ATI Radeon HD 5490

Power and Cooling — TDP, PSU recommendation, connector requirements

The ATI Radeon HD 5490 is an exceptionally power-frugal part, with a thermal design power of just 19 W. That figure places it in the same territory as many low-end laptop GPUs or integrated graphics solutions, making it suitable for slim desktop builds and office-oriented systems that prioritize quiet operation over brute force. The board is single-slot in design, which means it will not obstruct adjacent PCIe slots or require elaborate case airflow modifications. Its length of 170 mm (6.7 inches) further underscores its compact footprint, allowing installation in most small-form-factor chassis without clearance concerns.

Power delivery is remarkably simple: the card requires no auxiliary power connectors whatsoever. The data indicates that a 200 W power supply unit is sufficient for a system housing this GPU, which is a modest recommendation compared to the typical 500–650 W requirements seen in contemporary gaming cards. For a system with a low-power CPU and minimal peripherals, even a basic 200 W PSU provides ample headroom. The absence of external power connectors also reduces cable management complexity, a clear advantage for builders aiming for a clean interior.

The 40 nm process node, manufactured by TSMC, contributes directly to this low-power profile. With 292 million transistors packed into a 59 mm² die, the transistor density works out to 4.9M per mm² — a modest figure by modern standards but appropriate for the era. The TeraScale 2 architecture, belonging to the Evergreen generation (HD 5400), was designed with efficiency in mind, and the 19 W TDP reflects that design philosophy. In practice, the card will generate minimal heat, so even a basic passive or low-speed fan heatsink suffices. The slot width of single-slot also means no extra mounting brackets or dual-slot cooling shrouds are needed.

For system integrators, the PCIe 2.0 x16 bus interface is fully sufficient for the card’s bandwidth needs; the 8.000 GB/s memory bandwidth does not stress the bus. The power connector requirement of “None” is unambiguous — no six-pin or eight-pin PCIe power cables are required, simplifying upgrades in pre-built systems with limited PSU headroom. The end-of-life production status suggests that thermal solutions are mature and well-understood, with no surprises expected in cooling behavior.

Memory Subsystem — VRAM size/type, bus width, bandwidth and what it means for high resolutions

The HD 5490 comes equipped with 1024 MB of GDDR3 memory, a configuration that was entry-level at the time of its release. The memory type is GDDR3, running at an effective 1000 Mbps (500 MHz base clock). The memory bus is 64 bits wide, which is narrow by contemporary standards — most modern discrete GPUs use 128-bit or wider interfaces. This 64-bit bus, combined with the modest memory clock, yields a total bandwidth of 8.000 GB/s.

That bandwidth figure is the single most limiting factor for this card’s performance at higher resolutions. At 1080p, the pixel workload alone can approach or exceed 2.600 GPixel/s, and the card’s pixel rate is already capped at that figure. The texture rate of 5.200 GTexel/s is similarly constrained. The 8.000 GB/s bandwidth means that texture streaming and frame buffer access will become bottlenecks in GPU-bound scenarios, particularly when anti-aliasing or high-resolution textures are enabled.

For 720p and lower resolutions, the memory subsystem is adequate for casual gaming and media playback. However, at 1440p or 4K, the combination of a 64-bit bus and GDDR3 would result in severe frame pacing issues and stutter, as the card cannot move data fast enough to feed the shader array. The 1024 MB capacity is also marginal for modern game assets; many titles require more than 1 GB even at medium settings. For high-resolution work, the data clearly indicates that this card is not a viable option — the bandwidth is simply too low to sustain smooth frame delivery.

That said, for 2D desktop workloads, video playback, and legacy applications, the memory subsystem performs without complaint. The 8.000 GB/s bandwidth is more than enough for framebuffer operations at 1080p in non-3D contexts. The 64-bit bus also reduces PCB complexity and power consumption, aligning with the card’s overall design goals. Users should view this memory configuration as a deliberate trade-off: low power and cost at the expense of high-resolution gaming capability.

Ray Tracing and Feature Set — RT/tensor cores, API support from facts

The HD 5490 does not feature dedicated ray tracing cores or tensor cores — the data lists both as null. This is expected, as the TeraScale 2 architecture predates hardware-accelerated ray tracing by nearly a decade. Similarly, there is no support for variable rate shading, mesh shaders, or other modern DirectX 12 Ultimate features. The card’s feature set is firmly rooted in the DirectX 11 era.

