AMD Radeon HD 8750M vs AMD Radeon HD 8790M Comparison

The AMD Radeon HD 8750M and AMD Radeon HD 8790M are two mobile graphics solutions from the same Solar System (HD 8700M) generation, both built on the GCN 1.0 architecture and fabricated on TSMC's 28 nm process. Despite sharing the same Mars chip, 950 million transistors, and 77 mm² die size, benchmark results reveal a significant performance divergence that is not immediately obvious from their architectural similarity. The data shows that the HD 8750M holds a decisive lead in the single available OpenCL benchmark, while the HD 8790M counters with a superior memory subsystem and higher clock frequencies that do not translate into a benchmark victory in the recorded data.

Head-to-Head Benchmarks

The only direct benchmark comparison available is the Geekbench OpenCL test, and the results are stark. The AMD Radeon HD 8750M scores 5970 points, while the AMD Radeon HD 8790M scores 5017 points. This represents a 19% delta in favor of the HD 8750M, a substantial margin for two GPUs that share the same core configuration. The HD 8750M's victory is notable because the HD 8790M operates at higher base and boost clocks, yet the benchmark outcome favors the lower-clocked part.

The HD 8750M's OpenCL score places it at the 34th percentile among all GPUs, while the HD 8790M sits at the 33rd percentile. This percentile difference, while narrow, aligns with the benchmark delta. The HD 8750M's average benchmark score of 5970 is derived from that single OpenCL result, whereas the HD 8790M's average of 5691 is pulled down by its OpenCL score of 5017 despite also having a Vulkan score of 6365. The Vulkan result for the HD 8790M demonstrates that its hardware is capable of higher performance under different API workloads, but it does not appear in the head-to-head comparison.

Examining the nearest rivals provides additional context. The HD 8750M's closest competitor is the NVIDIA Quadro K4000, which averages 5982 points, a -0.2% delta from the HD 8750M. The HD 8750M also edges out the NVIDIA Quadro K620M (5957, 0.2%), the AMD Radeon HD 8730M (5955, 0.3%), and the NVIDIA Quadro K4000M (5986, -0.3%). For the HD 8790M, its nearest rival is the Intel Iris Pro Graphics P6300 at 5712 points (-0.4%), followed by the NVIDIA GeForce GTX 670MX (5721, -0.5%), the NVIDIA GeForce GTX 550 Ti (5731, -0.7%), and the NVIDIA Quadro M500M (5604, 1.6%). These rival comparisons show that while the HD 8750M competes within a tight cluster around 5950-5990 points, the HD 8790M's average is positioned against a different set of competitors, reflecting its lower average score.

Where Each One Wins

The benchmark data presents a clear dichotomy. The AMD Radeon HD 8750M wins the only head-to-head benchmark, Geekbench OpenCL, by 19%. This suggests that for compute-oriented OpenCL workloads, the HD 8750M is the superior choice. Its single benchmark result of 5970 is consistent across its entire testing profile, indicating reliable performance in that specific domain.

The AMD Radeon HD 8790M, despite losing the OpenCL comparison, demonstrates strength in other areas. Its Geekbench Vulkan score of 6365 is notably higher than its OpenCL score, indicating that under Vulkan API workloads, it can substantially outperform its own OpenCL capabilities. This Vulkan performance is not captured in the head-to-head comparison but represents a genuine advantage for applications that leverage Vulkan. The HD 8790M also benefits from a 2 GB memory allocation compared to the HD 8750M's 1024 MB, which is advantageous for texture-heavy workloads and larger datasets, even if it does not directly influence the recorded benchmarks.

For use-case splits, the HD 8750M is the clear winner for OpenCL compute tasks, scientific computing, and any application that relies on OpenCL acceleration. The HD 8790M, conversely, is positioned better for Vulkan-based gaming or rendering workloads, where its higher clock speeds and faster GDDR5 memory can be utilized. The HD 8790M's higher base clock of 850 MHz and boost clock of 900 MHz (versus 775 MHz and 825 MHz on the HD 8750M) provide a theoretical advantage in clock-bound scenarios, but the benchmark data does not reflect this in OpenCL.

Architecture Differences

Both GPUs are built on the GCN 1.0 architecture, use the Mars chip, and are manufactured on TSMC's 28 nm process with 950 million transistors on a 77 mm² die. The transistor density is identical at 12.3M / mm². The core configuration is also the same: 384 shading units, 24 texture mapping units, and 8 render output units. Neither GPU features ray tracing cores or tensor cores, as these are not listed in the specifications.

