AMD Radeon HD 8790M vs AMD Radeon R7 M445 Comparison

The AMD Radeon HD 8790M and the AMD Radeon R7 M445 represent two distinct generations of mobile graphics, with the former built on GCN 1.0 and the latter on GCN 3.0. The benchmark data, derived from Geekbench OpenCL scores, reveals a narrow but decisive victory for the R7 M445, which posts an average score of 5358 against the HD 8790M’s 5017. This translates to a 6.4% performance advantage for the newer part, a margin that places both cards in the lower third of all GPUs, with the HD 8790M at the 33rd percentile and the R7 M445 at the 31st.

Head-to-Head Benchmarks

The only direct comparison available is the Geekbench OpenCL test, where the AMD Radeon R7 M445 wins outright. The R7 M445 scores 5358, while the AMD Radeon HD 8790M scores 5017. The deltaPct of -6.4% indicates that the HD 8790M trails the R7 M445 by that margin. In absolute terms, the R7 M445 is 341 points ahead, a difference that, while not enormous, is consistent across the single benchmark sample.

Examining the rivals for each card provides additional context. The HD 8790M’s average score of 5691 is within 0.4% of the Intel Iris Pro Graphics P6300 (5712) and 0.5% of the NVIDIA GeForce GTX 670MX (5721). It is also 0.7% ahead of the NVIDIA GeForce GTX 550 Ti (5731) and 1.6% ahead of the NVIDIA Quadro M500M (5604). These tight deltas suggest that the HD 8790M sits in a very crowded performance band, where a few points separate it from several competitors.

For the R7 M445, its average score of 5358 places it nearly level with the Intel UHD Graphics P630 (5370), which is 0.2% faster. The R7 M445 is 0.7% ahead of the NVIDIA GeForce 840M (5322), 0.8% ahead of the NVIDIA GeForce 930A (5317), and 0.9% ahead of the NVIDIA GeForce GTX 980M (5308). This last comparison is particularly striking: the R7 M445, a low-end mobile part, edges out the GTX 980M, a high-end laptop GPU, by less than one percent in this particular OpenCL workload. However, this should not be overinterpreted, as the benchmark measures compute performance rather than gaming rasterization.

The single head-to-head result shows the R7 M445’s win is not a fluke of clock speeds alone. The R7 M445 boosts to 920 MHz, only 20 MHz higher than the HD 8790M’s 900 MHz boost. Yet the R7 M445 achieves its higher score despite having fewer shading units (320 vs. 384) and a narrower memory bus (64-bit vs. 128-bit). The architectural leap from GCN 1.0 to GCN 3.0 appears to compensate for the raw resource deficit, as does the larger 4 GB memory allocation versus the HD 8790M’s 2 GB.

The Verdict

From the data, the AMD Radeon R7 M445 is the faster GPU in compute-oriented OpenCL tasks. Its 6.4% lead over the HD 8790M is the only direct benchmark comparison, and it wins that test. For users prioritizing raw compute throughput in applications that leverage OpenCL, the R7 M445 is the clear choice. The HD 8790M, however, is not far behind, and its higher pixel rate (7.200 GPixel/s vs. 7.360 GPixel/s for the R7 M445) suggests it holds its own in fill-rate-bound scenarios, though this is not reflected in the single benchmark provided.

The R7 M445 also benefits from a newer feature set, supporting DirectX 12 (12_0) compared to the HD 8790M’s DirectX 12 (11_1). This makes the R7 M445 more future-proof for titles that utilize DirectX 12’s advanced features. The HD 8790M, while older, still maintains parity in OpenGL 4.6 and Vulkan 1.2.170 support, so legacy API compatibility is equal.

Given the benchmark results, the R7 M445 is the superior part for general compute performance. The HD 8790M is not obsolete, but it is measurably slower. The verdict is straightforward: if the choice is between these two and the workload relies on OpenCL, pick the R7 M445.

Where Each One Wins

The AMD Radeon R7 M445 wins in the sole compute benchmark, the Geekbench OpenCL test. Its higher average score of 5358 against the HD 8790M’s 5017 indicates better raw throughput for tasks like video encoding, physics simulations, or any GPU-accelerated application that uses OpenCL. The R7 M445 also has double the memory capacity (4 GB vs. 2 GB), which can be an advantage in workloads that require larger textures or datasets, even if the memory bandwidth is halved (32.00 GB/s vs. 64.00 GB/s).

The AMD Radeon HD 8790M, despite losing the compute test, has strengths in other areas. Its texture rate of 21.60 GTexel/s is higher than the R7 M445’s 18.40 GTexel/s, suggesting better performance in texture-heavy scenes where the GPU must sample many pixels. Its FP32 throughput of 691.2 GFLOPS also exceeds the R7 M445’s 588.8 GFLOPS, which could matter in applications that are not fully optimized for the newer architecture’s instruction set. However, these advantages do not translate into a benchmark win in the data provided.

The HD 8790M’s memory bandwidth of 64.00 GB/s is exactly double the R7 M445’s 32.00 GB/s. This is a significant gap that could benefit the HD 8790M in memory-intensive tasks like high-resolution anti-aliasing or large framebuffer operations, even though its total memory is smaller. The bus width difference (128-bit vs. 64-bit) is the root cause, and it gives the older card a theoretical advantage that the R7 M445 cannot overcome in some scenarios.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The AMD Radeon R7 M445 has a higher average benchmark score of 5358, compared to the AMD Radeon HD 8790M’s 5691. Note that the HD 8790M’s average includes two tests (OpenCL and Vulkan), while the R7 M445’s average is based on a single OpenCL test.

