The Sapphire PGS AMD FirePro W9000 is, on paper, the fastest workstation GPU on the market. How does it fare during our testing?
[Translated from VR-Zone Chinese]
When the GCN architecture enters the professional world, what kind of performance can be expected?
In August 2012, AMD released the new FirePro series professional graphics card HD 7000 series GCN architecture, which was finally officially introduced into the professional realm. This FirePro-generation GPU of the Evergreen-series is of the upgraded VILW-5 architecture, so this professional-grade display card takes performance and energy efficiency to new heights.
The FirePro W9000 W product-line has been tested to currently show the strongest performance. As many users would understand, the 28nm Tahiti XT GPU clock is set to 975MHz, up to 4.0/1.0 TFLOPS in single / double precision floating point operations, TDP 274 W; 384bit memory bus with 6GB GDDR5 memory, memory clock of 1375MHz, 264GB / s of bandwidth capacity.
Looking at the specifications and market, the deluxe Edition Radeon HD7970 strays quite far. However, the FirePro W9000 supports ECC error checking correction, in order to avoid operator error caused by results of the operation that are not normal, or ones that may reduce the speed of operation.
Backplane I / O section: due to being a specialized product, the video output panel is equipped with six mini-DisplayPort 1.2, so that one may set up the six-screen Eyefinity. The upper right part of the 3-pin DIN is an HD3D PRO connection interface.
Compared to the consumer-grade Radeon HD7970, the power input has been moved to the rear, which has a PCIe 6 pin, 6 +2 pin each. Originally the power input was into a synchronous interface connector able to be used with the FirePro S400 synchronization module which can utilize multi-GPU state, multi-screen output, rendering, or HD3D display signal synchronization.
Back at a glance: the abolition of the less desirable areas of expertise BIOS switch, but retains the multi-card Crossfire X Goldfinger connection, and up to a 4-WAY configuration.
In the next test, instead of the W9000, we prepared the similar Tahiti XT GPU Radeon HD 7970 GHz Edition, as well as the RV770 GPU, which is based on the Radeon HD4870 FirePro V8750 series of performance tests.
CPU: Intel Core i5-3570K @ 4800MHz 1.35v
Motherboard: ASUS Maximus V Formula
Memory: Samsung DDR-1600 4GB x 2 @ 2133MHz
Power Supply: THORTECH Thunderbolt 1200W
Storage: ADATA S511 120GB SSD
Graphics card / driver version: FirePro W9000 / 8.982.8.1
FirePro V8750 / 8.911.3.4
Radeon HD 7970 GE / 22.214.171.124
The first industry-standard Benchmark software – SPECviewperf 11.
SPECviewperf 11 provides test simulations, including the LightWave, CATIA, EnSight, Maya, Pro / ENGINEER, SolidWorks, TCVIS, Simens NX industry standard drafting, visualization, simulation, and 3D operating environment test simulations.
The difference in test conditions being the anti-aliasing (AA) closed test with 8x anti-aliasing. (Each vertical axis sub-tested by name)
Besides the SPECviewperf standards, we also utilized the SPECapc Organization developed SPEC test.
The previously mentioned A320M workstation platforms test the difference between the two examples. SPECapc favors the concept of the default script, standardized test steps and export test results to quantify the real tools. It can be said to be closest to real world performance. During this sequence we selected Lightwave, PTC Creo, MAYA, SIMENS NX along with the SPECapc script to be tested.
SPECapc for Creo 2.0: CREO 2.0 PTC has developed design solutions used in place of PRO / E, which covers 2D, 3D design and modeling. (Vertical axis display card model)
SPECapc for Creo 2.0: CREO 2.0
SPECapc for Maya 2012: the well-known 3D animation and rendering software. (Each vertical axis sub-tested by name)
SPECapc for Maya 2012
SPECapc for LightWave 9.6
(vertical axis is the name of the various sub-tests)
SPECapc for Siemens NX8
(Each vertical axis sub-tested by name)
Next Direct X and general-purpose operational testing: Due to the use of the RV770 GPU FirePro V8750 for OpenCL, the models which do not support Direct X 11 will not be included in this section.
DirectX section: We will use the 3DMark 11 with Unigine Heaven.
The OpenCL and DirectCompute section: We will use the LUXMARK 2.0 ComputeMark. Different test items in different colors bar as the difference between these various tests.
Summing up the above results, we can see that with the introduced GCN architecture the FirePro W9000 has not only established its powerful performance, but you can also see the professional card software optimized for performance and have a great advantage. You can see in most professional tests the FirePro W9000 substantially beats out its cousin, the HD 7970, and making use of both generation’s architecture, the FirePro V8750 also relies on the optimization of hot pursuit FLOPS performance more frequently than the HD 7970. So much so that its performance in the test displayed its clear victory.
On the other hand, you can see during the general part of the operation, it is relying completely on brute force. The HD 7970 GE W9000 computing clock has a significant influence on the final performance results. Regarding the performance of OpenCL, with DirectCompute-type of general operations, those who would like to participate in the Bitcoin Mining Gold Rush, may utilize this faster graphics card for in home-use, as may be the optimal choice.
Regarding professional application of the graphics card; in addition to its hardware architecture, another major value lies in its particular software optimization. Certainly there are various degrees of quality in graphics card software, the user’s opinion is sometimes affected by the software quality. If the authorized version causes trouble or loss due to completion dates being postponed, the benefits may be much higher after the purchase of a professional-grade graphics card. AMD and Sapphire PGS offer such services for the field of professional users, such as providing certification and related consulting; allowing for a complete solution to help specialized professionals to maximize output.