AMD Radeon R9 390

The AMD Radeon R9 390 is an end-of-life graphics card built on the 28 nm process node, featuring the Grenada chip with a GCN 2.0 architecture. Its 8 GB of GDDR5 memory is a defining characteristic, placing it in a unique position relative to its performance class. The card was released on June 17, 2015, as part of the Pirate Islands (R9 300) generation, succeeding the Volcanic Islands series and preceding Arctic Islands. The data indicates a product that was formidable in its time but now shows its age in modern benchmarks.

Memory Subsystem

The R9 390's most prominent feature is its 8 GB GDDR5 memory capacity, which is unusually large for a card of its performance tier. This is paired with a 512-bit memory bus, a configuration that was aggressive for the era. The combination yields a substantial memory bandwidth of 384.0 GB/s, driven by a memory clock of 1500 MHz (6 Gbps effective). This high bandwidth is a direct result of the wide 512-bit interface, allowing the card to move large amounts of data simultaneously.

For high-resolution gaming, this memory subsystem is both a strength and a potential bottleneck. The 8 GB capacity is more than sufficient for storing high-resolution textures and frame buffers, even at 4K, which is a significant advantage over cards with smaller pools. However, the raw compute power of the GPU itself may limit actual performance at these resolutions. The bandwidth, while impressive on paper, is a product of the older GCN 2.0 design. Benchmark results show a 3DMark Steel Nomad DX12 score of 955, which places the card in the 3rd percentile of all GPUs, indicating that despite the large memory buffer, the card's overall execution speed is limited. The 64.00 GPixel/s pixel rate and 160.0 GTexel/s texture rate are tied to the memory clock and architecture, and these figures suggest the card can fill a screen with data quickly, but the processing of the data is where it lags behind modern competitors. In practice, the memory subsystem allows for high-quality asset loading without stuttering, but the shader performance will often be the limiting factor in demanding modern titles.

Who Should Consider It

Given its benchmark percentile of 3, the R9 390 is not suited for modern high-refresh-rate or high-fidelity gaming. The data suggests this card is for a very specific use case: enabling high-resolution textures in older games or non-demanding eSports titles. For a 1080p resolution, the card can handle older AAA titles from its release era at high settings, leveraging its 8 GB memory to apply ultra texture packs. At 1440p, the card's computational limits become more apparent, and users would need to lower settings significantly. The 8 GB VRAM is the primary draw, as it allows for texture quality settings that would exceed the memory capacity of many of its contemporaries.

This is not a card for 4K gaming with modern titles; the 5.120 TFLOPS FP32 performance is insufficient to maintain playable frame rates. Instead, the R9 390 is best suited for a user who has a library of older games or plays undemanding games like MOBAs or older first-person shooters, where the high memory bandwidth can shine. It is also a candidate for a secondary system or a dedicated media PC, but its 275 W TDP makes it a power-hungry choice for such tasks. The comparison data shows it is only 7.2% ahead of the AMD Radeon HD 7850, a much older card, which underscores its position in the low end of modern performance charts. The card's dual-slot design and 275 mm length require a reasonably spacious case, but its focus is on providing a large memory buffer rather than raw speed.

Power and Cooling

The R9 390 carries a TDP of 275 W, a high figure that reflects the power-hungry GCN 2.0 architecture on a 28 nm process. This power draw necessitates a robust cooling solution, and the card ships as a dual-slot design. The physical dimensions are 275 mm (10.8 inches) in length, 109 mm (4.3 inches) in height, and 36 mm (1.4 inches) in width, making it a long card that requires adequate clearance in the chassis. For power delivery, the card requires a single 6-pin and a single 8-pin PCIe power connector. This is a standard configuration for high-power cards of its generation.

AMD recommends a 600 W power supply for a system with this card. This PSU recommendation is not just about the card's TDP; it accounts for the transient power spikes and the rest of the system's draw. The data does not include a specific cooler type, but the dual-slot width and 275 W TDP indicate that a capable air cooler is necessary to manage thermal output. The combination of a high TDP and a 28 nm process means the card will generate significant heat, and users should ensure their case has good airflow. The power connector requirement is a critical consideration for older power supplies, as the 8-pin connector is not always present on lower-wattage units from that era. The power draw is a clear drawback, but it is a consequence of the card's high memory bandwidth and wide bus, which require constant power to maintain.

How It Compares

The R9 390's nearest rivals, based on average benchmark scores, reveal its position in the market. The data shows a tight grouping, with the R9 390 scoring 955.

The AMD Radeon Pro WX Vega M GL scores 957, a delta of -0.2%, making it essentially a performance tie. This comparison is interesting because the WX Vega M GL is a professional mobile GPU, while the R9 390 is a desktop consumer card. The near-identical scores indicate that despite different architectures and target markets, their overall execution in this specific benchmark is the same. The R9 390 offers 8 GB of VRAM, which is a significant advantage over the WX Vega M GL's likely smaller pool, but the compute performance is matched.

