AMD Radeon HD 8690M vs AMD Radeon R7 M370 Comparison

The AMD Radeon R7 M370 and AMD Radeon HD 8690M are two mobile graphics solutions from different eras of AMD’s product stack, yet they share the same fundamental GCN 1.0 architecture and 28 nm TSMC manufacturing process. The data reveals a clear but narrow performance gap, with the R7 M370 holding a 15.1% lead in the only directly comparable benchmark. This analysis breaks down where each chip excels, what architectural choices drive the differences, and what the numbers mean for real-world usage scenarios.

Where Each One Wins

The benchmark data is unambiguous: the AMD Radeon R7 M370 wins the only head-to-head test available. In Geekbench OpenCL, the R7 M370 scores 7063 against the HD 8690M’s 6137, a 15.1% advantage. This single victory gives the R7 M370 a 1-0 win record in the head-to-head comparison. For compute-oriented workloads that rely on OpenCL acceleration—tasks like video encoding, image processing, or general-purpose GPU computation—the R7 M370 is the clear choice based on this metric.

The HD 8690M, however, does not emerge entirely without merit. Its average benchmark score of 6137 places it in the 36th percentile of all GPUs, while the R7 M370 sits at the 38th percentile. This close percentile positioning suggests that in the broader GPU landscape, both chips occupy a similar tier of performance. The HD 8690M’s nearest rivals include the Intel Iris Pro Graphics 6200 (0.3% ahead) and the NVIDIA GeForce MX230 (1% behind), indicating it trades blows with integrated and entry-level discrete solutions from the same era. The R7 M370, by contrast, sits near the NVIDIA GeForce GTX 675M (2.6% behind) and the AMD Radeon R7 M460 (2.3% ahead), showing it competes in a slightly higher performance bracket.

Where the HD 8690M might be preferable is in scenarios where power draw is a concern, though the FACT PACK provides no TDP figures to confirm this. What the data does show is that the HD 8690M achieves its performance with 320 shading units, 20 TMUs, and a smaller 56 mm² die, suggesting a leaner implementation. For legacy compatibility or drivers tuned for older GCN 1.0 hardware, the HD 8690M’s earlier release date (February 2013 vs. May 2015) might make it a more tested platform, but the benchmark evidence favors the R7 M370 in raw compute throughput.

Architecture Differences

Both GPUs are built on GCN 1.0 architecture and use TSMC’s 28 nm process node, which explains their identical transistor density of 12.3M per mm². The R7 M370 uses the “Litho” chip, while the HD 8690M uses the “Sun” chip, and this difference in silicon is where the performance gap originates. The R7 M370 packs 950 million transistors across a 77 mm² die, compared to the HD 8690M’s 690 million transistors on a 56 mm² die. This 260 million transistor difference translates directly into more execution resources.

The R7 M370 fields 384 shading units, 24 texture mapping units, and 8 ROPs. The HD 8690M counters with 320 shading units, 20 TMUs, and the same 8 ROPs. The R7 M370’s advantage of 64 additional shading units and 4 extra TMUs is the core architectural reason for its higher compute throughput. The pixel rate tells an interesting story: the HD 8690M actually edges out the R7 M370 in pixel fill rate at 7.800 GPixel/s versus 7.680 GPixel/s, thanks to its higher boost clock of 975 MHz versus 960 MHz. However, the R7 M370’s texture rate of 23.04 GTexel/s surpasses the HD 8690M’s 19.50 GTexel/s, and its FP32 performance of 737.3 GFLOPS beats the HD 8690M’s 624.0 GFLOPS.

Memory architecture presents the most significant divergence. Both cards use 2 GB of GDDR5, but the R7 M370 employs a 128-bit memory bus versus the HD 8690M’s 64-bit bus. This yields a massive bandwidth difference: 57.60 GB/s for the R7 M370 versus 32.00 GB/s for the HD 8690M. The HD 8690M compensates with a higher memory clock of 1000 MHz (4 Gbps effective) compared to the R7 M370’s 900 MHz (3.6 Gbps effective), but the narrower bus cannot overcome the R7 M370’s wider path. Both cards share the same API support: DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170, so software compatibility is identical.

Head-to-Head Benchmarks

The only directly comparable benchmark in the FACT PACK is Geekbench OpenCL, and the result is decisive. The R7 M370 scores 7063, while the HD 8690M scores 6137, producing a 15.1% delta in favor of the R7 M370. This is a substantial margin for two GPUs from the same architecture family, and it aligns with the raw compute specifications. The R7 M370’s FP32 output of 737.3 GFLOPS represents an 18.2% advantage over the HD 8690M’s 624.0 GFLOPS, which closely tracks the benchmark delta. The memory bandwidth gap is even larger—the R7 M370’s 57.60 GB/s is 80% higher than the HD 8690M’s 32.00 GB/s—yet the OpenCL score only shows a 15.1% difference, suggesting that compute workloads in this benchmark are not purely bandwidth-limited.

The R7 M370 also has a Geekbench Vulkan score of 6465, which the HD 8690M lacks entirely. This absence is notable; it suggests the HD 8690M either was not tested under Vulkan or does not support it in the benchmark environment, despite both cards listing Vulkan 1.2.170 in their API specs. The Vulkan score of 6465 is lower than the OpenCL score of 7063 for the R7 M370, indicating that the card performs better under OpenCL’s compute model than Vulkan’s combined graphics/compute workload. This data point is useful for users who prioritize OpenCL applications over Vulkan-based ones.

