Data Vortex Test-bed systems proved, in hardware, the scalability of the DV network

Our scalable network allows any number of nodes to be plugged into the switch, and our validation systems delivered huge performance improvements. Dv-enabled systems allow better performance for applications that need better data communication. For equal performance on a specific hard problem, our initial systems require less space and represent a 90% savings in power consumption. A Data Vortex-enabled system of equal size has a 32x to 100x performance improvement over a comparable system with the same number of x86 cores.

The purpose of the DV validation systems was to prove the large scale implementation capability of the Data Vortex topology. We succeeded at this task and the below data show our advantages compared to alternate networks and fabrics.

In 2023, we showcased that Data Vortex-connected accelerators can achieve solve graph searches in seconds, a stark contrast to the days it takes for a baseline competitor to accomplish the same (3500x performance!).

Linear Graph Performance on DV-connected Accelerators (2023)

Multi-Level DV Switch Performance Comparison (2019)

(Note: Performance on benchmarks is unaffected by multi-level switching, proving scalability)

Graph 500 Breadth First Search (2018)

Source of Graph 500 performance data of other systems

 GFFT System Performance Comparison (2016)

Source of GFFT performance data of other systems

 Giga Updates Per Second (GUPS) (2016)

Data Vortex White Paper: “HPCC Random Access Benchmark Excels on Data Vortex

Single port ping-pong BW each direction (packets/sec per port) (2016)

Source of IB performance data

Infiniband vs. Data Vortex Barrier Sync Times (2015)

3D FFT Performance Comparison (2015)

Source of Jaguar (“Titan”) data