Find a Driver

Treiber Download
Driver Autodetect- or -

AMD FireStream™ GPU Compute Accelerators

FireStream™ 9270

AMD FireStream™ 9270 GPU Compute Accelerator

Ultimate GPU Compute Performance:

	Over 1 TFLOPS raw single precision performance 240 GFLOPS raw double precision performance

Technology Need	AMD Stream Computing Solution
Double Precision Floating Point	The FireStream 9270 supports double precision floating point operations in hardware
High Performance per Watt	Over 5 GFLOPS per watt of single precision performance potential
Open Systems	Familiar 32 and 64 bit Linux and Microsoft® Windows® environments High level tools from multiple 3rd party developers
Multiple Software Entry Points	C-like development environment via extensions to open source Brook+ Low level access to hardware via open IL and ISA specifications in AMD Stream's CAL
Scalable Solutions	Stream software supports multiple GPUs per system
Commodity Components	FireStream leverages commodity GPU components designed for PC gamers resulting in a reliable GPU and memory roadmap

AMD FireStream™ 9270 Specifications
Number of GPUs	1
Stream Cores	800
Peak compute rate	1.2 TFLOPS (single), 240 GFLOPS (double)
Floating point formats	IEEE single & double precision
GPU local memory	2GB GDDR5 SDRAM
Memory interface	256-bit @ 850 MHz
Peak memory bandwidth	108.8 GB/s
System interface	PCIe x16 Gen 2
Auxiliary power connectors	Two 6-pin
Power consumption	160 watts typical, <220 watts peak
Board dimensions	4.376” x 9.5” (111.15 x 241.3 mm), dual-slot
Thermal solution	Active fan-sink

Optimized for computation
The AMD FireStream product line provides the industry's first double-precision floating point capability on a GPU. The AMD FireStream 9270 is our next generation DP-FP product. With 2GB GDDR5 memory on board and single-precision performance of 1.2 TFLOPS and double precision performance of 240 GFLOPS, the FireStream 9270 is ideal for the most demanding compute-intensive, data-parallel tasks.

Using ground-breaking 55nm process technology, even this large-memory board consumes less than 220 watts peak (160 watts typical), making it easy to deploy in a variety of server and workstation packages.

ATI Stream Software
Our ATI Stream software stack offers a complete top-down programming model ranging from the easily accessible Brook+, our implementation of the popular open source C-like Brook GPU programming language, to lower level performance tuning through our CAL (Compute Abstraction Layer) and IL (Intermediate Language) interfaces.

For more information on our ATI Stream software, please visit us at www.amd.com/stream.

Advantages
AMD FireStream GPU compute accelerators deliver all of the parallel processing power of the GPU to compute-intensive applications for scientists, engineers and consumers. Delivering 1.2 TFLOPS, developers can dramatically increase the performance of their data parallel applications.

System Requirements

PCIe 2.0 based server or workstation with available x16 lane graphics slot
- Minimum PCIe 1.0 x16
600W or greater power supply
512MB of system memory

AMD Warranty and Support

Enterprise-level support
Three year limited product repair/replacement warranty
Toll-free phone and email access to technical support team

FireStream™ 9250

AMD FireStream™ 9250 GPU Compute Accelerator

	Breaking the 1 TFLOPS barrier at under 120 watts!¹ Our 2nd generation stream computing GPU

Technology Need	ATI Stream Computing Solution
Double Precision Floating Point	The FireStream 9250 supports double precision floating point operations in hardware
High Performance per Watt	Up to 8 GFLOPS per watt of single precision performance potential
Open Systems	Familiar 32 and 64 bit Linux and Microsoft® Windows® environments High level tools from multiple 3^rd party developers
Multiple Software Entry Points	C-like development environment via extensions to open source Brook+ Low level access to hardware via open IL and ISA specifications in ATI Stream’s CAL
Scalable Solutions	Stream software supports multiple GPUs per system
Commodity Components	FireStream leverages commodity GPU components designed for PC gamers resulting in a reliable GPU and memory roadmap

