AMD Radeon HD 8690M vs AMD Radeon HD 8790M Comparison

The AMD Radeon HD 8690M and AMD Radeon HD 8790M are two closely related mobile GPUs from the same GCN 1.0 era, and the recorded data crowns an unexpected winner: the nominally lower-positioned HD 8690M. In the one head-to-head benchmark in the database, Geekbench OpenCL, the HD 8690M scores 6137 against 5017 for the HD 8790M, a commanding 22.3 percent margin. Yet the raw specification sheet tells a more complicated story, with the HD 8790M holding more shading units, more texture units, and double the memory bandwidth. This analysis breaks down where each card wins, why the benchmark result diverges so sharply from the spec sheet, and which buyer each one suits.

Where Each One Wins

The HD 8690M wins where it matters most in the recorded data: actual measured performance. Its Geekbench OpenCL score of 6137 is 22.3 percent higher than the HD 8790M's 5017, and its overall standing in the database reflects that gap. The HD 8690M sits in the 36th percentile versus all GPUs, while the HD 8790M lands in the 33rd percentile. The average benchmark scores tell the same story, 6137 for the HD 8690M versus 5691 for the HD 8790M, though the latter figure blends in a Vulkan result that has no direct counterpart for the HD 8690M.

The HD 8790M wins on paper. It carries 384 shading units to the HD 8690M's 320, 24 texture units to 20, and a 128-bit memory bus delivering 64.00 GB/s of bandwidth, exactly double the HD 8690M's 32.00 GB/s over its 64-bit bus. Its texture rate of 21.60 GTexel/s beats the HD 8690M's 19.50 GTexel/s, and its FP32 throughput of 691.2 GFLOPS exceeds the HD 8690M's 624.0 GFLOPS. In workloads that are bandwidth-hungry or texture-bound, these advantages are real and could matter. The catch is that the measured OpenCL result does not reward them, at least not in this database.

Pixel throughput is the one spec category the HD 8690M wins outright: 7.800 GPixel/s against 7.200 GPixel/s for the HD 8790M, a direct consequence of its higher clock speeds. Both cards share the same 8 render output units, so clock frequency is the differentiator there, and the HD 8690M's 925 MHz base and 975 MHz boost clearly outpace the HD 8790M's 850 MHz base and 900 MHz boost.

FAQ

Q: Which GPU is faster overall?

A: The AMD Radeon HD 8690M. It won the only head-to-head benchmark in the database, Geekbench OpenCL, by 22.3 percent (6137 versus 5017), and it holds a higher percentile ranking against all GPUs, the 36th percentile versus the 33rd.

Q: Does the HD 8790M's wider memory bus make it faster?

A: Not according to the recorded data. Despite a 128-bit bus and 64.00 GB/s of bandwidth, double the HD 8690M's 32.00 GB/s, the HD 8790M lost the Geekbench OpenCL test decisively. The wider bus remains an asset in bandwidth-sensitive workloads, but the measurements do not show it translating into a win here.

Q: How do the two compare against rival GPUs?

A: The HD 8690M's closest rivals include the Intel Iris Pro Graphics 6200 (6117 average score, within 0.3 percent), the NVIDIA RTX A400 (6078, 1 percent), the NVIDIA GeForce MX230 (6077, 1 percent), and the AMD Radeon HD 6950 (6210, 1.2 percent ahead). The HD 8790M competes with the Intel Iris Pro Graphics P6300 (5712, 0.4 percent ahead), the NVIDIA GeForce GTX 670MX (5721, 0.5 percent ahead), the NVIDIA GeForce GTX 550 Ti (5731, 0.7 percent ahead), and the NVIDIA Quadro M500M (5604, 1.6 percent behind).

Q: Are both GPUs still in production?

A: No. Both the HD 8690M and the HD 8790M are listed as end-of-life.

Q: Do the two cards support the same API feature set?

A: Yes. Both support DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170.

Q: What memory configuration does each card have?

A: Both ship with 2 GB of GDDR5 running at 1000 MHz, 4 Gbps effective. The difference is bus width: 64-bit on the HD 8690M and 128-bit on the HD 8790M.

Head-to-Head Benchmarks

The database contains one direct head-to-head result, and it is not close. In Geekbench OpenCL, the AMD Radeon HD 8690M scored 6137 while the AMD Radeon HD 8790M managed 5017, a 22.3 percent victory for the HD 8690M. That is a large gap between two GPUs positioned in adjacent tiers of the same product generation, and it inverts the expectation created by the spec sheet.

Context from the rivals list sharpens the picture. The HD 8690M's 6137 OpenCL score places it statistically neck-and-neck with the Intel Iris Pro Graphics 6200, which averages 6117 and sits just 0.3 percent behind. It also edges the NVIDIA RTX A400 (6078) and the NVIDIA GeForce MX230 (6077) by about 1 percent each, and trails only the AMD Radeon HD 6950 among its nearest rivals, which averages 6210, 1.2 percent ahead.

