Introduction to the AMD FirePro™ V7900
The AMD FirePro™ V7900 is AMD’s latest high-end workstation professional graphics card, addressing the same segments of the CAD/CAE, DCC and other professional markets as the ATI FirePro V7800 launched in 2010. Some of the key facts, differences and improvements relative to the older generation include:
• A ‘Cayman GL’ GPU with 1280 Stream Processors, and 2 GB GDDR5 memory.
• A single-slot form factor, with a power consumption of under 150 Watts
• Four DisplayPort connectors supporting DisplayPort 1.2 – a first in professional graphics
• PCI Express 2.1 Standard
• AMD technologies including PowerTune and GeometryBoost that deliver significant improvements in performance
AMD Eyefinity technology and new Workflows
AMD PowerTune technology
AMD PowerTune technology helps deliver higher performance that is optimized to the thermal limits of the GPU by dynamically adjusting the clock in real time based on an internally calculated GPU power usage.
AMD PowerTune technology is very different from existing and previous methods; rather than setting highest state GPU clock speeds based on a worst case TDP approach that can compromise performance in a majority of applications, AMD PowerTune technology can dynamically adjust the performance profile in real time to fit within the TDP envelope. An important point to note about PowerTune on the AMD FirePro™ V7900 and V5900 professional graphics cards is that the calibration information (that sets clocks based on power draw in the GPU) has been defined with today’s workstation applications in mind. Generally speaking, these applications have a very different GPU use profile from consumer GPUs. For example, they typically do a great deal of geometry processing, and significantly less pixel shading and texturing. When compared to previous generation workstation graphics accelerators, increases in performance on consumer applications and benchmarks are more modest than for professional applications.
AMD GeometryBoost technology
The AMD FirePro™ V7900 and V5900 professional graphics cards implement the new AMD GeometryBoost technology and so are designed to process geometry data at up to twice the rate of previous-generation GPUs. The implementation of this capability is driven by the need for professional graphics cards to process complex models and the emerging use of GPU tessellation (which creates or amplifies geometry within the GPU). AMD GeometryBoost has been accomplished by putting in place a dual graphics engine architecture which allows graphics workloads to be divided and dispatched for processing in a
highly efficient manner. The dual graphics engine is portrayed in the diagram below, and shows the presence of geometry and vertex assemblers in each engine (also linked to the tesselator units), that effectively instantiate the GeometryBoost feature.
VLIW4 thread processors
- 4-way co-issue
- All stream processing units now have equal capabilities (no more “T-unit”)
o Special functions (transcendental) occupy 3 of 4 issue slots
Allow better utilization than previous AMD VLIW5 design
- Similar performance with ~10% area reduction
- Simplified scheduling and register management
- Extensive logic re-use
Page 1 – Introduction
Page 2 – The V7900 Part I
Page 3 – The V7900 Part II
Page 4 – The Differences and Pictures
Page 5 – System Set-up and Software Used
Page 6 – SiSoftware Sandra 2011 Part I
Page 7 – SiSoftware Sandra 2011 Part II
Page 8 – Miscellaneous Professional Benchmarks
Page 9 – SPECviewperf® 11.0 and SPECapc’s
Page 10 – Conclusions and Award