The architectural differences manifest primarily in clock speeds and memory configuration, rather than in core design. The HD 8750M operates at a base clock of 775 MHz with a boost of 825 MHz, while the HD 8790M runs at 850 MHz base and 900 MHz boost. This 75 MHz difference in both base and boost clocks is a direct result of binning and power characteristics, not architectural changes.

Memory architecture is where the two diverge most significantly. The HD 8750M uses DDR3 memory with a clock of 900 MHz and an effective data rate of 1800 Mbps, resulting in a bandwidth of 28.80 GB/s. The HD 8790M employs GDDR5 memory at 1000 MHz with an effective rate of 4 Gbps, delivering 64.00 GB/s bandwidth. Both maintain a 128 bit bus width, but the GDDR5 memory on the HD 8790M more than doubles the bandwidth. This difference is crucial for memory-bound workloads, though it does not appear to benefit the HD 8790M in the OpenCL benchmark.

The pixel rate and texture rate also differ due to clock speeds. The HD 8750M achieves 6.600 GPixel/s and 19.80 GTexel/s, while the HD 8790M reaches 7.200 GPixel/s and 21.60 GTexel/s. The FP32 performance follows the same pattern: 633.6 GFLOPS for the HD 8750M versus 691.2 GFLOPS for the HD 8790M. These figures represent a roughly 9% theoretical advantage for the HD 8790M, yet the benchmark shows the opposite result.

Specification Differences

The two GPUs differ in several key specification fields. The memory size differs: the HD 8750M has 1024 MB, while the HD 8790M has 2 GB. Memory type also differs, with the HD 8750M using DDR3 and the HD 8790M using GDDR5. The memory clock is 900 MHz for the HD 8750M versus 1000 MHz for the HD 8790M, and the effective memory speed is 1800 Mbps versus 4 Gbps. Consequently, memory bandwidth is 28.80 GB/s for the HD 8750M and 64.00 GB/s for the HD 8790M.

Clock speeds differ across the board: base clock is 775 MHz versus 850 MHz, and boost clock is 825 MHz versus 900 MHz. The pixel rate is 6.600 GPixel/s versus 7.200 GPixel/s, and the texture rate is 19.80 GTexel/s versus 21.60 GTexel/s. FP32 performance is 633.6 GFLOPS versus 691.2 GFLOPS.

The bus interface also differs. The HD 8750M uses PCIe 3.0 x8, while the HD 8790M uses MXM-A (3.0). The HD 8790M is specified as an MXM Module with None for power connectors and Portable Device Dependent for display outputs, while the HD 8750M does not list these fields. The release dates differ by over a month: the HD 8750M was released on 2013-02-25 and the HD 8790M on 2013-03-31. Both share the same production status (End-of-life), predecessor (London), successor (Gem System), and API support (DirectX 12 (11_1), OpenGL 4.6, Vulkan 1.2.170).

FAQ

Q: Which GPU has the higher benchmark score in the head-to-head comparison?

A: The AMD Radeon HD 8750M wins the Geekbench OpenCL comparison with a score of 5970 versus 5017 for the HD 8790M, a 19% delta.

Q: Does the HD 8790M have any benchmark advantage over the HD 8750M?

A: In the recorded head-to-head data, the HD 8790M does not win any benchmark. However, it has a Geekbench Vulkan score of 6365, which is higher than its OpenCL score, though no Vulkan score is listed for the HD 8750M for comparison.

Q: What are the memory differences between the two GPUs?

A: The HD 8750M has 1024 MB of DDR3 memory with 28.80 GB/s bandwidth, while the HD 8790M has 2 GB of GDDR5 memory with 64.00 GB/s bandwidth. Both use a 128 bit bus.

Q: How do the clock speeds compare?

A: The HD 8750M runs at 775 MHz base and 825 MHz boost, while the HD 8790M runs at 850 MHz base and 900 MHz boost, giving the HD 8790M a 75 MHz advantage at both clock states.

Q: Are the core configurations identical?

A: Yes, both GPUs have 384 shading units, 24 TMUs, and 8 ROPs, and both use the Mars chip on a 28 nm process with 950 million transistors.

Q: What is the average benchmark score for each GPU?

A: The HD 8750M has an average benchmark score of 5970, while the HD 8790M has an average of 5691, reflecting the HD 8750M's higher OpenCL result.