The now available second generation FPGA driven Amazon EC2 instance (F2) | Amazon Web Services

Update 2/11/2025: Revised part of Genomics.

Equipped up to eight AMD Field-Pracble Gate Troys (FPGAS), AMD EPYC (Milan) processors with up to 192 cores, high bandwidth (HBM), up to 8 TIB of SSD based on instances and up to 2 TIB memory, new F2 two sizes and are ready to speed up your genomics, processing multimedia, large data, satellite communications, network, network Simulation of silicon and live video video.

Fast FPGA recapitulation
Here’s the way I explained the FPGA model when we look at the first generation of the Amazon Elastic Compute Cloud (Amazon EC2) FPGA instances

One of the most interesting routes to your own hardware solution is known as a field of programmable gate or FPGA. Unlike the purpose of a chip, which is designed with one function and then hard to implement it, FPGA is more flexible. It can be programmed in the field after it has been connected to the socket on the PC board. Each FPGA contained a fixed, final number of simple logic gates. FPGA programming is “simply” a matter of their connection and creation of the required logical functions (or, xor, etc.) or storage elements (flip-flops and shift regiders). Unlike the CPU, which is Essentielly Serial (with several parallel elements) and has fixed six and data paths (usually 32 or 64 bits), FPGA can be programmed to perform Mary in parallel, and the operations can almost be?

From this launch, AWS customers have used instances F1 to host Mary of various types of applications and services. With newer FPGA, more power and larger memory bandwidth, the new F2 instances are an even better host for highly parallelized and computing instant workload.

Each AMD Virtex UltraScale+ HBM VU47P FPGA has 2.85 million system logic cells and 9.024 DSP cuts (up to 28 DSP calculation peaks when processing INT8 values). The FPGA accelerator card associated with each instance F2 provides 16 GIB with high bandwidth and 64 GIB DDR4 memory on FPGA.

Inside F2
F2 are driven by processors 3. Generation AMD EPYC (Milan). Compared to F1 instances, they offer up to 3 times as many processes, up to twice as much system memory and NVMe storage and up to 4x network width. Each FPGA comes with 16 GIB high bandwidth (HBM) with up to 460 GIB/s bandwidth. Here are the sizes of instances and specifications:

High-end F2.48XLAGE The instance supports the AWS Cloud (CDI) digital interface for reliable transport of uncompressed live video between applications, with latency of an instance on an instance of up to 8 milliseconds.

Building FPGA applications
The AWS EC2 FPGA development kit contains tools that you will use for development, simulation, tuning, compilation and starting your hardware accelerated FPGA applications. For development and simulation, you can run a FPGA developer for FPGA developers for FPGA developers on memory -optimized or computing devices, and then use an instance of F2 for final tuning and testing.

The tools contained in the developer set support a number of developmental paradigms, tools, accelerator languages ​​and tuning. Regardless of your selection, you will eventually create an image of Amazon FPGA (AFI), which contains your own logic of acceleration and shell AWS, which implements access to FPGA memory, PCIe bus, interruption and external periphery. You can put AFIS in as many F2 instances as required, share with other AWS accounts or publish on AWS marketplace.

If you have already created an application that is running in F1 instances, you will need to update your development around the latest AM tools, then rebuild and verify before upgrading to F2 instances.

FPGA instance in action
Here are some great examples of how F1 and F2 can support unique and highly applicant workloads:

Genomic – The multinational pharmaceutical and biotechnology company Astrazeneca has used thousands of F1 instances to build the fastest genomic pipeline in the world that was able to process more than 400,000 samples of the entire genome in less than two months.

Communication satellite -The assaytiters range from inflexible and expensive physical infrastructure (modulators, demodulators, combinations, distribution, etc.) with agile FPGA -defined software. Using the Digital Signal (DSP) elements on FPGA, these solutions can be reconfigured in the field to support new progress and meet changing requirements. Key F2 features such as up to 8 FPGA instances, generous love bandwidth, and user virtual Ethernet support for data plan development (DPDK), support for processing multiple complicated progress in parallel.

Analytics -Integrates SQL Processing Unit (SPU) SQL with Presto, Apache Spark and other open source engine interviewing engine, adds faster processing of queries and efficiency in queries on the market at F2 instances.

What to know
Here are the last few things you should know about F2 cases:

Regions – Examples of F2 are available today on AWS regions in the US (N. Virginia) and Europe (London) with plans to extend availability to other regions over time.

Operating systems -F2 are instances only for Linux.

Purchase options -F2 are Avaidaable on request, on site, savings plan, reserved instance and reserved host form.

– Jeff;;

Updated 5 February 2025 -We have ficked F2.6xLange, the new smaller size of the EC2 F2 instances for applications that are more cost or more efficient when running on individual FPGA.