AMD Radeon HD 8690M vs AMD Radeon HD 8850M Comparison

FAQ

Q: How does the AMD Radeon HD 8850M compare to the AMD Radeon HD 8690M in raw compute performance?

A: The recorded Geekbench OpenCL score shows the HD 8850M at 7447 points versus 6137 points for the HD 8690M. This translates to a 21.3% advantage for the HD 8850M in the head-to-head measurement.

Q: Which GPU has a higher memory bandwidth, and what is the exact difference?

A: The HD 8850M has a 128-bit memory bus and delivers 64.00 GB/s, while the HD 8690M uses a 64-bit bus with 32.00 GB/s. The HD 8850M offers double the memory bandwidth.

Q: What is the transistor count difference between these two chips?

A: The HD 8850M, based on the Venus chip, integrates 1,500 million transistors on a 123 mm² die. The HD 8690M, using the Sun chip, has 690 million transistors on a 56 mm² die. The HD 8850M has more than twice the transistor count.

Q: Do both GPUs support the same graphics APIs?

A: Yes, both support DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170. The API feature set is identical between the two models.

Q: Which GPU has more shading units, texture mapping units, and ROPs?

A: The HD 8850M has 640 shading units, 40 TMUs, and 16 ROPs. The HD 8690M has 320 shading units, 20 TMUs, and 8 ROPs. The HD 8850M exactly doubles the HD 8690M in each of these categories.

Q: How do the two GPUs rank in the overall database percentile?

A: The HD 8850M sits at the 40th percentile among all GPUs, while the HD 8690M ranks at the 36th percentile. The HD 8850M is positioned higher in the overall performance distribution.

Where Each One Wins

Based solely on the recorded benchmark data, the AMD Radeon HD 8850M wins the only available test (Geekbench OpenCL) with a score of 7447 versus 6137, a 21.3% margin. The HD 8690M does not win any recorded benchmark test in this comparison.

For workloads that stress memory bandwidth, the HD 8850M’s 64.00 GB/s versus 32.00 GB/s gives it a clear edge in data-heavy compute tasks. The doubling of the memory bus width from 64-bit to 128-bit is a decisive structural advantage for the HD 8850M.

For tasks that rely on raw shading throughput, the HD 8850M’s 640 shading units versus 320, along with a 25.00 GTexel/s texture rate versus 19.50 GTexel/s, makes it the stronger candidate for graphics-bound workloads. The pixel rate also favors the HD 8850M at 10.00 GPixel/s versus 7.800 GPixel/s.

The HD 8690M does have a higher base and boost clock (925 MHz and 975 MHz versus 575 MHz and 625 MHz), but the HD 8850M compensates with more execution resources and a wider memory interface. The data shows no scenario in which the HD 8690M outperforms the HD 8850M based on the recorded benchmarks.

Architecture Differences

Both GPUs are built on GCN 1.0 architecture and manufactured by TSMC on a 28 nm process node. The core designs, however, differ substantially in scale.

The HD 8850M uses the Venus chip, which is a larger implementation with 1,500 million transistors on a 123 mm² die. The HD 8690M uses the Sun chip, which has 690 million transistors on a 56 mm² die. Despite the size difference, the transistor density is nearly identical: 12.2M / mm² for Venus and 12.3M / mm² for Sun. This indicates that the manufacturing process is the same, but the HD 8850M simply has a much larger silicon area devoted to compute resources.

The HD 8850M has exactly double the shading units (640 versus 320), double the texture mapping units (40 versus 20), and double the raster operation units (16 versus 8). This structural doubling is consistent across the entire execution pipeline.

The memory subsystem also differs fundamentally. The HD 8850M uses a 128-bit memory bus, while the HD 8690M uses a 64-bit bus. Both use GDDR5 memory at the same effective data rate of 4 Gbps and the same 1000 MHz memory clock. The bandwidth difference (64.00 GB/s versus 32.00 GB/s) is entirely attributable to the bus width difference.

The bus interface also differs: the HD 8850M connects via PCIe 3.0 x16, while the HD 8690M uses PCIe 3.0 x8. This could affect data transfer efficiency with the host system, particularly in mixed workloads.

