ATI Mobility Radeon HD 5670 Mac Edition

The ATI Mobility Radeon HD 5670 Mac Edition is a mobile GPU from AMD built around the Madison chip and the TeraScale 2 architecture. TSMC fabricated the die on a 40 nm process, integrating 627 million transistors into a 104 mm² package with a transistor density of 6.0M / mm². The product was released on 2010-04-21 and is now listed as end-of-life. It connects through PCIe 2.0 x16, and its display outputs are described as Portable Device Dependent. The generation field is Manhattan (Mobility HD 5600). In the FACT PACK, the benchmarks array is empty, the average benchmark score is 0, and percentileVsAllGpus is 50.

Memory Subsystem

The memory subsystem consists of 512 MB of GDDR3 on a 128-bit bus. The memory clock is 795 MHz, with an effective data rate of 1590 Mbps, producing 25.44 GB/s of bandwidth. That is the complete memory picture in the FACT PACK: no other capacity, bus width, or bandwidth variant is listed.

For high-resolution workloads, 512 MB is the most immediate constraint. A frame buffer of this size can hold only a limited amount of geometry, texture data, and render targets. As resolution rises, the amount of data stored per frame rises as well, and 512 MB can force lower texture quality or reduced detail levels. The bandwidth figure reinforces that limitation. At 25.44 GB/s over a 128-bit path, the memory controller cannot move data as quickly as wider or faster configurations. The 795 MHz memory clock is not high, so the effective throughput of 1590 Mbps is modest by the standards of later parts in the database.

The practical implication is that this GPU is better suited to lower resolutions and conservative settings. High-resolution anti-aliasing, high-quality shadows, and large texture packs would put pressure on both the capacity and the bandwidth of this subsystem. The 128-bit bus width means each memory access cycle transfers a relatively narrow chunk of data, so the bandwidth ceiling of 25.44 GB/s becomes a limiting factor when pixel counts and texture fetches increase. The data points to a memory design intended for moderate workloads rather than extreme high-resolution rendering.

Ray Tracing and Feature Set

The TeraScale 2 architecture is the foundation of this GPU, and the FACT PACK lists no RT core count and no tensor core count. Hardware-accelerated ray tracing and tensor-based AI features are therefore not part of the specification. The feature set is instead built from conventional shader and fixed-function resources: 400 shading units, 20 texture mapping units, and 8 ROPs. These combine into a pixel rate of 5.200 GPixel/s and a texture rate of 13.00 GTexel/s.

API support in the FACT PACK is DirectX 11.2 (11_0) and OpenGL 4.4. Vulkan is not listed. That means applications requiring Vulkan support are outside the documented feature set. The DirectX 11.2 (11_0) and OpenGL 4.4 support provide a baseline for software from the product's release era, but there is no indication of newer API capabilities. Because no RT or tensor cores are present, any ray-traced effects would have to be executed through the 400 shading units and the fixed-function pipeline, which is inefficient for that style of rendering. The data paints the picture of a conventional rasterization-oriented GPU.

Benchmark Performance

The benchmarks array in the FACT PACK is empty. No individual benchmark results are recorded. The average benchmark score is listed as 0, which is not a measured performance figure but a placeholder value. The percentileVsAllGpus entry of 50 places the GPU at the midpoint of the database's all-GPU ranking, even though no benchmark entries back that position.

The only quantitative performance indicators are the compute and fill-rate figures. FP32 throughput is listed at 520.0 GFLOPS. Pixel rate is 5.200 GPixel/s, and texture rate is 13.00 GTexel/s. These numbers are tied to the 400 shading units, 8 ROPs, and 20 TMUs respectively. The FP32 throughput of 520.0 GFLOPS suggests a low-to-mid-range execution envelope. The pixel rate of 5.200 GPixel/s indicates that the 8 ROPs can only output a limited number of pixels per second, which becomes more relevant at higher resolutions. The texture rate of 13.00 GTexel/s, backed by 20 TMUs, places a ceiling on how quickly texture data can be filtered.