API support includes DirectX 11.2 (with 11_0 feature level) and OpenGL 4.4. Vulkan support is absent, which means the card cannot take advantage of modern cross-platform graphics APIs that are increasingly common in newer game engines. DirectX 11.2 is still widely supported in many titles, but the lack of Vulkan excludes the card from a growing segment of the gaming library. OpenGL 4.4 provides compatibility with older applications and some productivity software, though it is not a substitute for the more modern API.

The absence of ray tracing and tensor cores means features like DLSS, FSR, or any form of AI-accelerated upscaling are unavailable. The card relies purely on traditional rasterization. The shading units number 80, with 8 texture mapping units and 4 render output units. This configuration yields a pixel fill rate of 2.600 GPixel/s and a texture fill rate of 5.200 GTexel/s. FP32 compute is rated at 104.0 GFLOPS — a figure that is negligible for compute-intensive workloads but sufficient for basic shader operations in older games.

Display outputs include 1x DVI, 1x HDMI 1.3a, and 1x VGA. The HDMI 1.3a standard supports 1080p output and basic audio passthrough, but lacks the bandwidth for 4K at high refresh rates. The VGA output is a nod to legacy monitors and projectors. In essence, the feature set is what one would expect from a low-end card of its generation: no advanced rendering capabilities, but full support for the APIs and display standards of its time.

How It Compares — position vs each nearest rival

The Fact Pack lists no nearest rivals for the ATI Radeon HD 5490. The nearestRivals array is empty, as are the benchmarks array. This absence of direct comparisons makes it impossible to position the card against specific competing products from NVIDIA, Intel, or other AMD generations. The percentileVsAllGpus field, however, provides a useful global context: the card sits at the 50th percentile among all GPUs in the database. That median placement suggests it outperforms roughly half of all listed GPUs — a surprising result for a low-end part, but one that reflects the large number of very old or integrated graphics entries in the database.

Without rival names, scores, or deltaPct values, any comparison must remain qualitative. The card’s 19 W TDP and 200 W PSU recommendation position it firmly in the low-power segment. Its 1024 MB GDDR3 memory and 8.000 GB/s bandwidth are typical of entry-level cards from its era, but they are far below the specifications of even mid-range cards from the same generation. The 50th percentile ranking implies that while the HD 5490 is not a performance leader, it is also not the absolute bottom of the barrel — it outperforms many integrated graphics solutions and older discrete GPUs.

The end-of-life production status suggests that the card has been superseded, but its 40 nm process node and 292 million transistor count indicate a mature design with well-established driver support. The absence of launch MSRP in the data means no pricing context can be provided. In the broader landscape of the Evergreen generation, the HD 5490 sits at the entry point, with the data showing no direct successor within the same die family. The predecessor is listed as Radeon R700, and the successor as Northern Islands — indicating a clear generational progression.

Benchmark Performance — analyze scores vs rivals with exact % deltas

The benchmark data for the ATI Radeon HD 5490 is entirely absent. The benchmarks array is empty, and the avgBenchmarkScore is 0. This means no synthetic or real-world performance scores are available for direct analysis. The percentileVsAllGpus figure of 50 provides the only quantitative performance context, but without a benchmark score or rival comparisons, it cannot be translated into specific deltas or percentage differences.

The compute capabilities, however, can be interpreted from the raw specification data. The FP32 performance of 104.0 GFLOPS is a theoretical ceiling for shader work. The pixel rate of 2.600 GPixel/s and texture rate of 5.200 GTexel/s represent the maximum fill rates under ideal conditions. These figures are consistent with a card designed for 720p gaming at low to medium settings in titles from its release era (October 2012). At 1080p, the card would struggle to maintain playable frame rates in most 3D games, given the bandwidth and fill rate constraints.

The lack of benchmark scores means the 50th percentile ranking cannot be validated against specific workloads. It is plausible that the ranking is heavily weighted by the card’s low TDP and broad compatibility, rather than raw performance. In the absence of rival deltas, the analysis must rely on the specification-derived limits: the 64-bit memory bus, 4 ROPs, and 8 TMUs all point to a card that is roughly an order of magnitude slower than contemporary mid-range offerings. The 19 W TDP, however, means that a system equipped with this card will have exceptionally low thermal output and power draw — a trade-off that may be acceptable for basic office productivity, legacy gaming, or media center use.

Given the empty benchmarks and nearestRivals fields, the data cannot support any percentage-based performance comparisons. The page must therefore present the HD 5490 on its own merits, using the theoretical fill rates and bandwidth as the primary performance indicators. The 50th percentile ranking offers a neutral reference point, but without further context, it remains an ambiguous statistic. The card’s end-of-life status and 2012 release date further contextualize its performance: it is a product of its time, designed for a specific low-power niche, and not intended to compete with performance-oriented GPUs.