AMD Radeon R7 M350 vs NVIDIA GeForce GTX 670MX Comparison

Head-to-Head Benchmarks

The recorded data shows a clear, consistent advantage for the AMD Radeon R7 M350 across both benchmark workloads. In the Geekbench OpenCL test, the AMD part scores 6,991 points against 6,125 points for the NVIDIA GeForce GTX 670MX, a decisive 14.1% margin. This is not a narrow win; it represents a substantial performance gap in compute-oriented tasks, where the R7 M350's architecture and driver optimization appear to deliver superior raw throughput.

The Vulkan workload tells a similar story, though the gap narrows considerably. The R7 M350 posts 5,662 points versus 5,316 points for the GTX 670MX, a 6.5% advantage. While still a win for AMD, the reduced delta in Vulkan suggests that NVIDIA's Kepler architecture handles the lower-level API overhead relatively better than it does OpenCL, or that the R7 M350's advantage shrinks when both cards are pushed closer to their respective limits in graphics-centric workloads.

Interpreting these scores through the database's percentile rankings, the R7 M350 sits at the 36th percentile of all GPUs, while the GTX 670MX sits at the 33rd percentile. The absolute delta of 606 points in average benchmark score (6,327 versus 5,721) places the AMD card roughly 10.6% ahead on aggregate. The head-to-head record is 2 wins for the R7 M350, 0 wins for the GTX 670MX, so there is no benchmark category in the database where the NVIDIA part emerges ahead.

Looking at the nearest rival clusters provides additional context for each card's standing. The R7 M350's average score of 6,327 places it statistically tied with the AMD Radeon Pro WX 4100 (6,330, 0% delta) and marginally ahead of the NVIDIA Quadro K620 (6,282, 0.7% delta). It also edges out the NVIDIA GeForce RTX 5070 Ti SUPER and the RTX 4070 Ti SUPER AD102, both at 6,270 with a 0.9% delta. This suggests that the R7 M350, despite its modest specifications, lands in a performance tier that overlaps with much more recent hardware in these specific compute benchmarks.

The GTX 670MX's average of 5,721 places it just ahead of the Intel Iris Pro Graphics P6300 (5,712, 0.2% delta) and the AMD Radeon HD 8790M (5,691, 0.5% delta). It trails the NVIDIA GeForce GTX 550 Ti by a negligible 0.2% (5,731 versus 5,721) and leads the NVIDIA Quadro M500M by 2.1% (5,604). The GTX 670MX is therefore clustered with older mid-range mobile and desktop parts, whereas the R7 M350 reaches into a slightly higher performance bracket.

Architecture Differences

The two GPUs come from different architectural generations and are built for different market segments, which explains much of the benchmark divergence. The AMD Radeon R7 M350 uses the Meso chip on the GCN 3.0 architecture, part of the Gem System (R7 M300) generation. The NVIDIA GeForce GTX 670MX uses the GK104 chip on the Kepler architecture, part of the GeForce 600M generation. Both are fabricated on a 28 nm process at TSMC, so the process node is identical, but the chip designs diverge sharply.

Transistor counts reveal a significant difference in complexity. The GTX 670MX packs 3,540 million transistors on a 294 mm² die, while the R7 M350 uses 1,550 million transistors on a 125 mm² die. The NVIDIA chip is more than twice the physical size and houses more than double the transistor count. Transistor density is nearly identical (12.4 million per mm² for AMD versus 12.0 million per mm² for NVIDIA), indicating that the size difference comes from sheer scale rather than packing efficiency.

The compute resources follow the transistor disparity. The GTX 670MX has 960 shading units, 80 texture mapping units, and 24 ROPs. The R7 M350 has 384 shading units, 24 TMUs, and 8 ROPs. Despite having 2.5 times the shader count, the GTX 670MX only achieves 1,153.9 GFLOPS of FP32 throughput versus 779.5 GFLOPS for the R7 M350, a ratio of 1.48. This is due to the GTX 670MX's low clock speeds: 601 MHz base and boost, compared to the R7 M350's 1000 MHz base and 1015 MHz boost. The AMD part runs nearly 70% higher clocked, which partially compensates for its smaller execution engine.

Memory configurations differ substantially in both capacity and type. The R7 M350 has 4 GB of DDR3 on a 64-bit bus, yielding 16.00 GB/s of bandwidth. The GTX 670MX has 3 GB of GDDR5 on a 192-bit bus, yielding 67.20 GB/s, which is 4.2 times higher bandwidth. The NVIDIA card's memory subsystem is clearly superior for bandwidth-bound workloads, yet the benchmark results show the AMD card winning, indicating that these specific Geekbench tests are not primarily bandwidth-limited.

