FPGA Main Board Accelerators

The Storage Networking Industry Association (SNIA) defines computation storage as “Architectures that provide Computational Storage Services coupled to storage offloading host processing and/or reducing data movement”, with two foundational constructs, Computational Storage Devices (CSx) and Computational Storage Services (CSS).

The IA-420F is built on BittWare ’s enterprise-class foundation stones: existing products and processes that are already proven through extensive validation. When it’s time to ramp to production, this means a faster turnaround and utilizing the same hardware you have been using for development. 

The IA-840F is built on BittWare ’s enterprise-class foundation stones: existing products and processes that are already proven through extensive validation. When it’s time to ramp to production, this means a faster turnaround and utilizing the same hardware you have been using for development.

The Bittware IA-860m is built on BittWare’s enterprise-class foundation stones: existing products and processes that are already proven through extensive validation. When it’s time to ramp to production, this means a faster turnaround and utilizing the same hardware you have been using for development.

The Bittware IA-780i is a compact 400G FPGA Accelerator Card. The Intel Agilex 7 I-Series FPGAs are optimized for applications that are bandwidth intensive and require high performance processor interfaces. 

The Bittware IA-440i and IA-640i are compact 400G FPGA Accelerator Cards. The Intel Agilex 7 I-Series FPGAs are optimized for applications that are bandwidth intensive and require high performance processor interfaces. 

The Bittware IA-865m is built on BittWare’s enterprise-class foundation stones: existing products and processes that are already proven through extensive validation. When it’s time to ramp to production, this means a faster turnaround and utilizing the same hardware you have been using for development.

The Bittware PCIe 5.0 NIC Card has an Intel Agilex® 7 FPGA and LMS ÜberNIC with CXL. This fully hardware-accelerated network stack finds first use in data centers for the financial services industry (FSI) and others. This is the first FPGA-powered CXL NIC.

The BittWare 520C is a PCIe accelerating OpenCL-programmable FPGA board built using the Intel Stratix 10 with F1760 NF43 package and a PCIe Gen3 x16 host interface. With fast, independent optical I/O, establish dense, directly coupled FPGA clusters.

The BittWare 520N is a PCIe accelerating OpenCL-programmable board built using the Intel Stratix 10 F1760 NF43 FPGA with four 100G QSFP Ethernet ports and a PCIe Gen3 x16 host interface. With fast, independent optical I/O, establish dense, directly coupled FPGA clusters.

The BittWare 520N-MX accelerator board built with an Intel Stratix 10 MX FPGA is well suited to 100G line rate network packet processing and compute-intensive data center applications that demand high memory bandwidths.

The Bittware 520R-MX FPGA PCIe Accelerator Board is optimized for sensor processing applications with massive real-time data ingest requirements. The FPGA is a Stratix 10 MX2100 with integrated HBM2 memory and 48 optical links for 480 Gbps Optical Input high-speed input to the FPGA, and OCuLink connectors allow for expansion.

The Bittware 520NX PCIe FPGA Accelerator Card has 8GB HBM2 and a Stratix 10 NX2100 FPGA which is AI-optimized for high-bandwidth, low-latency artificial intelligence. This revolutionary accelerator delivers a unique combination of capabilities needed to implement low latency and larger AI models.

The Gidel Proc10S high performance scalable compute acceleration board is powered by an Intel Stratix 10 FPGA with up to 200 Kle, 260 GB DDR4 RAM, and an option for a SoC Quad-core 64-bit ARM Cortex-A53 MPCore processor. Peak single precision performance reaches 10 TFLOPS.

Gidel’s Proc10M / Proc10N module is designed to enable easy use and immediate accessibility to the powerful Intel Stratix 10 FPGA with HBM2 on-chip memory technology for both host and embedded systems.

The BittWare 385-SoC is an ARM-based System-on-Chip FPGA accelerator board built on the Intel Arria 10 SX F34. It features dual QSFP ports, two banks of 4G DDR4 SDRAM external memory, and 8-Lane PCI-Express Gen 3 in a NIC form factor.