When examining the rival landscape, the R7 M370’s average benchmark score of 6764 places it within a tight cluster: the AMD FirePro M5100 is 1% behind, the NVIDIA GeForce GT 1010 is 1% ahead, and the AMD Radeon R7 M460 is 2.3% ahead. The HD 8690M’s average of 6137 sits near the Intel Iris Pro Graphics 6200 (0.3% ahead) and the NVIDIA RTX A400 (1% ahead). This positioning underscores that the R7 M370 competes with mid-range mobile GPUs from 2015-2016, while the HD 8690M aligns with the upper end of integrated graphics from the same period.

FAQ

Q: Which GPU has higher raw compute performance?

A: The AMD Radeon R7 M370 delivers 737.3 GFLOPS of FP32 performance, compared to the HD 8690M’s 624.0 GFLOPS, a difference that translates to a 15.1% higher Geekbench OpenCL score (7063 vs 6137).

Q: Do both GPUs support the same modern APIs?

A: Yes, both list DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170 in their specifications, providing identical API feature sets for software compatibility.

Q: How do the memory systems differ?

A: The R7 M370 uses a 128-bit bus giving 57.60 GB/s bandwidth, while the HD 8690M uses a 64-bit bus for 32.00 GB/s. Both have 2 GB of GDDR5, but the HD 8690M runs its memory faster at 1000 MHz (4 Gbps effective) versus 900 MHz (3.6 Gbps effective).

Q: Which GPU has a higher pixel fill rate?

A: The HD 8690M edges out the R7 M370 with 7.800 GPixel/s versus 7.680 GPixel/s, due to its higher boost clock of 975 MHz compared to 960 MHz, despite having fewer shading units.

Q: What is the transistor count difference?

A: The R7 M370 contains 950 million transistors on a 77 mm² die, while the HD 8690M has 690 million on a 56 mm² die, both manufactured on TSMC’s 28 nm process with identical 12.3M / mm² transistor density.

Q: Which GPU has better Vulkan benchmark performance?

A: Only the R7 M370 has a recorded Geekbench Vulkan score of 6465; the HD 8690M has no Vulkan benchmark data in the FACT PACK, so no direct comparison is possible.

The Verdict

The data points to a clear recommendation: choose the AMD Radeon R7 M370 for any workload that benefits from compute throughput or memory bandwidth. Its 15.1% OpenCL lead, substantial 80% bandwidth advantage, and higher FP32 output make it the stronger performer in the only direct comparison available. The R7 M370’s 38th percentile ranking versus the HD 8690M’s 36th percentile further confirms its superior standing among all GPUs.

However, the HD 8690M is not without a niche. Its higher pixel rate of 7.800 GPixel/s versus 7.680 GPixel/s suggests it might perform marginally better in pixel-bound rendering scenarios, even if its overall compute is weaker. For users with applications that are heavily pixel-fill-rate dependent, the HD 8690M could hold a slight edge. Additionally, its earlier release date (February 2013) means it has had more time in the field, potentially benefiting from longer driver maturation cycles. The HD 8690M’s smaller die size of 56 mm² versus 77 mm² might also imply lower power consumption, though no TDP data exists to verify this.

The R7 M370’s Vulkan score of 6465, while lower than its OpenCL score, demonstrates broader benchmark coverage and potentially better driver support for modern APIs. Its successor, Polaris Mobile, and predecessor, Solar System, indicate it sits in a transitional generation, whereas the HD 8690M’s successor is the Gem System, the exact family the R7 M370 belongs to. This lineage suggests the R7 M370 represents an evolution of the HD 8690M’s design philosophy, with more transistors, wider memory, and higher compute resources.

For most users, the R7 M370 is the superior choice based on benchmark evidence. For those specifically optimizing for pixel fill rate or seeking a leaner silicon footprint, the HD 8690M remains a viable, though dated, alternative. The 1-0 head-to-head record does not tell the full story—the R7 M370’s advantages in memory and compute are decisive enough to recommend it without reservation in compute-heavy tasks.

Specification Differences

| Specification | AMD Radeon R7 M370 | AMD Radeon HD 8690M |

|---|---|---|

| Chip | Litho | Sun |

| Generation | Gem System (R7 M300) | Solar System (HD 8600M) |

| Transistors | 950 million | 690 million |

| Die Size | 77 mm² | 56 mm² |

| Base Clock | 875 MHz | 925 MHz |

| Boost Clock | 960 MHz | 975 MHz |

| Memory Clock | 900 MHz (3.6 Gbps effective) | 1000 MHz (4 Gbps effective) |

| Memory Bus Width | 128 bit | 64 bit |

| Memory Bandwidth | 57.60 GB/s | 32.00 GB/s |

| Shading Units | 384 | 320 |

| TMUs | 24 | 20 |

| Pixel Rate | 7.680 GPixel/s | 7.800 GPixel/s |

| Texture Rate | 23.04 GTexel/s | 19.50 GTexel/s |

| FP32 | 737.3 GFLOPS | 624.0 GFLOPS |

| Release Date | 2015-05-04 | 2013-02-28 |

| Predecessor | Solar System | London |

| Successor | Polaris Mobile | Gem System |

| Geekbench OpenCL | 7063 | 6137 |

| Geekbench Vulkan | 6465 | N/A |

| Percentile | 38 | 36 |

| Average Benchmark Score | 6764 | 6137 |