AMD FirePro
AMD FirePro was a brand of professional-grade graphics processing units (GPUs) developed by Advanced Micro Devices (AMD) for workstations, servers, and high-performance computing environments, primarily targeting applications in computer-aided design (CAD), computer-generated imagery (CGI), engineering simulation, media and entertainment, and scientific visualization.[1] These GPUs were engineered for stability, with features like optimized drivers, Independent Software Vendor (ISV) certifications for professional software such as Autodesk AutoCAD and SolidWorks, and support for multi-display configurations up to 4K resolutions, distinguishing them from consumer-oriented Radeon cards.[2] The FirePro line originated from AMD's acquisition of ATI Technologies in October 2006, which integrated ATI's established expertise in professional graphics into AMD's portfolio.[3] The brand was formally introduced in 2010 alongside the transition from ATI branding, with early models like the FirePro V7900 and V5900 based on the Evergreen architecture, delivering mid-to-high-end performance for design and visualization tasks without exceeding standard power envelopes.[4] Over the years, the series expanded to include the W-series desktop cards and S-series for servers, incorporating advanced architectures such as Graphics Core Next (GCN); notable releases included the FirePro W9100 with 16GB of GDDR5 memory in 2014 for handling large datasets in 4K workflows, and the W9100 32GB variant in 2016, which set a record for memory capacity in professional GPUs at the time.[5][6] FirePro cards emphasized enterprise reliability, including ECC memory support in select models, remote management capabilities via AMD Software: PRO Edition, and compatibility with virtual desktop infrastructure (VDI) for secure, multi-user environments.[7] The brand was retired in July 2016 at SIGGRAPH, succeeded by the Radeon Pro series, which continued the professional focus under a unified naming scheme aligned with AMD's consumer products while preserving certifications and driver optimizations.[8][9]History
Origins and Early Development
The FireGL line originated in 1995 with the German company Spea Software AG, which introduced the first FireGL graphics card as a high-performance solution for 3D graphics in professional applications. This initial board, known as the Fire GL, utilized the 3Dlabs GLINT coprocessor, an early OpenGL accelerator designed to deliver workstation-level 3D rendering on desktop systems. Targeted at CAD, visualization, and multimedia workflows, it supported high-resolution displays up to 1280 × 1024 pixels with true-color imaging and a 75 Hz refresh rate, offering superior 3D graphics performance compared to high-end workstation cards of the time. The card's focus on OpenGL compatibility enabled efficient handling of complex 3D models in software like AutoCAD, marking a key milestone in bringing advanced 3D acceleration to Intel-based personal workstations running Windows NT or DOS.[10] In November 1995, Spea Software AG was acquired by Diamond Multimedia Systems for approximately £71 million, integrating the FireGL brand into Diamond's portfolio and shifting its development toward sustained workstation rendering capabilities. Under Diamond's ownership from late 1995 to 2001, the FireGL series evolved with models like the Fire GL 1000 Pro, which incorporated chips such as the 3Dlabs Permedia 2 for enhanced 3D acceleration and support for professional rendering tasks. These cards emphasized reliability for demanding environments, including PCI and early AGP interfaces with 8 MB of memory, prioritizing OpenGL optimization for CAD/CAM/CAE applications over consumer gaming features. This period solidified FireGL's reputation for stable, high-fidelity 3D output in professional settings, with a focus on texture mapping and hardware-accelerated rendering to reduce CPU load in visualization pipelines.[11] Diamond's FireGL division, reorganized as FGL Graphics, was acquired by ATI Technologies in March 2001 for $10 million ($2.7 million in cash plus $7.3 million contingent on performance), bringing the brand and its 35 engineers under ATI's control to bolster its professional graphics offerings. ATI quickly transitioned FireGL to its own architectures, introducing cards like the FireGL X2 in 2003 based on the TeraScale 1 microarchitecture (Radeon 9700 series), which provided unified shader support and improved floating-point precision for CAD and CGI workloads. These boards optimized both OpenGL and emerging DirectX APIs for professional software, enabling faster rendering of complex scenes in tools for digital content creation and engineering design, while maintaining certifications for stability in multi-application environments.