The HD 8790M's 5017 OpenCL score, by contrast, falls well below its own rival cluster. Its nearest competitors average between 5604 and 5731: the Quadro M500M at 5604 sits 1.6 percent behind, while the Iris Pro P6300, the GTX 670MX, and the GTX 550 Ti all sit marginally ahead at 0.4, 0.5, and 0.7 percent respectively. In other words, the HD 8790M's OpenCL result underperforms even its modest peer group, whereas the HD 8690M's result holds up well against its own.

The HD 8790M does have a second recorded benchmark, Geekbench Vulkan, where it scored 6365. There is no comparable HD 8690M Vulkan entry in the database, so no direct comparison can be drawn, but the figure shows the chip is not uniformly weak across API backends. Its average benchmark score of 5691 across both tests reflects that mixed picture.

Specification Differences

The two cards differ in nearly every performance-relevant specification despite belonging to the same architecture.

Clock speeds favor the HD 8690M: 925 MHz base and 975 MHz boost, against 850 MHz base and 900 MHz boost for the HD 8790M. Memory clocks are identical at 1000 MHz (4 Gbps effective), but the bus differs sharply: 64-bit on the HD 8690M for 32.00 GB/s of bandwidth, versus 128-bit on the HD 8790M for 64.00 GB/s.

Shader resources favor the HD 8790M: 384 shading units and 24 texture units, against 320 shading units and 20 texture units on the HD 8690M. Both have 8 ROPs. The throughput figures split accordingly: the HD 8690M leads in pixel rate (7.800 GPixel/s versus 7.200 GPixel/s) because of its clock advantage, while the HD 8790M leads in texture rate (21.60 GTexel/s versus 19.50 GTexel/s) and FP32 compute (691.2 GFLOPS versus 624.0 GFLOPS) on the strength of its larger shader array.

The physical and platform specs also diverge. The HD 8690M is a smaller chip: 56 mm² of die area and 690 million transistors, versus 77 mm² and 950 million transistors for the HD 8790M. Both share the same transistor density of 12.3M per mm². Interface-wise, the HD 8690M connects over PCIe 3.0 x8, while the HD 8790M is an MXM Module using the MXM-A (3.0) interface with no power connectors and portable-device-dependent display outputs. Both carry 2 GB of GDDR5, both are end-of-life, and neither has a recorded launch MSRP.

Architecture Differences

Architecturally, these are siblings with different budgets. Both are built on AMD's GCN 1.0 architecture, both were fabricated by TSMC on its 28 nm process, and both belong to the Solar System generation of mobile parts, the HD 8690M within the HD 8600M stack and the HD 8790M within the HD 8700M stack. Both trace their lineage to the same predecessor, London, and both point to the same successor, Gem System.

The internal silicon tells the story of two distinct chips. The HD 8690M uses the Sun chip, a compact 56 mm² die with 690 million transistors. The HD 8790M uses the larger Mars chip at 77 mm² and 950 million transistors, spending that transistor budget on additional GCN compute units: 384 shading units versus 320. The shared 12.3M per mm² density confirms both came from the same process generation, with Mars simply being a scaled-up implementation of the same design philosophy.

Feature support is identical. Both expose DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170, and neither carries RT cores or tensor cores, as expected for GCN 1.0 hardware. Release timing was nearly simultaneous: the HD 8690M arrived on 2013-02-28, with the HD 8790M following on 2013-03-31.

The interesting architectural tension is clock versus resources. Sun runs its smaller array faster and pairs it with a narrower bus; Mars runs a bigger array slower and feeds it through a bus twice as wide. In the recorded OpenCL test, the faster, smaller configuration won decisively.

The Verdict

Based strictly on the measured data, the AMD Radeon HD 8690M is the pick. It won the only head-to-head benchmark by 22.3 percent, ranks three percentile points higher against the full GPU database (36th versus 33rd), and its OpenCL score holds its own against a rival cluster that includes the Intel Iris Pro Graphics 6200 and the NVIDIA GeForce MX230. For any workload resembling that OpenCL test, the data says the HD 8690M delivers more measured performance despite its smaller die and narrower bus.

The HD 8790M still has a defensible case, but only on the specification sheet. Buyers whose workloads lean on memory bandwidth or texture throughput may prefer its 64.00 GB/s bus, 21.60 GTexel/s texture rate, and 691.2 GFLOPS of FP32 compute, all of which exceed the HD 8690M's figures. Its Vulkan result of 6365 also suggests API-dependent behavior that the single OpenCL matchup does not capture. It is the only one of the two recorded in that test, so the comparison remains incomplete there.

Both cards are end-of-life, identically featured at the API level, and equipped with the same 2 GB of GDDR5. The database verdict is straightforward: on measured performance, the HD 8690M wins, and by a wide margin; on theoretical resources, the HD 8790M is the larger, wider, better-fed chip whose potential the recorded benchmarks did not confirm.