Both chips are designated end-of-life and belong to the Solar System generation, with the HD 8850M in the HD 8800M series and the HD 8690M in the HD 8600M series. Both share the same predecessor (London) and successor (Gem System) in the database.

Specification Differences

The following specification fields differ between the two GPUs:

• Chip: Venus (HD 8850M) versus Sun (HD 8690M)
• Generation: HD 8800M (HD 8850M) versus HD 8600M (HD 8690M)
• Transistors: 1,500 million versus 690 million
• Die Size: 123 mm² versus 56 mm²
• Transistor Density: 12.2M / mm² versus 12.3M / mm² (minor difference)
• Base Clock: 575 MHz versus 925 MHz
• Boost Clock: 625 MHz versus 975 MHz
• Memory Bus Width: 128 bit versus 64 bit
• Memory Bandwidth: 64.00 GB/s versus 32.00 GB/s
• Shading Units: 640 versus 320
• Texture Mapping Units: 40 versus 20
• Raster Operation Units: 16 versus 8
• Pixel Rate: 10.00 GPixel/s versus 7.800 GPixel/s
• Texture Rate: 25.00 GTexel/s versus 19.50 GTexel/s
• FP32 Performance: 800.0 GFLOPS versus 624.0 GFLOPS
• Bus Interface: PCIe 3.0 x16 versus PCIe 3.0 x8
• Release Date: 2013-03-31 versus 2013-02-28

Fields that are identical include memory size (2 GB), memory type (GDDR5), memory clock (1000 MHz, 4 Gbps effective), API support (DirectX 12 (11_1), OpenGL 4.6, Vulkan 1.2.170), TDP (not specified), and production status (end-of-life).

Head-to-Head Benchmarks

The only recorded head-to-head benchmark is Geekbench OpenCL, where the AMD Radeon HD 8850M scores 7447 and the AMD Radeon HD 8690M scores 6137. The HD 8850M wins by a 21.3% margin.

To put this in context, the HD 8850M’s nearest rivals in the database include the Intel UHD Graphics 750 at 7441 (only 0.1% behind), the AMD Radeon R7 350 at 7425 (0.3% behind), the NVIDIA GeForce GTX 1650 at 7472 (0.3% ahead), and the Intel Arc A310 at 7550 (1.4% ahead). The HD 8850M sits in a tight cluster where a 1.4% swing separates it from the next higher-ranked GPU.

The HD 8690M’s nearest rivals include the Intel Iris Pro Graphics 6200 at 6117 (0.3% behind), the NVIDIA RTX A400 at 6078 (1% behind), the NVIDIA GeForce MX230 at 6077 (1% behind), and the AMD Radeon HD 6950 at 6210 (1.2% ahead). The HD 8690M occupies a lower performance tier, roughly 18% below the HD 8850M’s cluster.

The 21.3% gap between the two AMD mobile GPUs is substantial. It aligns with the structural differences: the HD 8850M has double the shading units, double the TMUs, double the ROPs, and double the memory bandwidth. The HD 8690M’s higher clock speeds (925 MHz base and 975 MHz boost versus 575 MHz and 625 MHz) are not enough to overcome the resource deficit. Even with a 54% higher boost clock, the HD 8690M’s FP32 throughput of 624.0 GFLOPS trails the HD 8850M’s 800.0 GFLOPS by 22%, which closely matches the benchmark delta.

The pixel rate difference (10.00 GPixel/s versus 7.800 GPixel/s) and texture rate difference (25.00 GTexel/s versus 19.50 GTexel/s) point to the HD 8850M being better suited for fill-rate-limited scenarios. The memory bandwidth doubling is particularly relevant for workloads that stream large datasets, such as image processing or compute kernels that access textures frequently.

In the database, the HD 8850M achieves 1 win out of 1 head-to-head benchmark, while the HD 8690M records 0 wins. The HD 8850M’s 40th percentile ranking versus the HD 8690M’s 36th percentile ranking confirms that the performance gap is meaningful in the broader GPU landscape, though both sit in the lower half of the distribution. The data indicates that for any application relying on OpenCL compute, the HD 8850M is the superior choice by a consistent margin.