Some more highlights:

1GB on-board GDDR3 memory
Memory export
Double precision floating point support
120W power consumption; 55nm process
Asynchronous DMA
PCIe 2.0 x16 interface
BIOS settings optimized for stream processing
API and OS Support
Open system approach with ATI Stream SDK

Optimized for computation
The AMD FireStream product line provides the industry's first double-precision floating point capability on a GPU. The AMD FireStream 9250 is our second generation DP-FP product. With 1GB GDDR3 memory on board and single-precision performance of 1 TFLOPS, it is ideal for the most demanding compute-intensive, data-parallel tasks.
Using ground-breaking 55nm process technology, even this large-memory board consumes less than 120 watts peak, making it easy to deploy in a variety of server and workstation packages.

Advantages
AMD FireStream GPU compute accelerators deliver all of the parallel processing power of the GPU to compute-intensive applications for scientists, engineers and consumers. Delivering 1 TFLOPS, developers can dramatically increase the performance of their data parallel task applications.

System Requirements

PCIe 2.0 based server or workstation with available x16 lane graphics slot
Minimum PCIe 1.0 x16
500W or greater power supply
512MB of system memory

AMD Warranty and Support

Enterprise-level support
Three year limited product repair/replacement warranty
Toll-free phone and email access to technical support team

Download the AMD FireStream™ 9250 Datasheet

FireStream™ 9170

AMD FireStream™ 9170 GPU Compute Accelerator

	Our latest generation GPU compute accelerator
	Our latest generation GPU compute accelerator

Technology Need	ATI Stream Computing Solution
Double Precision Floating Point	The FireStream 9170 supports double precision floating point operations in hardware
High Performance per Watt	Over 5 GFLOPS per watt of single precision performance potential Over 1 GFLOP/W in double precision
Open Systems	Familiar 32 and 64 bit Linux and Windows environments High level tools from multiple 3^rd party developers
Multiple Software Entry Points	C-like development environment via extensions to open source Brook+ Low level access to hardware via open IL and ISA specifications in ATI Stream’s CAL
Scalable Solutions	Stream software supports multiple GPUs per system
Commodity Components	FireStream leverages commodity GPU components designed for PC gamers resulting in a reliable GPU and memory roadmap

Some more highlights:

320 stream cores (compute units or ALUs)
2GB on-board GDDR3 memory
Memory export
Double precision floating point support
<100W power consumption; 55nm process
Asynchronous DMA
PCIe 2.0 x16 interface
BIOS settings optimized for stream processing
API and OS Support
Open system approach with ATI Stream SDK

Advantages
AMD FireStream stream processors deliver all of the parallel processing power of the GPU to compute-intensive applications for scientists, engineers and consumers. Using up to 500 GFLOPS developers can dramatically increase the performance of their data parallel task applications.

System Requirements

PCIe 2.0 based server or workstation with available x16 lane graphics slot
Minimum PCIe 1.0 x16
500W or greater power supply
512MB of system memory

AMD Warranty and Support

Enterprise-level support
Three year limited product repair/replacement warranty
Toll-free phone and email access to technical support team

Download the AMD FireStream™ 9170 Datasheet

FireStream™ 9350/9370

AMD FireStream™ GPUs offer revolutionary floating-point performance across a wide range of computing applications. Utilizing the multi-billion-transistor ASICs developed for Radeon consumer graphics boards, FireStream GPUs are designed and built by AMD to meet the demanding performance and reliability requirements of High Performance Computing (HPC) clusters that scale to thousands of nodes. Designed for use in scalable servers, blades, and PCIe chassis, FireStream GPUs are available in systems from leading server OEMs. Priced competitively, FireStream GPUs offer unparalleled value for HPC applications.