Q: How much faster is the R7 M445 in the Geekbench OpenCL test?

A: The R7 M445 scores 5358 versus the HD 8790M’s 5017, giving the R7 M445 a 6.4% lead as indicated by the deltaPct of -6.4% (negative indicating the HD 8790M is behind).

Q: Does the HD 8790M have any performance advantage over the R7 M445?

A: Yes, in theoretical throughput metrics. The HD 8790M has a higher texture rate (21.60 GTexel/s vs. 18.40 GTexel/s) and higher FP32 compute (691.2 GFLOPS vs. 588.8 GFLOPS), though these do not translate to a win in the provided benchmark.

Q: Which GPU supports a newer version of DirectX?

A: The AMD Radeon R7 M445 supports DirectX 12 (12_0), while the AMD Radeon HD 8790M supports DirectX 12 (11_1). This gives the R7 M445 access to more advanced DirectX 12 features.

Q: What is the memory configuration difference?

A: The HD 8790M has 2 GB of GDDR5 on a 128-bit bus with 64.00 GB/s bandwidth. The R7 M445 has 4 GB of GDDR5 on a 64-bit bus with 32.00 GB/s bandwidth. The R7 M445 has double the capacity but half the bandwidth.

Q: How do these cards compare to their nearest rivals?

A: The HD 8790M is within 1.6% of four rivals, ranging from 0.4% behind the Intel Iris Pro Graphics P6300 to 1.6% ahead of the NVIDIA Quadro M500M. The R7 M445 is within 0.9% of its rivals, from 0.2% behind the Intel UHD Graphics P630 to 0.9% ahead of the NVIDIA GeForce GTX 980M.

Architecture Differences

The two GPUs come from different architectural generations. The AMD Radeon HD 8790M uses the Mars chip based on GCN 1.0, while the AMD Radeon R7 M445 uses the Meso chip based on GCN 3.0. This is the most fundamental difference. GCN 3.0 introduced improvements in instruction scheduling and memory compression that are not present in GCN 1.0, which likely explains the R7 M445’s benchmark win despite its smaller resource pool.

The transistor counts differ substantially. The HD 8790M packs 950 million transistors on a 77 mm² die, while the R7 M445 has 1,550 million transistors on a 125 mm² die. Transistor density is nearly identical (12.3M / mm² vs. 12.4M / mm²), meaning the R7 M445 is simply a larger chip. The R7 M445’s larger die and transistor budget allow for more complex logic, even though it has fewer shading units (320 vs. 384) and TMUs (20 vs. 24). Both have 8 ROPs.

The R7 M445 is built on the same 28 nm TSMC process as the HD 8790M. The core clock behavior is similar, with the HD 8790M boosting to 900 MHz and the R7 M445 boosting to 920 MHz. However, the base clocks differ: 850 MHz for the HD 8790M and 780 MHz for the R7 M445. The R7 M445 relies more on its boost clock to achieve performance, while the HD 8790M has a higher floor.

The R7 M445 supports FP16 compute at a 1:1 ratio (588.8 GFLOPS), while the HD 8790M has no listed FP16 capability. This is a notable architectural addition in GCN 3.0 that could benefit certain compute workloads, though the provided benchmarks do not test it. The R7 M445’s DirectX 12 (12_0) support also indicates a more modern feature level compared to the HD 8790M’s DirectX 12 (11_1).

Specification Differences

The two cards differ in several key specifications. The chip is different: Mars for the HD 8790M, Meso for the R7 M445. The architecture advances from GCN 1.0 to GCN 3.0. The generation names also differ, with the HD 8790M belonging to the Solar System (HD 8700M) family and the R7 M445 to the Gem System (R7 M400) family.

Clock speeds vary slightly. The HD 8790M has a base clock of 850 MHz and a boost of 900 MHz, whereas the R7 M445 has a lower base of 780 MHz but a higher boost of 920 MHz. Memory clocks are identical at 1000 MHz (4 Gbps effective). The memory size doubles from 2 GB on the HD 8790M to 4 GB on the R7 M445, but the bus width halves from 128-bit to 64-bit, resulting in bandwidth dropping from 64.00 GB/s to 32.00 GB/s.

Compute resources are reduced in the R7 M445: shading units drop from 384 to 320, TMUs from 24 to 20, while ROPs stay at 8. Pixel rates are nearly identical (7.200 GPixel/s vs. 7.360 GPixel/s), but texture rate falls from 21.60 GTexel/s to 18.40 GTexel/s. FP32 throughput also drops from 691.2 GFLOPS to 588.8 GFLOPS, though the R7 M445 adds FP16 support at 588.8 GFLOPS.

The form factor changes from an MXM Module (with MXM-A 3.0 interface and no power connectors) to an IGP (integrated graphics processor) with a PCIe 3.0 x8 bus interface. The HD 8790M’s transistor count is 950 million on a 77 mm² die, while the R7 M445 has 1,550 million transistors on a 125 mm² die. The release dates are also different, with the HD 8790M launching in 2013 and the R7 M445 in 2016. Both are end-of-life products.