[12][13][14] AMD's $5.4 billion acquisition of ATI in October 2006 integrated the FireGL lineup into AMD's broader professional graphics strategy, preserving its focus on workstation-grade performance amid the convergence of CPU and GPU technologies. This merger allowed FireGL to leverage AMD's resources for enhanced driver support and optimization, continuing its evolution as a dedicated professional series until the brief transition to the FirePro branding in 2008. Key early milestones, such as the GLINT-based 3D acceleration and the pivot to dual OpenGL/DirectX support, established foundational concepts for certified, application-specific graphics acceleration in professional computing.[3][15]Launch of FirePro Brand
In 2008, AMD introduced the FirePro brand as a unified professional graphics lineup, replacing the legacy ATI FireGL series for 3D workloads and the FireMV series for multi-monitor applications, drawing from the pre-acquisition FireGL heritage established in the early 2000s. The launch marked AMD's strategy to streamline its professional offerings under a single banner, with the initial FirePro 3D Series (designated V000) based on the TeraScale architecture derived from the 2007 R600 GPU. This rebranding aimed to provide certified, optimized performance for enterprise environments while leveraging consumer-grade silicon efficiencies for cost competitiveness.[16][17] The FirePro series initially targeted professional workloads such as computer-aided design (CAD), computer-generated imagery (CGI), digital content creation (DCC), and emerging general-purpose GPU (GPGPU) computing through stream processing capabilities first enabled by the R600 architecture. These cards supported early parallel computing tasks in scientific visualization and simulation, emphasizing reliability over gaming-oriented features. AMD positioned FirePro as a bridge between high-end consumer graphics and workstation demands, with hardware enabling up to four displays via DisplayPort and DVI outputs.[17][16] Key models released between 2008 and 2010 included the entry-level FirePro V3700 and V3750, mid-range V5700, high-end V8700, and flagship V8750, all featuring GDDR3 or GDDR5 memory configurations up to 2GB. These cards came with certified drivers optimized for independent software vendor (ISV) applications, such as AutoCAD and SolidWorks, ensuring stability and performance validation through AMD's collaboration with developers. For instance, the V8750, launched in July 2008 with an MSRP of $1,799, delivered enhanced 3D rendering speeds for complex models.[18][19][20] AMD's market strategy for FirePro emphasized workstation integration, including support for error-correcting code (ECC) memory to prevent data corruption in critical computations and double-precision floating-point performance at one-quarter single-precision rates for scientific and engineering simulations. The series also incorporated CrossFire technology for multi-GPU configurations, enabling scalable performance in professional setups. Early adoption occurred in sectors like architecture, engineering, and media production, with OEMs such as Dell and HP incorporating FirePro cards into certified workstations shortly after launch.[16][17]Evolution and Discontinuation
In 2011, AMD released the Vx900 series based on TeraScale 3 architecture, with high-performance models like the V7900 and V5900. In 2012, AMD advanced its FirePro lineup by transitioning to the Graphics Core Next (GCN) architecture, which enhanced compute capabilities for professional workloads. By August 2012, the Wx000 series introduced full GCN implementation, exemplified by the FirePro W9000 featuring 6 GB GDDR5 memory, delivering up to 1.95 billion triangles per second in geometry processing for CAD and visualization tasks.[21][7] From 2013 to 2015, AMD expanded the FirePro portfolio to address diverse professional ecosystems, including integrated solutions and advanced compute features. The D-Series debuted in July 2013 as dual-GPU configurations tailored for Apple's Mac Pro, with models like the D300, D500, and D700 providing scalable graphics performance equivalent to W7000, W8000, and W9000 cards, respectively, optimized for creative and engineering applications.[22][23] In 2014, the Wx100 and Wx300 series, such as the W8100 and W9100, incorporated Heterogeneous System Architecture (HSA) support to enable seamless CPU-GPU collaboration for heterogeneous computing tasks like simulation and rendering.[24] Meanwhile, the S-Series targeted server environments, with the FirePro S7150 launched in early 2016 featuring hardware virtualization via Multiuser GPU (MxGPU) technology, allowing up to 16 concurrent virtual desktops per card for secure, remote workstation access in data centers.