AMD Radeon HD 7470A

The AMD Radeon HD 7470A is an end-of-life graphics module from the All-In-One (HD 7000) generation, built around the Redwood chip using TSMC's 40 nm process and AMD's TeraScale 2 architecture. Its delivery form is an MXM Module with an MXM-A (3.0) bus interface, and its display outputs are marked Portable Device Dependent. In the benchmark record, the benchmarks list is empty, the average benchmark score is 0, and the percentile versus all GPUs is 50.

Benchmark Performance

The benchmarks array for this card contains no submitted test runs, so the data shows no frames-per-second values and no synthetic score. The average benchmark score is recorded as 0, which in this entry is a consequence of the empty benchmark list rather than a measured result. The only global placement number is the percentile versus all GPUs: 50. That value places the card at the exact midpoint of the tracked GPU population, but it is an aggregate rank, not a workload-derived performance figure.

With no measured scores available, the silicon specification defines the card's compute boundaries. The pipeline contains 400 shading units, 20 texture mapping units, and 8 raster operation units. Peak FP32 throughput is 620.0 GFLOPS. The rasterization ceiling is listed as 6.200 GPixel/s, and the texture fill ceiling is 15.50 GTexel/s. These are theoretical peak values, not achievable framerates, and the gap between peak fill and real-world performance depends on driver state, application shading complexity, and memory access patterns.

The 620.0 GFLOPS figure is the highest compute figure in the specification. It is tied directly to the 400 shading units and represents the maximum single-precision arithmetic the shader array can complete in one second. The 20 texture mapping units deliver the listed 15.50 GTexel/s rate, while the 8 ROPs deliver 6.200 GPixel/s. The relationship between these two rates is set by the fixed hardware counts, and both are upper bounds that will be reached only in optimal conditions.

The memory clock is the only clock explicitly present in the entry: 800 MHz, with an effective data rate of 1600 Mbps. No base clock, boost clock, or game clock is recorded. Because the shader clock is not available, the performance data cannot be normalized to a clock-per-MHz basis, and any comparison that requires operating frequency is not possible from this record.

The API specifications provide the software reach of the card. DirectX 11.2 with the 11_0 feature level is listed, along with OpenGL 4.4. Vulkan is null in the entry, so the data does not claim Vulkan support. The 11_0 feature level defines the minimum DirectX feature set that applications can rely on, and it is the more informative number for compatibility assessment because it reflects the shader model and resource-binding limits.

Because the nearestRivals list is empty, there are no deltaPct values to cite. The record does not support a statement such as "a specific percentage faster than a competitor," and there is no named rival score to place beside the HD 7470A. The only comparative number remains the percentile of 50, which is exactly centered.

Memory Subsystem

Memory capacity is 1024 MB of DDR3. The bus width is 64 bits, and the registered bandwidth is 12.80 GB/s. The effective memory data rate is 1600 Mbps, which the entry pairs with a memory clock of 800 MHz. When a memory bus is 64 bits wide, the total bandwidth is bound by that narrow datapath; in this case the listed 12.80 GB/s is the full budget for all frame buffer reads, writes, texture fetches, and display refresh traffic.

At 1024 MB, the frame buffer size is fixed and relatively modest relative to modern high-end parts, though the data does not include a comparison set. For high resolutions, the limiting factor is more likely the bandwidth than the raw capacity. High resolution increases the number of pixels that must be written to the frame buffer and the volume of texture data fetched per frame. A 64-bit bus with a 12.80 GB/s transfer ceiling will be stressed by those workloads much sooner than a wider bus with the same memory type.

The DDR3 memory type, the 800 MHz clock, and the 1600 Mbps effective rate are all part of the same memory configuration. The effective rate is the double-data-rate figure, while the 800 MHz value is the base memory clock used to derive it. Because the memory type and bus width are fixed, the 12.80 GB/s bandwidth is a constant specification for every unit of this module.

For anything approaching high-resolution rendering, the combination of 1024 MB capacity and 12.80 GB/s bandwidth is the structural constraint. The data does not include measured resolution or settings results, so any statement about specific quality levels remains an inference from the memory specification. The math of the memory subsystem is straightforward: narrow bus, standard DDR3 data rate, and modest total bandwidth.

Who Should Consider It

The generation string, All-In-One (HD 7000), identifies the intended system class for this module. The form factor is MXM-A (3.0), and the slot width is listed as MXM Module, which reinforces that this card is meant to be integrated into a chassis rather than installed into a desktop PCIe slot. The display outputs are Portable Device Dependent, meaning the host system dictates the physical display connections rather than the card itself.

The production status is End-of-life, and the release date is 2012-01-04. That date places the product at an earlier point in the database's collection, and the end-of-life status indicates it is no longer an active product in the market. The data does not provide a successor or predecessor, so there is no offical lineage position beyond the All-In-One generation label.

Given the 400 shaders, 620.0 GFLOPS FP32, and 6.200 GPixel/s pixel rate, the card is positioned as a low-power computing module. With 1024 MB of DDR3 memory on a 64-bit bus, the expected use is a system that needs graphical output without demanding sustained high-resolution workloads. The empty benchmark list means there is no measured evidence to recommend this card for high-detail gaming or compute acceleration.

The percentile of 50 indicates that, among all GPUs tracked in the database, this card occupies the middle of the distribution. That is a neutral position: it is not an extreme low-end point, but it is also not situated near the top. For an all-in-one system with moderate graphical requirements and a compact MXM form factor, the specification pattern is coherent. For a user seeking strong high-resolution performance, the data does not support that expectation.

Power and Cooling

The thermal design power for this module is 18 W. That is the only power-consumption figure in the entry. No power connector type is listed, and no suggested PSU wattage is recorded. Because the record lacks a PSU recommendation, there is no supported statement about a required system power supply size.

The slot width is MXM Module, and the bus interface is MXM-A (3.0). These form factor details mean power delivery and cooling are designed around the host system's module slot rather than a standalone bracket with attached fans. There are no length, height, or width figures for this module in the data, so the physical dimensions available to a cooling solution are not specified.

With only 18 W of TDP, the thermal load is low. The absence of a power connector field suggests the module does not require a separate external power feed, but the pack does not explicitly describe the connector situation beyond the null field. The absence of a suggested PSU figure likewise leaves system integration requirements unquantified. What the data does show is a low-power module whose thermal management belongs to the all-in-one chassis.

How It Compares

The nearestRivals field for this card is empty. There are no rival names, no rival scores, and no deltaPct values in the record. As a result, this entry does not support per-rival comparison paragraphs; there are no named competitors to position against.

The only comparative metric present is the percentile versus all GPUs of 50. A percentile of 50 places the card at the exact midpoint of the database's GPU distribution. That is a positional statement, not a measured speed difference, and it cannot be expanded into percentages of advantage or deficit relative to another product.

The silicon details in the record—the Redwood chip, TeraScale 2 architecture, 40 nm process, 627 million transistors, 104 mm² die size, and 6.0M/mm² transistor density—provide context for the module's class. Those figures describe manufacturing scale and architectural generation, but none of them is a benchmark score. Without rival entries or benchmark submissions, the HD 7470A stands as a mid-percentile, low-power MXM module in the database, with its specification sheet providing the only concrete basis for assessment.