AMD Radeon HD 8730M vs AMD Radeon R7 M350 Comparison

The AMD Radeon R7 M350 and AMD Radeon HD 8730M are two end-of-life mobile GPUs separated by roughly two years, and the recorded data paints a clear picture: the newer R7 M350 wins the only direct benchmark the two share, while the older HD 8730M holds one genuine hardware advantage in memory bandwidth. Both are entry-level parts from the same manufacturer, built on the same 28 nm TSMC process, and both sit in the mid-30s percentile range against the full database of GPUs. For anyone comparing these chips today, the differences are real but modest, and this analysis breaks down exactly where each one earns its keep.

Head-to-Head Benchmarks

The database contains a single direct comparison between these two chips, and it goes decisively to the R7 M350. In Geekbench OpenCL, the R7 M350 scores 6991 against 5955 for the HD 8730M, a 17.4 percent margin in favor of the newer chip. That is a meaningful gap for parts that otherwise share the same shading unit count, and it reflects the higher clocks and newer architecture covered below.

Context matters here. The HD 8730M's 5955 average score places it in a tight cluster with its own nearest rivals: the NVIDIA Quadro K620M at 5957 (a dead-even delta of 0 percent), the AMD Radeon HD 8750M at 5970 (0.3 percent behind the 8730M), the Intel UHD Graphics 730 at 5929 (0.4 percent behind), and the NVIDIA Quadro K4000 at 5982 (0.5 percent ahead). In other words, the 8730M trades blows within a sub-1-percent band with parts spanning integrated graphics and older workstation cards.

The R7 M350's average score of 6327 puts it in the 36th percentile against all GPUs in the database, versus the 34th percentile for the HD 8730M. The M350 also has a second recorded result the older card lacks: a Geekbench Vulkan score of 5662, notably lower than its own OpenCL figure, which suggests compute workloads through OpenCL play to its strengths better than Vulkan-based paths. Its nearest rivals by average score include the AMD Radeon Pro WX 4100 at 6330 (delta of 0 percent, effectively identical), the NVIDIA Quadro K620 at 6282 (0.7 percent), and two modern high-end NVIDIA cards whose average scores of 6270 sit 0.9 percent below the M350's figure, a quirk of how averages cluster in the database rather than a claim of comparable gaming capability.

The headline: one benchmark, one winner, the R7 M350 by 17.4 percent.

Architecture Differences

Both GPUs come from AMD and both were fabricated by TSMC on a 28 nm process, but nearly everything above the silicon node diverges.

The R7 M350 uses the Meso chip, built on GCN 3.0 architecture, and belongs to the Gem System generation (R7 M300). The HD 8730M uses the Mars chip on GCN 1.0, part of the Solar System generation (HD 8700M). The generational lineage is direct: the Solar System generation is listed as the predecessor of the R7 M350's Gem System line, and the HD 8730M's successor is listed as Gem System. These are, in effect, adjacent steps on the same mobile product ladder.

The Meso die is substantially larger and denser in absolute terms: 125 mm² with 1,550 million transistors, versus 77 mm² and 950 million transistors for Mars. Transistor density is nearly identical, at 12.4M per mm² for the M350 against 12.3M per mm² for the 8730M, which is what you would expect from two dies on the same process node. The M350 simply spends more silicon.

The execution resources, however, are the same on paper: 384 shading units, 24 TMUs, and 8 ROPs on both chips. Neither has RT cores or tensor cores. The difference is how fast those resources run. The R7 M350 clocks at 1000 MHz base and 1015 MHz boost, while the HD 8730M runs at 650 MHz base and 700 MHz boost. That clock advantage drives every theoretical throughput number: the M350 delivers 8.120 GPixel/s and 24.36 GTexel/s against 5.600 GPixel/s and 16.80 GTexel/s, and 779.5 GFLOPS of FP32 against 537.6 GFLOPS. The M350 also records FP16 throughput at 779.5 GFLOPS (a 1:1 ratio with FP32); the 8730M has no FP16 figure recorded.

Memory is where the older card fights back. The HD 8730M pairs 2 GB of DDR3 with a 128-bit bus, yielding 28.80 GB/s of bandwidth. The R7 M350 has twice the capacity at 4 GB but only a 64-bit bus, dropping bandwidth to 16.00 GB/s. The M350's memory runs at 1000 MHz (2 Gbps effective) versus 900 MHz (1800 Mbps effective) for the 8730M, but the halved bus width more than cancels that clock edge. This is the single clearest hardware win for the older chip.

Software support also differs. The R7 M350 supports DirectX 12 at feature level 12_0, while the HD 8730M tops out at 12 (11_1). Both support OpenGL 4.6 and Vulkan 1.2.170. The bus interface is identical, PCIe 3.0 x8 on both.

Release dates frame the gap: the HD 8730M arrived in March 2013, the R7 M350 in May 2015. Both are now end-of-life.