[25] Throughout this period, key evolutions in FirePro included deeper integration of Eyefinity technology for seamless multi-monitor configurations—supporting up to six displays from a single card—and DirectGMA for low-latency data transfers between GPU memory and system applications, reducing bottlenecks in real-time workflows like financial trading and media editing.[7][26] AMD announced the discontinuation of the FirePro brand in July 2016, transitioning workstation products to the Radeon Pro lineup and server offerings to Radeon Instinct to unify branding under the broader Radeon ecosystem.[8] This rebranding streamlined marketing and development for both professional and consumer markets, aligning with the introduction of Polaris-based Radeon Pro WX series cards that addressed emerging needs in VR, 4K content creation, and design engineering. Existing FirePro hardware continued to receive driver support post-transition.[8]Features
Differences from Radeon Line
The AMD FirePro graphics cards were targeted at professional markets such as computer-aided design (CAD), computer-generated imagery (CGI), digital content creation (DCC), and general-purpose GPU (GPGPU) computing in workstations and servers, whereas the Radeon line focused on gaming and general consumer computing applications.[27] In terms of hardware, FirePro cards included support for error-correcting code (ECC) memory in select models to ensure data integrity during critical simulations and computations, a feature absent in most consumer Radeon GPUs. Additionally, FirePro provided higher double-precision floating-point (FP64) performance, with rates up to 1/4 of single-precision (FP32) capabilities in professional variants compared to 1/16 in equivalent Radeon consumer models, enabling more accurate scientific and engineering workloads.[28][29] Driver support for FirePro utilized the AMD Catalyst Pro edition, which offered certified independent software vendor (ISV) optimizations for stability in professional applications, including longer support cycles and enhanced OpenGL performance, in contrast to Radeon drivers that prioritized DirectX optimizations for gaming. FirePro cards underwent rigorous certification testing with professional software from vendors like Autodesk and Adobe to guarantee compatibility and reliability, an emphasis not present in the Radeon lineup.[27][30] FirePro designs often featured single-slot and low-profile form factors suitable for dense server environments, along with lower thermal design power (TDP) ratings for sustained operation, differing from the multi-slot, high-performance cooling configurations typical of Radeon gaming cards. These optimizations supported multi-monitor setups via Eyefinity technology in professional contexts.[27]Multi-Monitor Support
AMD Eyefinity technology, introduced in 2010 with the FirePro V-series graphics cards, enables support for multiple independent displays driven by a single GPU, utilizing DisplayPort and Mini-DisplayPort outputs to create expansive workspaces.[31] Early implementations in models like the FirePro V5800 provided up to three displays through three native DisplayPort connectors, facilitating basic multi-monitor setups for professional workflows.[32] Key configurations supported by Eyefinity include bezel correction, which adjusts image positioning to account for monitor bezels and ensure seamless visuals across screens; uniform scaling to maintain consistent image quality; and flexible topology arrangements, such as landscape or portrait orientations, ideal for CAD model reviews and large-scale control room deployments. Hardware advancements, particularly native DisplayPort 1.2 in later models like the FirePro W9000, allow for 4K resolutions (up to 4096x2160) at 60 Hz per port, while adapters enable compatibility with DVI and HDMI interfaces for broader connectivity.[33] In professional environments, FirePro's multi-monitor capabilities are optimized for applications like digital signage, financial trading desks requiring rapid data visualization, and engineering simulations, delivering high-resolution output across displays without performance degradation.[1] The GCN architecture era, starting in 2012, further expanded Eyefinity to reliably support up to six displays, incorporating stereoscopic 3D compatibility via AMD HD3D technology for immersive professional visualization.[34] Unlike consumer-oriented Radeon implementations, FirePro Eyefinity prioritizes driver stability and certification for sustained multi-monitor operation in mission-critical settings.[1]Compute and Acceleration Technologies
