IT168 Evaluation GPU Computing, also known as GPGPU, is a general-purpose computation on a graphics processing unit and refers to the use of GPUs to perform operations in engineering, digital media, and scientific research applications traditionally handled by a central processing unit (CPU). Since the parallel performance of the GPU is very powerful and the shipment volume is much higher than that of chips such as FPGAs, it has obvious cost advantages. In recent years, GPU computing has become synonymous with high-performance computers to some extent. Developers can take advantage of the GPU's computing power through a variety of tools, of which OpenCL is currently the most popular open standards for heterogeneous computing, ensuring that users' investments will not result in a significant amount of code rewriting due to hardware vendor changes. The Sapphire PGS AMD FirePro S series is designed for heterogeneous servers and small workstations and is capable of computing-intensive applications, helping researchers and engineers get results faster. With the AMD GCN Next-Generation Graphics Core (GCN) microarchitecture, the OpenCL optimized driver provided by the Sapphire AMD FirePro GPU can provide a cost-effective experience for a large number of ultra-computing applications. In the Sapphire PGS AMD FirePro S product line, there is a dual GPU product, the FirePro S10000, which has 3,584 stream processors, mainly for compute-intensive applications. The specifications are as follows: In the field of high-performance computing, there are several hard indicators for measuring performance: single- and double-precision floating-point throughput, memory bandwidth, and memory capacity. Floating-point performance determines the maximum computational performance that a GPU can achieve. Memory bandwidth determines access-intensive applications. Performance, and memory capacity determines the maximum data set size that can be handled with maximum efficiency. AMD's FirePro S10000 server graphics card is AMD's most powerful dual-GPU server graphics card ever. Impressively, its 1.48 TFLOPS double-precision peak floating-point performance is 7.8 times higher than its competitor's dual-CPU solution, and it can easily handle the most demanding compute-intensive data parallel tasks. With 6GB of GDDR5 memory, up to 480GB/s of memory bandwidth, and single-precision peak floating-point performance up to 5.91 TFLOPS, the Sapphire PGS AMD FirePro S10000 can easily handle a variety of compute-intensive workloads. Many high-performance applications involve the processing of large-scale data sets. For example, the National Institutes of the Alamos of the United States (LANL for short) has an open source program called SNAP, which is another program agent named PARTISN, and processes data. The set is quite large. But generally, these high-performance applications are far behind us. We can understand that there are not many, and the GPU computing tool HashCat is quite close to us. HashCat uses the GPU's powerful computing power to brute force password cracking. Support for more than 150 types of hash encodings is supported. As the software is continuously updated, more hash encoding types are supported. HashCat series software supports CPU, AMD GPU, and NVIDIA GPU for password cracking in hardware, and provides versions for Windows and Linux platforms. Drivers only need to install the graphics driver proposed in addition to the official version. The software is currently divided into three versions: HashCat, oclHashCat, and oclGaussCrack. And we used oclHashCat binaries for AMD in this test. The version number is: oclHashCat-2.01. With the support of Sapphire Technology, we obtained the sapphire PGS AMD FirePro S10000 6GB sample card from Sapphire PGS AMD FirePro with the exclusive dealer KF in China, and tested oclHashCat 2.01. For comparison, we added the sapphire PGS AMD. FirePro W8000 8GB comparison data: Background introduction: The so-called hash table is a hash table, which is a data structure directly accessing a memory storage location according to a key value (Key Value). For example, a map container in C++ is an application of a hash table. With the high-density configuration of dual GPUs, the results of the Sapphire PGS AMD FirePro S10000 are almost twice as fast as the comparable Sapphire PGS AMD FirePro W8000. Of course, the main task of the latter is that the workstation application is not designed for super-computing. Not the same, there are such test results are understandable. Currently, this server card for high-density computing is being provided by KF Kingfone as a solution. Users can contact the company's branch offices in Shenzhen, Guangzhou, Beijing, and Shanghai for further information. @ For more information please visit Sapphire PGS website: http://www.SapphirePGS.com/ Technical support email: ws. China's only designated sales service provider King Fung Electronics Website: Jingfeng Lynx flagship store: http://lianjingsm.tmall.com/ Sapphire PGS Jingdong Buy Website: Jingdong Sapphire PGS FirePro Product List Shenzhen KF Jingfeng: 0755 – 8375 9168 / 0755-83295070 (21st Floor, Block B, Jiahe Huaqiang Building, Shennan Middle Road, Futian District, Shenzhen) Guangzhou KF Jingfeng: 020 – 3849 8045 / 0755-83295070 (G306, B Floor, Pacific Digital Plaza, 1-7 Shipai West Road, Tianhe District, Guangzhou) Shanghai KF Jing Feng: 021 – 5490 1989 / 0755-83295070 (Room 1608, 16th Floor, Pacific Digital Plaza Phase 1, Huijia Building, 41 Caoxi North Road, Xuhui District, Shanghai) Beijing KF Jingfeng / 0755-83295070 (room A133, 6th Floor, Beijing Zhongguancun Electronics City) 12v trolling motor battery,solar powered golf cart,valve regulated lead acid battery,lithium marine batteries ,12v deep cycle marine battery,valve regulated lead acid vrla batteries EMoreShare International Trade (Suzhou) Co., Ltd , https://www.emoresharesystem.com
28 nm
GCN architecture (Graphics Core Next) bus interface PCI-Express 3.0 memory memory bit width 384 bit 2
Memory Memory: GDDR5
Memory bandwidth: 480 GB/s
Capacity: 6GB floating-point operations 5.91 TFLOPS single-precision floating-point operations
1.48 TFLOPS Double-precision Floating-Point Output Screen 5 Output Interface 1x Dual-Link DVI
4x Mini DisplayPort 1.2 Resolution 40962160 Pixel DisplayPort Resolution
25601600 Pixel Dual Link DVI Resolution Application Programming Interface OpenGL 4.4
OpenCL 1.2
DirectX? 12
Shader Model 5.0 VDI Virtual Application RemoteFX
VMware ESX
Citrix XenServer Product Features AMD PowerTune Kinetic Technology
AMD ZeroCore Power Technology Power Saving Technology
AMD Eyefinity Wide Area Technology 2.0
Data Correction Memory (ECC)
Streaming processing technology complete 30bit processing and transmission
8, 10, 16-bit per pixel color composition
Quad HD 4K Display Device Support
DirectGMA supports Partially Resident Textures (PRTs)
DOPP
Professional Software Certification Single Driver Architecture Thermal Solution Fan Cooling Size 2 Occupied Slots
FH/FL Power Requirements <375 W Certification Specification FCC, CE, C-Tick, BSMI, KCC, UL, VCCI, RoHS, WEEE System Requirements 750 W Recommended Host Power Supply
1 x available PCI-Express X16 footprint
2048MB system memory
2 x 8-pin video card power connection (PEG standard)
Microsoft? Windows? 8, 7, Vista (32/64-bit), Server 2008 R2 (64-bit) or Linux? (32/64-bit)
Product Accessories DVD-ROM (or Network) Driver Download and Installation CD Active Mini DisplayPort to Dual-Link DVI Cable (Dual Link)
Quick installation instructions