The BittWare 510T Compute Acceleration Card is an FPGA co-processor that delivers maximum performance-per-watt for intensive compute applications such as the datacenter. In a GPU format, and with a 16-lane PCIe Gen 3 card, it is powered by dual Intel Arria 10 floating point FPGAs. 

BittWare’s A10PL4 is a low-profile PCIe x8 card based on the Intel Arria 10 GX FPGA. The Arria 10 features high density and a power-efficient FPGA fabric. Included are high-speed transceivers, hard floating-point DSP blocks, and embedded PCIe x8 Gen3.

The BittWare A10SA4 PCIe accelerator board built with a Intel Arria 10 GX 1150 FPGA is well suited to large FPGA loads in the data center. The A10SA4 supports the Open Computing Language (OpenCL™) programming model, which enables a high-level software-like development flow, which greatly simplifies FPGA development. 

The Proc10A™ system is a flexible, high performance, low-power FPGA platform based on Intel’s powerful Arria 10 FPGA. The Proc10A offers high I/O throughput and high powered on-board processing and data management for low latency for applications such as HPC high performance computing, storage, networking, and high-end imaging.

The HawkEye is a PCIe 8 lane Gen 3 low profile board based on Intel’s Arria 10 FPGA. This compute accelerator has up to 18 GB DDR4 on-board memory, up to four SFP+ links for a maximum of 56 Gbs. The Arria 10 FPGA provides up to 480K LEs and IEEE floating-point numbers.

Gidel offers a number of ultra-compact high-performance FPGA modules that can be mounted on a PCIe carrier board. The modules provide a complete FPGA envelope ready for use while leaving flexibility to tailoring the peripheral system precisely. Gidel provides templates and powerful development tools that enable quick development and optimal FPGA utilization. 

QD-F910 is a low-profile PCIe card based on the Intel Arria 10 GX FPGA. The Arria 10 boasts high densities and a power-efficient FPGA fabric married with a rich feature set including high-speed transceivers, hard floating point DSP blocks, and embedded Gen3 PCIe. 

The BittWare 385A is a low profile 1.5 Teraflops FPGA accelerator board built on the Intel Arria 10 1150 GX. It features dual QSFP ports, 8GB DDR4 SDRAM external memory, and 8-Lane PCI-Express Gen 3.

BittWare’s 385A-SFP is an Intel Arria 10 based FPGA network accelerator card for 8 lane, single width, PCI Express Gen 3, with six SFP+ ports, rated at 10 Gbps.

With flexibility to serve applications like 5G, RADAR, test and measurement and satellite communications, BittWare’s RFX-8440 and RFX-8441 cards feature the latest third-generation Xilinx RFSoC UltraScale+ Zynq XCZU43 in an E1156 package, a powerful single-chip adaptable radio platform providing up to 6 GHz of direct RF sampling.

BittWare 250-SoC system-on-a-chip exploits the NVMe-over-Fabric (NVMEoF) technology of the Xilinx Zynq UltraScale+ MPSoC FPGA. It also has 64-bit ARM processors.

The BittWare 250-U2 is a proxy in-line accelerator (PIA) in the U2 form factor. FPGAs enable accelerated computing, network, and database access.

The BittWare 250S+ is powered by a Xilinx KU15P Ultrascale FPGA (FFVA1156 in default configuration speed grade 2). It is a fully programmable, flash SSD, near-storage, localized FPGA accelerator with up to 4 M.2 NVMe SSDs or M.2 to U.2 SSD Cables and DDR4 or MRAM.

The BittWare XUPP3R PCIe accelerator board built with a Xilinx UltraScale+  FPGA is designed for high-performance, high-bandwidth, and reduced latency applications demanding massive data flow and packet processing. 

The BittWare XUPPL4 PCIe accelerator board built with a Xilinx UltraScale+  FPGA is designed for high-performance, high-bandwidth, and reduced latency applications demanding massive data flow and packet processing.

The BittWare XUPVV4 PCIe accelerator board built with a Xilinx UltraScale+  FPGA is designed for high-density datacenter applications. The UltraScale+ VU13P offers up to 3.8 million logic elements.