Because no nearest rivals are listed, no percentage deltas can be computed. There are no scores to compare, no rival names to cite, and no deltaPct values to analyze. The 50th percentile placement is a broad database position, not a head-to-head performance ranking. The empty benchmarks list and zero average score mean the performance story rests entirely on the architectural and fill-rate data.

How It Compares

The nearestRivals field is empty. There are no named competitors in the FACT PACK, no rival scores, and no deltaPct values to calculate. This GPU cannot be positioned against specific alternatives such as other mobile GPUs from the same generation. The only placement information is the percentileVsAllGpus value of 50, which is a general database rank rather than a comparison to any identified product.

The absence of nearest rivals means statements such as "ahead of" or "behind" cannot be made from the data. No comparative tier can be constructed, and no ordering list can be assembled. What remains is the product's own specification: 400 shading units, 20 TMUs, 8 ROPs, 512 MB of GDDR3, a 128-bit bus, and 25.44 GB/s of bandwidth. Those attributes, combined with the 15 W TDP, define the GPU's envelope. The empty rival list is itself informative: it suggests the database has no closely matched GPU to place beside this Mac Edition part, or that no comparative evaluation was performed.

FAQ

Q: What architecture and chip does this GPU use?

A: It uses the TeraScale 2 architecture with the Madison chip, fabricated at TSMC on a 40 nm process with 627 million transistors and a 104 mm² die.

Q: How much memory does it have?

A: It has 512 MB of GDDR3 on a 128-bit bus, with a 795 MHz memory clock, a 1590 Mbps effective data rate, and 25.44 GB/s of bandwidth.

Q: Does it support ray tracing or Vulkan?

A: The FACT PACK lists no RT core count and no tensor core count. Vulkan is not listed. API support is DirectX 11.2 (11_0) and OpenGL 4.4.

Q: What is the power draw?

A: The listed TDP is 15 W. No power connector, suggested PSU, or slot width is recorded.

Q: When was it released?

A: It was released on 2010-04-21. Its predecessor is M9x, and its successor is Vancouver.

Q: What is its benchmark standing?

A: The benchmarks array is empty, the average benchmark score is 0, and percentileVsAllGpus is 50.

Power and Cooling

The TDP is 15 W, which is the only power figure in the FACT PACK. This low value points toward modest thermal output, but the data provides no cooling details. No power connector is listed, no suggested PSU is listed, and no slot width is recorded. The bus interface is PCIe 2.0 x16, and display outputs are Portable Device Dependent, meaning the host portable device determines the physical output configuration.

Because no power connector is listed, an auxiliary power connection should not be assumed. The absence of a suggested PSU also indicates that this is not a desktop-style upgrade component with a dedicated power recommendation. Instead, the power delivery and thermal management are part of the surrounding system. The 15 W TDP is the design target, and it suggests that cooling can be modest by desktop standards, but the exact cooler or heatsink design is not recorded. For a Mac Edition mobile part, the practical power and cooling behavior depends on the host device's construction, which is outside the scope of the FACT PACK.

Who Should Consider It

This GPU is for users who need a Mac Edition mobile part with a 15 W TDP, TeraScale 2 architecture, and DirectX 11.2 (11_0) and OpenGL 4.4 support. The 512 MB GDDR3 frame buffer and 25.44 GB/s bandwidth argue against high-resolution, high-detail workloads. At lower resolutions, the 400 shading units, 20 TMUs, and 8 ROPs are a more reasonable match for the memory subsystem. The FP32 throughput of 520.0 GFLOPS and pixel rate of 5.200 GPixel/s suggest that simple scenes and moderate settings are the realistic target.

Users who require Vulkan support or hardware ray tracing should look elsewhere, as the FACT PACK lists neither. The end-of-life production status means the product is no longer in active production, and the 50th percentile database position offers no measured performance assurance. The Portable Device Dependent display output wording means compatibility is tied to the specific Mac portable system. In short, the intended use case is a legacy Mac portable environment where the 15 W power envelope and the documented API support matter more than raw speed or modern feature acceleration.