AMD Radeon HD 6990

AMD Radeon HD 6990

AMD Radeon HD 6990: A Legend of the Past in the Era of Modern GPUs. A Comprehensive Review in 2025

Introduction

The AMD Radeon HD 6990, released in 2011, became a symbol of an era when manufacturers competed to create the most powerful solutions for enthusiasts. Despite its venerable age, it still piques the interest of retro hardware lovers and collectors. But how relevant is this model in 2025? Let’s delve into the details.


1. Architecture and Key Features

TeraScale 3: Two Hearts in One Shell

The HD 6990 is built on the TeraScale 3 architecture, housing two Cayman XT graphics processors (GPUs) on a single circuit board. This solution allowed AMD to create a flagship card with record-breaking performance for its time.

- Process Technology: 40 nm (for comparison, modern GPUs use 5–7 nm).

- Clock Speed: Base 830 MHz, with the potential to boost to 880 MHz in "Turbo" mode.

- CrossFireX: Support for multi-GPU configuration (up to 4 GPUs), but for the HD 6990, one card is sufficient— it already contains two chips.

What’s Missing?

- Ray Tracing (RTX): 2010s technologies did not support hardware ray tracing.

- FidelityFX/DLSS: No equivalent AI scaling or image optimization.

Unique Feature: A BIOS switch for toggling between silent mode and maximum performance.


2. Memory: GDDR5 and Its Capabilities

The HD 6990 is equipped with 4 GB of GDDR5 (2 GB per GPU) with a 256-bit bus for each chip.

- Bandwidth: 160 GB/s (total for two GPUs).

- Actual Efficiency: Due to memory being split between chips, the available memory for gaming is 2 GB per GPU, which is critically low in 2025, even for 1080p in modern titles.

Tip: This is sufficient for running older games (2010–2015), but in new releases, even on minimum settings, dips may occur due to insufficient VRAM.


3. Gaming Performance: Nostalgia vs. Reality

In 2025, the HD 6990 is chosen for retro gaming. FPS examples (High settings, 1080p):

- The Witcher 3 (2015): ~35–40 FPS (without HD textures).

- GTA V (2013): ~50–60 FPS.

- CS:GO: ~120–150 FPS.

Higher Resolutions:

- 1440p: In games post-2017—less than 30 FPS.

- 4K: Only for indie projects or older strategy games.

Ray Tracing: Not supported, either in hardware or through drivers.


4. Professional Tasks: Capabilities and Limitations

The card is compatible with OpenCL 1.2, but its potential for professional tasks is modest:

- Video Editing: Suitable for use in Adobe Premiere Pro at resolutions up to 1080p, but 4K rendering will take an unacceptable amount of time.

- 3D Modeling: In Blender or Maya—only for basic scenes.

- Scientific Calculations: Lack of support for modern APIs (e.g., ROCm) makes it useless for machine learning.

Main Drawback: No equivalent to CUDA, limiting compatibility with specialized software.


5. Power Consumption and Heat Dissipation

- TDP: 375 W—on par with modern RTX 4090!

- Power Supply Recommendations: At least 750 W with a buffer (better to aim for 850 W for stability).

- Cooling: Turbine system with a single fan. Noise level—up to 55 dB under load.

Case Advice: Good ventilation is essential. Minimum 3 case fans (2 intake, 1 exhaust). Avoid compact cases—the card is 29.5 cm long and will take up two slots.


6. Comparison with Competitors

2011:

- NVIDIA GeForce GTX 590: Competitor with similar specifications but lower TDP (365 W). In games, the HD 6990 was 10–15% faster.

2025:

- AMD Radeon RX 6400 ($100): A lower-end modern model. Improved power efficiency, supports DirectX 12 Ultimate, and comes with 4 GB of GDDR6.

- NVIDIA GeForce GTX 1650 ($150): Better driver optimization and more VRAM.

Conclusion: The HD 6990 falls short even when compared to budget newcomers in 2025 for modern tasks.


7. Practical Tips

- Power Supply: Only from reputable brands (Seasonic, Corsair). Cheap PSUs may not handle the load.

- Compatibility: PCIe 2.0 x16. Modern motherboards support the card but check for BIOS updates.

- Drivers: The last version is from 2021 (Adrenalin 21.5.2). Windows 10/11 work, but there may be issues with newer games.

Important: The card does not support UEFI Secure Boot—booting the OS may present problems.


8. Pros and Cons

Pros:

- Legendary status and collectibility.

- High performance in 2010s games.

- Unique design with a lit logo.

Cons:

- Outdated architecture.

- High power consumption.

- Lack of support for modern technologies.


9. Final Verdict: Who Is the HD 6990 For?

- Retro Gamers: For immersing in games from 2000–2015 without needing to upgrade an old PC.