The FireStream 9370 offers maximum GPU performance with 4GB of DDR5 memory in a 2-slot configuration. The FireStream 9350 offers maximum performance / slot with 2GB DDR5 memory in a 1-slot configuration. Both boards are passively cooled and include a single Display Port output. (For workstation installations that require active cooling of the GPU, AMD’s ATI FirePro™ professional graphics boards are software compatible and offer comparable features.) AMD FireStream GPUs support the open-standard OpenCL programming environment, enabling development of high performance applications that are portable across GPU platforms.

Technology Need	AMD FireStream™ Computing Solution
High DPFP performance	528 GFLOPS double precision
High performance per Watt	2.4 GFLOPS / Watt
Open standards	OpenCL, Direct Compute
Performance optimization tools	OpenCL SDK
Professional-quality components and manufacturing	All FireStream boards are made by AMD
Scalable server configurations	Passive heatsink simplifies thermal design
Extended product life	3-year availability; 3-year limited warranty

Features and Benefits

Feature	Benefit
40 nm fabrication process, 2.1 billion devices	Unprecedented DPFP on a single chip
Up to 20 SIMD execution units	2.64 Single Precision TFLOPS
Industry standard OpenCL software	Performance-optimized, portable application software
Made by AMD	Assured quality and reliability
PCIe 2.1 Host Interface	8 GB/S Host-GPU bandwidth
150W TDP, 1-slot FirePro 9350	Maximum DPFP performance per cm3
AMD Value	Superior technology at lower cost

Specifications

Specifications	AMD FireStream™ 9350	AMD FireStream™ 9370
Number of GPUs	1	1
Memory Capacity	2GB DDR5	4 GB DDR5
Double Precision Floating Point	400 GFLOPS	528 GFLOPS
Single Precision Floating Point	2.0 TFLOPS	2.64 TFLOPS
TDP	150W	225W
Stream Cores	1440	1600
SIMD Processors	18	20
Core Clock Frequency	700 MHz	825 MHz
System Interface	PCIe 2.1	PCIe 2.1
Memory Bandwidth	128 GB / S	147 GB / S
Memory Clock Frequency	1.0 GHz	1.2 GHz
Dimensions	4.376” x 9.5”; Single slot	4.376” x 10.5”; Dual slot
Aux. Power Connector	6-pin (2x3)	8-pin (2x4)
Thermal solution	Passive heat-sink	Passive heat-sink
Display output	1 Display Port	1 Display Port

AMD FireStream Software (Visit www.amd.com/developer for more information.)

Math Libraries	Description
AMD Core Math Library for Graphics Processors (ACML-GPU)	AMD GPU-accelerated version of ACML
	1D, 2D, 3D-FFT
AMD Accelerated Parallel Processing Math Libraries (APPML)	BLAS Level 3 functions
GPU Automatically Tuned Linear Algebra Software (GATLAS)	GATLAS attempts to find OpenCL™ matrix multiply kernels with good performance through an auto-tuning benchmark methodology.
ViennaCL	The Vienna Computing Library (ViennaCL) is a scientific computing library written in C++ and based on OpenCL™. It includes common linear algebra operations (BLAS 1 and 2) and solutions of large systems of equations by iterative methods.
Video Codecs
MainConcept OpenCL™ H.264/AVC Encoder SDK	OpenCL™-accelerated MPEG-2, VC-1 and H.264/AVCvideo encoding.
Language Bindings	Allow you to write the OpenCL host code in your choice of programming languages. GPU kernels are written in OpenCL
C++	CAL++ (for ATI CAL)
Fortran	HMPP Workbench from CAPS enterprise
JAVA	JavaCL
Matlab	IPT_ATI_PROJECT
.NET	OpenCL .Net OpenTK
PYTHON	CLyther PyGWA (for ATI CAL) PyOpenCL Pythoncl Compiling Python to a hybrid execution environment(ACM)

For more information on other AMD based HPC solutions, please visit http://sites.amd.com/uk/business/it-solutions/compute-intensive-hpc/Pages/compute-intensive-hpc.aspx.