Processor specialist Advanced Micro Devices on Nov. 12 launched the FirePro S10000 graphics card, which the company said exceeds 1 teraflops, or 1 trillion floating-point operations per second, of double-precision floating-point performance.
The processor, aimed at high-performance computing (HPC) environments, also brings 5.91 TFLOPS of peak single-precision and 1.48 TFLOPS of double-precision floating-point calculations, compared with the Nvidia Tesla K10, which is capable of up to 4.58 TFLOPS of peak single-precision and 190 GFLOPS peak double-precision floating-point performance.
Equipped with a 6GB GDDR5 frame buffer and a 384-bit interface, the FirePro S10000 delivers up to 1.5 times the memory bandwidth of the comparable competing dual-GPU solution, compared to the Nvidia Tesla K10. In addition, DirectGMA support removes CPU bandwidth and latency bottlenecks to improve communication between both graphics processing units (GPUs). This also enables P2P data transfers between devices on the bus and the GPU, bypassing the need to traverse the host's main memory, use the CPU or incur additional redundant transfers over PCI Express. The result is high throughput, low-latency transfers that allow for improved compute times of complex calculations that require high accuracy.
"The demands placed on servers by compute and graphics-intensive workloads continue to grow exponentially as professionals work with larger data sets to design and engineer new products and services," said David Cummings, AMD senior director and general manager of professional graphics. "The AMD FirePro S10000, equipped with our Graphics Core Next Architecture, enables server graphics to play a dual role in providing both compute and graphics horsepower simultaneously. This is executed without compromising performance for users while helping reduce the total cost of ownership for IT managers."
The FirePro S10000, which is equipped with AMD next-generation Graphics Core Next (GCN) architecture, is also suited for virtual desktop infrastructure (VDI) and workstation graphics deployments. The GCN architecture is the world's first 28nm GPU architecture, which enables AMD to fit up to 4.3 billion transistors--the most basic building blocks of a GPU-into approximately the same space once needed to fit 2.6 billion. The architecture also includes AMD ZeroCore Power technology, which ensures that unused or idle GPUs are as efficient as possible.
With two GPUs in one dual-slot card, the FirePro S10000 enables high GPU density in the data center for VDI and helps increase overall processing performance, makes it easier to sustain compute and facilitate graphics-intensive workloads. The processor also includes support for OpenCL, a compute programming language widely used among developers looking to take full advantage of the combined parallel processing capabilities.