Toll free - (866) 338-1768



Wolverine Coprocessors


The Family

Convey’s Wolverine family of coprocessors are compact, general purpose application accelerators that provide a “drop-in” performance boost, dramatically increasing the performance of x86 systems.  

  • Based on Field Programmable Gate Array technology

  • PCI Express form factor cards can be installed in a variety of off-the-shelf servers

  • Designed for enterprise applications, with a robust design and system tools delivering ease of integration and application development productivity

Click here to view card specifications

On-board Bandwidth
WX690NM Xilinx Virtex-7 XC7VX690T 0 GB 40 GB/s 75W
WX690 Xilinx Virtex-7 XC7VX690T 16/32/64 GB 40 GB/s 75W
WX690SP Xilinx Virtex-7 XC7VX690T 16/32/64 GB 40 GB/s 150W
WX2000 Xilinx Virtex-7 XC7V2000T 16/32/64 GB 40 GB/s 150W

Reduced Total Cost of Ownership

Convey Application Accelerators dramatically increase the performance of an x86 system, allowing a single accelerated system to do the work of 20-40 conventional servers.  This allows a handful of servers to replace many racks of equipment, reducing both acquisition and facilities costs. A Convey Accelerated Image Resizer (based on the company’s Wolverine 690 card), can resize images at rates up to 70 times faster than a conventional server.  Systems with an Accelerated Image Resizer can deliver:

  • 91% lower capital costs

  • 97% reduced power costs

  • 95% reduced facilities costs

The reduced capital expense combined with savings in operational costs delivers over a 90% reduction in 3 year cost of ownership compared to conventional servers alone.

Globally Shared Vitual Memory

Wolverine coprocessors share a virtual address space with threads running on the host system:GloballySharedMemory.jpg

  • Wolverine on-board memory is mapped into the virtual address space of the attached process

  • Coprocessor infrastructure handles virtual-to-physical translation

  • Host and coprocessor have full access to all data, using the same pointers

GSVM makes programming the coprocessor easier while protecting the host system against errant references.