The NVIDIA GeForce RTX 3050 6 GB scores 1003, which is 4.8% higher than the R9 390. This is a notable delta, showing that a modern entry-level card like the RTX 3050 outperforms the older AMD card. The RTX 3050 also has a much lower TDP and access to modern features like hardware ray tracing, which the R9 390 lacks. While the R9 390 has more VRAM (8 GB vs 6 GB), the RTX 3050's architectural efficiency makes it the superior performer in this test.

The NVIDIA GeForce 410M also scores 1003, a 4.8% delta. This is a peculiar pairing, as the 410M is a very old, low-end mobile card. The identical scores suggest the benchmark is not strongly differentiating between these two cards, or that the 410M's score may be from a specific driver version or test condition. In practice, the R9 390's 8 GB VRAM and 512-bit bus are vastly superior, but the data shows a parity in this specific metric. This highlights that a single benchmark score can misrepresent real-world capabilities.

The AMD Radeon HD 7850 scores 891, which is 7.2% lower than the R9 390. This is a more logical comparison, as both are older AMD desktop cards. The R9 390 is a clear step up from the HD 7850, which is expected given its later release and larger memory configuration. The 7.2% lead is modest, however, indicating that the architectural improvements between generations were not massive in this specific test, and the R9 390's advantage lies primarily in its memory subsystem and higher core counts (2560 shading units vs the HD 7850's lower count).

Benchmark Performance

The primary benchmark data point for the R9 390 is a score of 955 in the 3DMark Steel Nomad DX12 test. This score places the card in the 3rd percentile of all GPUs, which is a strong indicator that it performs worse than 97% of other graphics cards in the database. This is a low position, confirming the card's status as an end-of-life product with limited modern capabilities. The average benchmark score is identical to the single test score at 955, suggesting there is a consistent performance level for this card in this workload.

The deltas to its rivals are critical for interpretation. The R9 390 is 0.2% behind the AMD Radeon Pro WX Vega M GL, a negligible difference. This means that in this test, the two cards are functionally equivalent. The 4.8% gap to both the NVIDIA GeForce RTX 3050 6 GB and the NVIDIA GeForce 410M is more significant. The R9 390 is nearly 5% slower than these cards, which is a clear performance deficit. Conversely, the R9 390 is 7.2% ahead of the AMD Radeon HD 7850. This shows that while the card is behind modern low-end parts, it still holds a lead over a previous-generation mid-range card. The data implies that the R9 390's raw compute power (5.120 TFLOPS FP32) is not competitive, and the high VRAM (8 GB) and bandwidth (384.0 GB/s) do not translate to better scores in this DX12 test. The card's architectural age, lacking modern features like async compute optimizations, likely hampers its performance in this workload.

FAQ

Q: What is the launch MSRP of the AMD Radeon R9 390?

A: The launch MSRP is 329 USD.

Q: How much video memory does the R9 390 have, and what type is it?

A: It has 8 GB of GDDR5 memory on a 512-bit bus, providing a bandwidth of 384.0 GB/s.

Q: What is the card's performance percentile relative to all other GPUs?

A: The card sits in the 3rd percentile, indicating it performs better than only 3% of all GPUs in the database.

Q: Does the R9 390 support hardware ray tracing?

A: No, the card does not have any RT cores or tensor cores, and it is based on the GCN 2.0 architecture.

Q: What power supply is recommended for a system using this card?

A: The suggested PSU is 600 W, and the card requires a 1x 6-pin + 1x 8-pin power connector configuration.

Q: What is the card's performance delta compared to the NVIDIA GeForce RTX 3050 6 GB?

A: The R9 390 scores 4.8% lower than the RTX 3050 6 GB in the benchmark.

Ray Tracing and Feature Set

The R9 390 does not include any dedicated ray tracing cores or tensor cores, as these are not listed in the specifications and the architecture is GCN 2.0, which predates hardware ray tracing. This is a significant limitation for modern gaming, as titles with ray tracing effects will run without hardware acceleration, forcing the card to rely on compute shaders, which would be extremely slow. The card's feature set is instead focused on its memory and compute capabilities. It has 2560 shading units, 160 texture mapping units, and 64 ROPs, which are the core building blocks for traditional rasterization.

For API support, the card is listed with DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.2.170. The DirectX 12 support is feature level 12_0, which is the baseline for DX12 and does not include some of the more advanced features found in higher feature levels. This means the card can run DX12 titles, but may not support all optional features. Vulkan 1.2.170 support is good, allowing for modern cross-platform API usage. The display outputs are 2x DVI and 1x HDMI 1.4a and 1x DisplayPort 1.2. The HDMI 1.4a output limits 4K output to 30 Hz, while the DisplayPort 1.2 can handle 4K at 60 Hz. This is another sign of the card's age, as modern cards typically have HDMI 2.1 and DisplayPort 1.4 or 2.0. The PCIe 3.0 x16 interface is standard for the era, and the card's bandwidth is sufficient for its performance level. The feature set is thus a mix of high memory capacity and outdated connectivity and compute features, making it unsuitable for modern ray-traced games but capable of running older titles with high texture settings.