- Enthusiasts: Those building a “time machine” rig with hardware from the height of CrossFire/SLI.

- Budget Builds: If the card is obtained for free or at a symbolic price (in the second-hand market—$30–50).

Why You Shouldn’t Buy the HD 6990 in 2025?

It won't handle modern games and software, and its operation will be costly due to high power consumption. Better to consider even budget newcomers like the RX 6500 XT or Intel Arc A380.


Conclusion

The AMD Radeon HD 6990 is a relic that reminds us of a time when power was valued over efficiency. In 2025, it should be viewed as an artifact of IT history rather than as a working tool. However, for those looking to breathe new life into an old PC or build a retro system, it can still serve as the heart beating to the rhythm of the 2010s.

Basic

Label Name
AMD
Platform
Desktop
Launch Date
March 2011
Model Name
Radeon HD 6990
Generation
Northern Islands
Bus Interface
PCIe 2.0 x16
Transistors
2,640 million
Compute Units
24
TMUs
?
Texture Mapping Units (TMUs) serve as components of the GPU, which are capable of rotating, scaling, and distorting binary images, and then placing them as textures onto any plane of a given 3D model. This process is called texture mapping.
96
Foundry
TSMC
Process Size
40 nm
Architecture
TeraScale 3

Memory Specifications

Memory Size
2GB
Memory Type
GDDR5
Memory Bus
?
The memory bus width refers to the number of bits of data that the video memory can transfer within a single clock cycle. The larger the bus width, the greater the amount of data that can be transmitted instantaneously, making it one of the crucial parameters of video memory. The memory bandwidth is calculated as: Memory Bandwidth = Memory Frequency x Memory Bus Width / 8. Therefore, when the memory frequencies are similar, the memory bus width will determine the size of the memory bandwidth.
256bit
Memory Clock
1250MHz
Bandwidth
?
Memory bandwidth refers to the data transfer rate between the graphics chip and the video memory. It is measured in bytes per second, and the formula to calculate it is: memory bandwidth = working frequency × memory bus width / 8 bits.
160.0 GB/s

Theoretical Performance

Pixel Rate
?
Pixel fill rate refers to the number of pixels a graphics processing unit (GPU) can render per second, measured in MPixels/s (million pixels per second) or GPixels/s (billion pixels per second). It is the most commonly used metric to evaluate the pixel processing performance of a graphics card.
26.56 GPixel/s
Texture Rate
?
Texture fill rate refers to the number of texture map elements (texels) that a GPU can map to pixels in a single second.
79.68 GTexel/s
FP64 (double)
?
An important metric for measuring GPU performance is floating-point computing capability. Double-precision floating-point numbers (64-bit) are required for scientific computing that demands a wide numeric range and high accuracy, while single-precision floating-point numbers (32-bit) are used for common multimedia and graphics processing tasks. Half-precision floating-point numbers (16-bit) are used for applications like machine learning, where lower precision is acceptable.
637.4 GFLOPS
FP32 (float)
?
An important metric for measuring GPU performance is floating-point computing capability. Single-precision floating-point numbers (32-bit) are used for common multimedia and graphics processing tasks, while double-precision floating-point numbers (64-bit) are required for scientific computing that demands a wide numeric range and high accuracy. Half-precision floating-point numbers (16-bit) are used for applications like machine learning, where lower precision is acceptable.
2.601 TFLOPS

Miscellaneous

Shading Units
?
The most fundamental processing unit is the Streaming Processor (SP), where specific instructions and tasks are executed. GPUs perform parallel computing, which means multiple SPs work simultaneously to process tasks.
1536
L1 Cache
8 KB (per CU)
L2 Cache
512KB
TDP
375W
Vulkan Version
?
Vulkan is a cross-platform graphics and compute API by Khronos Group, offering high performance and low CPU overhead. It lets developers control the GPU directly, reduces rendering overhead, and supports multi-threading and multi-core processors.
N/A
OpenCL Version
1.2
OpenGL
4.4
DirectX
11.2 (11_0)
Power Connectors
2x 8-pin
Shader Model
5.0
ROPs
?
The Raster Operations Pipeline (ROPs) is primarily responsible for handling lighting and reflection calculations in games, as well as managing effects like anti-aliasing (AA), high resolution, smoke, and fire. The more demanding the anti-aliasing and lighting effects in a game, the higher the performance requirements for the ROPs; otherwise, it may result in a sharp drop in frame rate.
32
Suggested PSU
750W

Benchmarks

FP32 (float)
Score
2.601 TFLOPS

Compared to Other GPU

FP32 (float) / TFLOPS
2.693 +3.5%
2.55 -2%
2.509 -3.5%