Where Each One Wins

The R7 M350 wins on raw processing. The 17.4 percent Geekbench OpenCL lead comes straight from the spec sheet: same 384 shading units, but running roughly 45 percent faster at boost clock, delivered through a newer GCN revision. Compute-oriented workloads, anything measured through OpenCL, favor the M350 clearly. Its 779.5 GFLOPS of FP32 compute versus 537.6 GFLOPS makes it the stronger choice for GPU compute tasks, and its 4 GB of memory is more comfortable than 2 GB for workloads that actually allocate texture and buffer data. The full DirectX 12_0 feature level also gives it access to effects and API paths the 8730M cannot expose.

The HD 8730M wins on memory bandwidth, full stop. At 28.80 GB/s against 16.00 GB/s, it has 80 percent more bandwidth, which matters in scenarios that stream large amounts of data through the memory subsystem: higher resolution textures, larger framebuffers, or bandwidth-sensitive operations. Texture fill rate is the pressure point, since the 8730M's 16.80 GTexel/s is well below the M350's 24.36 GTexel/s, so its bandwidth edge only pays off when the workload is memory-bound rather than shader-bound.

That is the honest split from the data: the M350 wins compute and feature level, the 8730M wins bandwidth. Everything else, from bus interface to ROP count, is a wash.

The Verdict

For a direct upgrade question, the data supports the R7 M350 without hesitation. It wins the only shared benchmark by 17.4 percent, sits two percentile points higher against the database (36 versus 34), offers double the VRAM, delivers substantially higher pixel and texture fill rates, and supports the newer DirectX 12_0 feature level. It is the later, faster step in the same product lineage.

The HD 8730M's case rests on bandwidth alone. If a specific workload is demonstrably memory-bandwidth-bound and does not need the M350's compute throughput, the 128-bit bus delivers 28.80 GB/s where the M350 manages 16.00 GB/s. That is a narrow argument, but it is the only one the recorded data supports.

Both cards are end-of-life, both lack modern hardware features like ray tracing units, and neither is competitive beyond the mid-30s percentile range. Between the two, the R7 M350 is the stronger all-round part; the 8730M is the bandwidth play.

FAQ

Q: Which GPU is faster overall? A: The Radeon R7 M350. It wins the shared Geekbench OpenCL test 6991 to 5955, a 17.4 percent advantage, and its average benchmark score of 6327 beats the HD 8730M's 5955.

Q: Which card has more memory? A: The R7 M350, with 4 GB of DDR3. The HD 8730M has 2 GB. Both use DDR3.

Q: Which has more memory bandwidth? A: The HD 8730M. Its 128-bit bus delivers 28.80 GB/s, while the R7 M350's 64-bit bus manages 16.00 GB/s despite slightly faster memory clocks.

Q: Do they have the same number of shader units? A: Yes. Both have 384 shading units, 24 TMUs, and 8 ROPs. The performance difference comes from clocks and architecture, not resource counts.

Q: What DirectX feature level does each support? A: The R7 M350 supports DirectX 12 at feature level 12_0. The HD 8730M supports DirectX 12 at feature level 11_1. Both support OpenGL 4.6 and Vulkan 1.2.170.

Q: Are these cards still in production? A: No. Both are listed as end-of-life. The HD 8730M launched in March 2013 and the R7 M350 in May 2015.

Specification Differences

| Field | AMD Radeon R7 M350 | AMD Radeon HD 8730M |

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

| Chip | Meso | Mars |

| Architecture | GCN 3.0 | GCN 1.0 |

| Generation | Gem System (R7 M300) | Solar System (HD 8700M) |

| Transistors | 1,550 million | 950 million |

| Die size | 125 mm² | 77 mm² |

| Transistor density | 12.4M / mm² | 12.3M / mm² |

| Base clock | 1000 MHz | 650 MHz |

| Boost clock | 1015 MHz | 700 MHz |

| Memory clock | 1000 MHz (2 Gbps effective) | 900 MHz (1800 Mbps effective) |

| Memory size | 4 GB | 2 GB |

| Memory bus width | 64 bit | 128 bit |

| Memory bandwidth | 16.00 GB/s | 28.80 GB/s |

| Pixel rate | 8.120 GPixel/s | 5.600 GPixel/s |

| Texture rate | 24.36 GTexel/s | 16.80 GTexel/s |

| FP32 compute | 779.5 GFLOPS | 537.6 GFLOPS |

| FP16 compute | 779.5 GFLOPS (1:1) | Not recorded |

| DirectX | 12 (12_0) | 12 (11_1) |

| Release date | May 2015 | March 2013 |

| Predecessor | Solar System | London |

| Successor | Polaris Mobile | Gem System |

| Avg benchmark score | 6327 | 5955 |

| Percentile vs all GPUs | 36 | 34 |

Identical between the two: manufacturer (AMD), foundry (TSMC), 28 nm process node, 384 shading units, 24 TMUs, 8 ROPs, no RT or tensor cores, DDR3 memory type, PCIe 3.0 x8 bus interface, OpenGL 4.6, Vulkan 1.2.170, and end-of-life production status.