AMD FirePro graphics cards introduced stream processing capabilities with the R600 architecture in 2007, enabling parallel computing tasks for professional workloads through the initial FireStream lineup, which laid the foundation for GPGPU acceleration in subsequent FirePro products.[35] This architecture supported general-purpose computations on GPUs, allowing FirePro cards to handle data-intensive operations beyond traditional graphics rendering. DirectGMA (Direct Graphics Memory Access) technology in AMD FirePro GPUs facilitates low-latency data transfers directly between the CPU and GPU memory, bypassing the host system's main memory to minimize overhead and CPU utilization.[36] Optimized for professional applications such as video editing and real-time post-production, DirectGMA enables efficient peer-to-peer communication with third-party devices like SDI I/O cards, reducing latency in workflows involving high-bandwidth video streams.[37] The Heterogeneous System Architecture (HSA) was integrated into FirePro starting with 2015 GCN-based cards, providing a unified memory model that allows seamless collaboration between CPU and GPU without explicit data copying.[38] This enables developers to leverage both processors as a cohesive heterogeneous system for compute tasks, improving efficiency in parallel programming environments.[39] High-end FirePro models offer double-precision floating-point support reaching up to 1 TFLOPS, essential for accurate scientific simulations including molecular modeling.[7] For instance, the FirePro W9000 achieves 998.4 GFLOPS in double precision, supporting precision-demanding computations in fields requiring high numerical stability. ECC memory in select FirePro server variants further enhances reliability for compute-intensive operations by detecting and correcting data errors.[40] FireStream provided the early compute lineage for FirePro from 2008, with its integration allowing the professional line to evolve into a platform supporting OpenCL for GPGPU workloads akin to CUDA environments.[41] AMD's adoption of OpenCL 1.0 in 2008 enabled FirePro and FireStream hardware to execute portable, parallel code across GPUs, facilitating high-performance computing in professional settings without vendor lock-in.[42] By 2012, FireStream capabilities were fully folded into FirePro, unifying compute acceleration under the professional graphics brand.[43]Software and Certification
AMD FirePro graphics cards utilized enterprise-grade drivers tailored for professional environments, providing extended support for up to three years following each release to maintain stability and compatibility in demanding workflows. These drivers were specifically optimized for professional applications leveraging OpenGL 4.x and DirectX 11/12, ensuring consistent rendering and compute performance in fields such as CAD, engineering, and media production.[44][45] A key aspect of the FirePro software ecosystem was its comprehensive ISV certification program, involving rigorous testing and validation with over 100 independent software vendor applications, including examples like CATIA for mechanical design, Maya for 3D animation, and ANSYS for simulation and analysis. This collaborative process between AMD and ISVs focused on driver development to deliver enhanced stability, reduced crashes, and peak performance in certified software, often spanning 2-12 weeks of multi-wave testing per certification cycle.[46][47][48] Community-driven soft-mods allowed users to adapt consumer Radeon drivers for FirePro hardware, enabling gaming capabilities on professional cards, but such modifications posed risks including warranty invalidation and potential instability in certified professional applications.[49] Enterprise management was facilitated through the AMD Catalyst Control Center Pro Edition, which offered advanced features for remote monitoring, system optimization, and deployment configuration in large-scale professional setups.[50] After the 2016 rebranding of the professional graphics line, FirePro hardware received continued legacy support via the Radeon Pro Software suite until approximately 2021, maintaining ISV certifications and enterprise stability during that period. Legacy driver support ended around 2021, with no further updates planned thereafter.[51][52]Products
Workstation Series