The BittWare XUPVV8 PCIe accelerator board built with a Xilinx UltraScale+ VU13P FPGA is well suited to high density applications. The UltraScale+ VU13P offers up to 3.8 million logic elements. This power density makes thermal management difficult, but Bittware solves this problem  with the Viper platform.

The BittWare XUPVVH PCIe accelerator board built with a Xilinx UltraScale+ VU35P/37P FPGA is well suited to high density data center applications that demand high memory bandwidths.

The BittWare XUPVVP PCIe accelerator board built with a Xilinx UltraScale+ FPGA is designed for ultra high power applications. The UltraScale+ VU13P offers up to 3.8 million logic elements.

The PCIe8 CML-ECL is a PCI Express 8-lane interface that enables fast DMA and synchronous I/O to transfer CML and ECL data between an external device and a host computer. 8 bidirectional channels are supported; four CML and four ECL.

The EDT PCI CDa board provides one high-speed 16-bit parallel channel or an optional 16 synchronous serial channels between an external device and a PCI local bus computer (or optional 16-channel synchronous serial interface).

The EDT ECL / LVDS-E / RS-422-E is a mezzanine board that pairs with an EDT main board for PCI Express for high-speed data transfer. This “E-series” board options are ECL, LVDS or RS-422 inputs and outputs in groups of four, with an E1/T1 option.

The EDT PCIe4 CDa is a PCI Express 4-lane interface that enables fast DMA and synchronous I/O to transfer differential data between an external device and a host computer, LVDS or RS422. FPGA is Intel Arria II GX (EP2AGX45D) combining PCIe DMA at 700 Mbs with UI functionality at 150 Mbs I/O, which is via integrated FIFO buffers.

The EDT PCIe8g3 S5-40G supports PCI Express – PCIe, Gen3 - x8 and has one 40G QSFP+ and supports 1/10/40 GbE, OC3/12/48/192 (STM1/4/16/64), or OTU1/2/2e/2f. The ports link to the SERDES user-interface for serialization and deserialization, and clock recovery and its own 10 – 210 MHz programmable reference clock.

The EDT PCIe8g3 S5-10G supports PCI Express – PCIe, Gen3 - x8 and has four 10G SFP+ and 1 – 10 GbE or OC3/12/48/192 (STM1/4/16/64), or OTU1/2/2e/2f. The ports link to the SERDES user-interface for serialization and deserialization, and clock recovery and its own 10 – 210 MHz programmable reference clock.

The EDT PCIe8g3 A5-10G supports PCI Express – PCIe, Gen3 - x8 and has two 10G SFP+ and 1 – 10 GbE or OC3/12/48/192 (STM1/4/16/64), or OTU1/2/2e/2f. This low-profile board is available as a half-height or full-height backpanel. The port links to the SERDES user-interface for serialization and deserialization, and clock recovery and its own 1 – 808 MHz programmable reference clock.

EDT Bridge, Backpanel, and Time Boards attache to EDT main boards.

Not just more gates, but better data highways, more programmable elements, and 8x GDDR6 memories delivering 4Tbps of aggregate bandwidth for the most demanding compute-intensive applications up to 41K INT8 MACs and 134 INT4 TOPS with multiple floating-point and integer numerical formats.

The Gidel JPEG Compression FPGA IP Core encodes JPEG at high-performance. The compression IP is fast processing, low latency, and compact silicon utilization. The IP can compress high-performance camera image streams on a small FPGA device, or alternatively, multiple instances can run on a single larger FPGA device.

Gidel’s lossless compression FPGA IP Core is real-time compression for Color Filter Arrays such as Bayer, Monochrome, and RGB images and videos. The IP enables compression of multiple cameras or sensors at pixel clock rates exceeding 1 Giga pixels.

The Bittware AX-840p embodies AMD Xilinx Versal ACAPs FPGAs, which integrate networked, power-optimized cores on an adaptable platform for the most challenging compute and networking applications. 7nm Architecture: Versal Premium architecture combines high bandwidth and compute density on an adaptable platform.

Bittware IA-440p features AMD Xilinx Versal ACAPs that integrate networked, power-optimized cores on an adaptable platform for the most challenging compute and networking applications. The 7nm Versal Premium Architecture combines high bandwidth and compute density on an adaptable platform.