact liquid cooling is especially valuable for data centers to increase rack density, reduce overall cooling costs and cool the next generation of high-performance processors.

To provide customers with the best solutions, a „building block“ approach of three modules is used, including passive Cold Plate Loops, Rack Manifolds and Coolant Distribution Units. This gives customers the enormous advantage of product flexibility in integrating liquid cooling into any computer environment or server configuration

Cold plate loops
Passive cold plate loops are mounted directly into the server to capture heat and include passive cold plate(s), tubing, and quick disconnects which connect to the rack manifold. Cold plates are available for any combination of CPU, GPU, and memory in any server, with adaptations for VR, ASIC, FPGA, and other components.

Rack manifold
Rack Manifold manages liquid distribution between the Coolant Distribution Unit and any number of Passive Coldplate Loops. Rack Manifolds are made of reliable stainless steel, rendering them incredibly robust. Combined with dry-break, dripless Quick Disconnects, these manifolds are safe and effective building blocks when paired with Passive Coldplate Loops.
Rack Manifolds are organized for a manual connection at the front or back of the rack. Manifolds are flexible and can be arranged vertically or horizontally within a rack.

Coolant distribution unit
Depending on load requirements and availability of facility water, a variety of coolant distribution units (CDUs) are offered, including CHx (Liquid-to-Liquid), AHx (Liquid-to-Air) and customer-specific options.