The Workstation Series of AMD FirePro graphics cards encompassed desktop GPUs optimized for professional applications in computer-aided design (CAD), 3D modeling, and visualization, evolving from the earlier FireGL line acquired by ATI Technologies in 2002.[53] Prior to ATI's involvement, the FireGL brand originated in the mid-1990s with cards produced by Diamond Multimedia, utilizing 3Dlabs Permedia chipsets such as the Permedia 2 for models like the FireGL 1000 Pro, which featured up to 8MB of VRAM and targeted early 3D rendering tasks in professional workstations.[54] By the late 1990s and into 2001, subsequent iterations like the FireGL 1 and FireGL 10 advanced to 32-128MB of SGRAM or VRAM configurations, supporting enhanced texture mapping and OpenGL acceleration for emerging CAD and scientific visualization workflows.[55] Following ATI's acquisition of the FireGL team, the series transitioned to in-house architectures, with the FireGL lineup from 2001 to 2007 focusing on TeraScale 1-based cards for mid-range CAD performance. A representative model, the FireGL V7200 released in 2006, utilized the R580 GPU with 512MB GDDR3 memory, a 100W TDP, and support for DirectX 9.0c and OpenGL 2.0, enabling efficient handling of complex assemblies in applications like SolidWorks. These cards emphasized stability and driver certification for professional software, bridging the gap to higher-end rendering without the consumer-oriented features of the Radeon line. The rebranding to FirePro in 2008 introduced the 3D V000 series (2008-2010), built on the TeraScale 2 architecture for improved compute capabilities in CAD and digital content creation. Entry-level options like the FirePro V3800 featured 512MB DDR3 memory, 400 stream processors at 650 MHz, and a low 43W TDP, suitable for 2D/3D design review.[56] Higher-end variants, such as the FirePro V8750 with 2GB GDDR5 memory and 800 stream processors, delivered up to 1.2 TFLOPS of single-precision floating-point performance, while the top-tier FirePro V9800 reached approximately 2.4 TFLOPS for demanding simulation tasks.[57][58] This series prioritized ISV certifications for tools like AutoCAD, ensuring reliable performance in engineering environments.[59] From 2011 to 2012, the Vx900 and Wx000 series marked a shift to the Southern Islands architecture, targeting high-end visualization and media workflows with greater memory capacity. The flagship FirePro W9000, launched in 2012, incorporated 6GB GDDR5 memory on a 384-bit bus, 2048 stream processors, and 3.994 TFLOPS of single-precision performance, facilitating large-scale 3D modeling and real-time rendering.[60][7] These models supported ECC memory in select configurations to maintain data integrity during extended computations, a feature critical for precision engineering.[61] The Wx100 and Wx300 series (2013-2016), based on the Graphics Core Next (GCN) architecture, extended capabilities for ultra-high-resolution and compute-intensive tasks. Mid-range examples like the FirePro W7100 offered 8GB GDDR5 and 1792 stream processors for balanced CAD performance, while the FirePro W8100 provided 8GB GDDR5 and 2560 stream processors for advanced assembly visualization in SolidWorks.[62] At the pinnacle, the FirePro W9100 in its 32GB GDDR5 variant, released in 2016, supported workflows involving massive datasets and 8K-resolution outputs, achieving 5.2 TFLOPS single-precision and 2.6 TFLOPS double-precision for applications like structural analysis in AutoCAD and Siemens NX.[63][64] Across the series, optimizations for professional software ensured certified stability, with ECC support in high-end models like the W9000 and W9100 safeguarding against errors in critical design iterations.[2]Mobile Series
The Mobility FireGL series represented AMD's early efforts in professional mobile graphics prior to the FirePro brand launch in 2008, targeting portable CAD and design applications. These cards, part of the MX series, featured discrete GPUs optimized for laptop workstations with limited power envelopes. For instance, the FireGL V5200, released in 2005, utilized a 90 nm process with 256 MB GDDR3 memory and a 128-bit bus, delivering 425 MHz core clock speeds suitable for mobile engineering tasks.[65] The FirePro Mobile series, spanning 2008 to 2016, built on this foundation by rebranding and enhancing ATI's professional mobile lineup with AMD's Graphics Core Next (GCN) architecture, sharing core designs with the desktop workstation series for consistent performance tuning. These GPUs emphasized portability, incorporating 1-2 GB GDDR5 memory across 128-bit buses and thermal management to balance sustained workloads with battery efficiency in laptops. Key models included the FirePro M4000 (2012, GCN 1.0, 512 shaders at 600 MHz, 1 GB GDDR5, 33 W TDP) and M6000 (2012, GCN 1.0, 640 shaders at 750 MHz, 2 GB GDDR5, 43 W TDP), which supported DirectX 11 and OpenGL 4.6 for professional rendering.