AMD Radeon HD 8690M

The AMD Radeon HD 8690M is a GCN 1.0 GPU built on TSMC's 28 nm process. Based on the Sun chip and the Solar System (HD 8600M) generation, it sits between the London predecessor and the Gem System successor. It contains 690 million transistors on a 56 mm² die, for a transistor density of 12.3M / mm². The GPU comes with 2 GB of GDDR5 memory on a 64-bit bus and reports a Geekbench OpenCL score of 6114, placing it at the 34th percentile of all GPUs in the database. Released on February 28, 2013, it is listed as End-of-life.

How It Compares

The closest rival in the database is Intel Iris Pro Graphics 6200, which averages 6163 on the same OpenCL test. The HD 8690M trails that result by 0.8%, the smallest margin among its listed nearest rivals.

Against NVIDIA GeForce GTX 770M, the data shows a 1.5% advantage for the Radeon. The GTX 770M averages 6026, so the HD 8690M is the higher-scoring part in that pair, although the delta is narrow.

The AMD FirePro W600 averages 6205, which is 1.5% higher than the HD 8690M. This places the AMD parts at essentially the same level in this benchmark, with the FirePro holding a marginal edge.

The highest-scoring rival is AMD Radeon HD 6950, with an average of 6210. The HD 8690M is 1.5% behind that figure. Across the listed nearest rivals, the largest delta is 1.5%, and the HD 8690M sits in a tightly clustered performance band.

Power and Cooling

The record for the Radeon HD 8690M contains no TDP value, no suggested PSU rating, and no power connector specification. That means no wattage target or power-supply size can be quoted from the database. The only bus-related requirement listed is PCIe 3.0 x8, which defines the interface between the GPU and the rest of the system.

For thermal context, the database gives a TSMC 28 nm process, a die size of 56 mm², and a transistor count of 690 million. The architecture is GCN 1.0, and the core clocks are listed at 925 MHz base and 975 MHz boost. The memory runs at 1000 MHz with 4 Gbps effective GDDR5 signaling across a 64-bit bus. These are the only figures relevant to the power envelope; the absence of a TDP means any further wattage estimate would not be supported by the data.

No slot width, cooler dimensions, or display outputs are included in the record, so physical mounting and connector guidance is also unavailable. The production status is End-of-life, and the power-related fields are all empty. Without a suggested PSU or connector data, the database cannot provide a power-supply selection for this GPU.

Benchmark Performance

The only benchmark result is Geekbench OpenCL. The HD 8690M scores 6114, which is also its average benchmark score. The all-GPU percentile rank is 34, placing it below the majority of tested GPUs.

Relative to its nearest rivals, the HD 8690M is 0.8% behind the Intel Iris Pro Graphics 6200 (6163), 1.5% ahead of the NVIDIA GeForce GTX 770M (6026), and 1.5% behind both the AMD FirePro W600 (6205) and the AMD Radeon HD 6950 (6210). The largest difference in either direction is therefore 1.5%. This means the HD 8690M's benchmark position is not defined by a large advantage or disadvantage; it is clustered with the listed rivals at essentially the same performance level.

The GPU's peak FP32 compute rate is 624.0 GFLOPS. Pixel fill rate is 7.800 GPixel/s, and texture fill rate is 19.50 GTexel/s. Core clocks are 925 MHz base and 975 MHz boost. The memory system provides 32.00 GB/s of bandwidth from 2 GB of GDDR5 on a 64-bit bus, with a 1000 MHz memory clock and 4 Gbps effective data rate. For OpenCL workloads that depend heavily on memory bandwidth, the 32.00 GB/s figure and the 64-bit bus are the primary constraints.

Because the dataset contains no game benchmarks, all performance conclusions are limited to the Geekbench OpenCL score. That score places the HD 8690M at the 34th percentile, which is modest for the database as a whole, but the tight 1.5% margins around its nearest rivals show that it is competitive within its immediate class.

Who Should Consider It

The HD 8690M is an End-of-life GPU with a 34th-percentile OpenCL score and a 2 GB GDDR5 memory configuration. Users should consider it for workloads that fit a modest compute and memory envelope: light OpenCL tasks, legacy system use, or applications where the database's nearest-rival margins of 1.5% are irrelevant. It is not a data point for high-fidelity or high-bandwidth rendering, because the texture rate of 19.50 GTexel/s, pixel rate of 7.800 GPixel/s, and 32.00 GB/s memory bandwidth are all on the low end.

At the settings level, the fill-rate figures argue for lower-detail configurations and smaller render targets. The 2 GB frame buffer is large enough for a practical amount of data, but the 64-bit bus and 32.00 GB/s bandwidth will limit how quickly textures and geometry can move. The FP32 rate of 624.0 GFLOPS gives a clear compute ceiling for OpenCL tasks; before choosing this GPU, a user should confirm that the target workload stays within that ceiling.

Among the nearest rivals, the HD 8690M is essentially interchangeable on raw score. It trails the AMD FirePro W600 and AMD Radeon HD 6950 by 1.5% and leads the NVIDIA GeForce GTX 770M by 1.5%. A user whose platform already matches a listed rival would not see a meaningful benchmark difference from switching to the HD 8690M. The decision should therefore be based on compatibility and availability, not on the 1.5% deltas. The End-of-life status means the part is no longer in production, so availability may be limited to existing inventory.

Ray Tracing and Feature Set

The HD 8690M has no RT core count and no tensor core count in the database. That means hardware-accelerated ray tracing and tensor processing are not indicated in this product record. The feature set is based on GCN 1.0, with 320 shading units, 20 texture units, and 8 ROPs. The PCIe interface is PCIe 3.0 x8.

API support is more modern than the GPU's release date might suggest: DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170 are all listed. Vulkan 1.2.170 and OpenGL 4.6 are the main software-facing features. The DirectX 12 entry is specifically listed as 12 (11_1), which records the feature level as 11_1 rather than a higher feature level.

No FP16 performance is listed; only FP32 at 624.0 GFLOPS is provided. The memory system is the same 2 GB GDDR5 configuration on a 64-bit bus, with 32.00 GB/s bandwidth. The record does not include display outputs, so the number or type of video connections cannot be stated. For users who need modern API entry points and can work within the limits of GCN 1.0, the HD 8690M offers a coherent feature set; for ray tracing or tensor workloads, it does not.