[66][67] A prominent example was the FirePro M6100, launched in 2013 on GCN 2.0 architecture with 768 shaders at up to 1.1 GHz, 2 GB GDDR5, and support for AMD Eyefinity multi-monitor technology adapted for laptop displays. These cards integrated into mobile workstations such as Dell Precision M6800 and 7510 models, as well as HP ZBook 15 G3 and 17 G2 series, enabling on-the-go CAD, simulation, and content creation. Optimizations included dynamic thermal throttling to extend battery life during light tasks while maintaining ISV certifications for applications like Adobe Premiere Pro, where AMD's OpenCL acceleration improved video editing workflows through collaborations starting in 2013.[68][69][70] Despite these advances, FirePro mobile GPUs faced inherent limitations from power constraints, typically delivering 20-30% lower performance than their desktop counterparts in sustained benchmarks due to TDPs capped at 50 W or below to accommodate laptop cooling and portability. The series concluded in 2016 with the transition to the Radeon Pro WX mobile lineup, which succeeded FirePro for professional mobile use.[71][72]| Model | Launch Year | Architecture | Shaders | Memory | TDP |
|---|---|---|---|---|---|
| FirePro M4000 | 2012 | GCN 1.0 | 512 | 1 GB GDDR5 | 33 W |
| FirePro M6000 | 2012 | GCN 1.0 | 640 | 2 GB GDDR5 | 43 W |
| FirePro M6100 | 2013 | GCN 2.0 | 768 | 2 GB GDDR5 | ~50 W |
Server Series
The AMD FirePro Server Series encompassed graphics processing units (GPUs) optimized for data center environments, emphasizing high-performance computing (HPC), virtualization, and remote graphics delivery. Introduced as an extension of the FirePro brand, these products targeted server deployments where parallel processing and resource sharing were critical, distinguishing them from workstation-focused models by supporting multi-user scenarios and dense compute clusters.[25] The FireStream Series, active from 2008 to 2012, served as early compute accelerators for HPC applications, leveraging GPU stream processing for scientific simulations and data analysis. A representative model, the FireStream 9270 (launched November 2008), featured a RV770-based architecture with 2 GB GDDR5 memory on a 256-bit interface, delivering 1.2 TFLOPS of single-precision performance and 0.24 TFLOPS in double-precision, while consuming 160 W via PCIe 2.0 x16. These cards enabled dense server installations for tasks like molecular modeling, marking AMD's initial push into GPGPU computing before integration into broader FirePro lines.[73][74] Succeeding FireStream, the FirePro S-Series (2013–2016) introduced specialized server GPUs with hardware virtualization via AMD Multiuser GPU (MxGPU) technology, facilitating virtual desktop infrastructure (VDI) and cloud-based graphics for up to 32 users per card in partitioned modes. The S7150 (January 2016), built on the Tonga architecture, offered 8 GB ECC GDDR5 memory (256-bit interface, 160 GB/s bandwidth), 2048 stream processors, and 150 W TDP in a low-profile, passively cooled PCIe 3.0 x16 design, supporting OpenCL 2.0 for compute workloads. Its dual-GPU variant, the S7150 x2, doubled memory to 16 GB and performance while maintaining 265 W TDP, ideal for VDI in enterprise servers. Earlier in the series, the S9170 (July 2015) stood out with 32 GB GDDR5 memory (512-bit interface, 320 GB/s bandwidth) on a Hawaii architecture, providing 5.24 TFLOPS single-precision and 2.62 TFLOPS double-precision at 275 W, excelling in memory-intensive HPC like AI training and large-scale simulations. These models prioritized error-correcting code (ECC) memory and power efficiency for 24/7 data center operations.[75][76][25] Complementing the S-Series, the FirePro Remote Series focused on remote workstation access in server settings, using PCoIP protocol for secure, low-latency graphics streaming. The R5000 (February 2013), based on the Pitcairn LE GPU, included 2 GB GDDR5 memory (256-bit, 102.4 GB/s bandwidth), 768 stream processors, and 150 W TDP via PCIe 3.0 x16, enabling up to 16 simultaneous users per card in data centers for collaborative design and engineering tasks. It supported four 1920x1200 displays remotely, with local mini-DisplayPort outputs for monitoring.[77][78] By 2016, the Server Series transitioned into the Radeon Pro and Radeon Instinct lines, with FirePro branding retired to streamline AMD's professional portfolio amid growing demand for AI and cloud compute.[8]| Model | Launch Year | Architecture | Memory | Key Performance | TDP | Interface |
|---|---|---|---|---|---|---|
| FireStream 9270 | 2008 | RV770 | 2 GB GDDR5 | 1.2 TFLOPS (single) | 160 W | PCIe 2.0 x16 |
| FirePro S7150 | 2016 | Tonga | 8 GB GDDR5 ECC | 3.77 TFLOPS (single) | 150 W | PCIe 3.0 x16 |
| FirePro S7150 x2 | 2016 | Tonga (dual) | 16 GB GDDR5 ECC | 7.54 TFLOPS (single) | 265 W | PCIe 3.0 x16 |
| FirePro S9170 | 2015 | Hawaii | 32 GB GDDR5 | 5.24 TFLOPS (single), 2.62 TFLOPS (double) | 275 W | PCIe 3.0 x16 |
| FirePro R5000 | 2013 | Pitcairn LE | 2 GB GDDR5 | 1.27 TFLOPS (single) | 150 W | PCIe 3.0 x16 |