Pixel and texture rates also favor the GTX 670MX. The NVIDIA card achieves 12.02 GPixel/s and 48.08 GTexel/s, versus 8.120 GPixel/s and 24.36 GTexel/s for the R7 M350. The GTX 670MX leads by 48% in pixel throughput and 97% in texture throughput. This is a notable discrepancy given that the R7 M350 wins the overall benchmarks, suggesting that the OpenCL and Vulkan tests reward compute and driver efficiency more than raw rasterization throughput.

Power characteristics are only recorded for the GTX 670MX, which has a TDP of 75 W and requires no power connectors. The R7 M350 has no TDP figure in the database. The GTX 670MX uses a PCIe 3.0 x16 interface, while the R7 M350 uses PCIe 3.0 x8, a halved lane count that could theoretically limit bandwidth in some scenarios, though the benchmark data does not indicate a penalty.

API support shows a generation gap. The R7 M350 supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.2.170. The GTX 670MX supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The feature level difference in DirectX (12_0 versus 11_0) reflects the architectural advances in GCN 3.0, though both cards are end-of-life products in the database.

The Verdict

The data points to a straightforward conclusion: the AMD Radeon R7 M350 outperforms the NVIDIA GeForce GTX 670MX in every recorded benchmark. The R7 M350 wins the OpenCL test by 14.1% and the Vulkan test by 6.5%, with an average score advantage of 10.6%. The percentile ranking confirms this, with the AMD card at the 36th percentile versus the NVIDIA card at the 33rd.

However, the comparison is not flattering to either card in absolute terms. Both sit in the lower third of the GPU performance distribution. The R7 M350's nearest rivals include workstation and desktop parts from both vendors, while the GTX 670MX's nearest rivals are older or integrated solutions. The R7 M350's average score of 6,327 is competitive with the AMD Radeon Pro WX 4100 and the NVIDIA Quadro K620, which suggests it delivers compute performance comparable to entry-level professional GPUs despite its mobile-oriented design.

The GTX 670MX, with an average score of 5,721, is closely matched with the NVIDIA GeForce GTX 550 Ti (5,731, a 0.2% deficit) and the Intel Iris Pro Graphics P6300 (5,712, a 0.2% lead). This places it in a performance tier that is now clearly dated. The R7 M350, while also end-of-life, is a newer design (release date 2015-05-04 versus 2012-09-30) and benefits from architectural improvements in GCN 3.0.

For users choosing between these two based strictly on the recorded data, the R7 M350 is the superior choice in compute benchmarks. The GTX 670MX offers higher memory bandwidth, more shading units, and faster pixel/texture rates, but these specifications do not translate into wins in the tests that matter here. The R7 M350's higher clock speeds and newer architecture appear to be the deciding factors.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The AMD Radeon R7 M350 has an average benchmark score of 6,327, compared to 5,721 for the NVIDIA GeForce GTX 670MX, a difference of 606 points.

Q: What is the largest performance gap between the two cards in a single test?

A: The largest gap is in Geekbench OpenCL, where the R7 M350 scores 6,991 versus 6,125 for the GTX 670MX, a 14.1% advantage.

Q: How do the two cards compare in terms of memory bandwidth?

A: The GTX 670MX has significantly higher memory bandwidth at 67.20 GB/s (GDDR5, 192-bit bus) versus 16.00 GB/s (DDR3, 64-bit bus) for the R7 M350.

Q: Which GPU has more shading units?

A: The GTX 670MX has 960 shading units, while the R7 M350 has 384 shading units. Despite this, the R7 M350 wins both recorded benchmarks.

Q: What is the DirectX support difference between the two?

A: The R7 M350 supports DirectX 12 (12_0), while the GTX 670MX supports DirectX 12 (11_0), indicating a higher feature level on the AMD part.

Q: How does the R7 M350 compare to its nearest rival, the AMD Radeon Pro WX 4100?

A: The R7 M350 scores 6,327 on average, essentially tied with the Radeon Pro WX 4100 at 6,330, a 0% delta.

Where Each One Wins

The AMD Radeon R7 M350 wins in all recorded benchmark categories, but the nature of its victories points to specific strengths. In the OpenCL workload, the 14.1% margin suggests that the GCN 3.0 architecture's compute pipeline, combined with the 1015 MHz boost clock, provides a substantial advantage in general-purpose GPU computing. The FP32 throughput of 779.5 GFLOPS, while lower in absolute terms than the GTX 670MX's 1,153.9 GFLOPS, is evidently more efficiently utilized in these tests. The R7 M350's 4 GB memory capacity, though on a narrow 64-bit bus, may also help in workloads that require larger working sets.

The Vulkan test shows a narrower 6.5% win, indicating that the R7 M350's advantage persists but diminishes in a lower-level graphics API. This could reflect the GTX 670MX's Kepler architecture being relatively more competitive in raw graphics throughput, given its higher pixel rate (12.02 GPixel/s versus 8.120 GPixel/s) and texture rate (48.08 GTexel/s versus 24.36 GTexel/s). For users running Vulkan-based applications, the R7 M350 still leads, but the margin is less comfortable.

The NVIDIA GeForce GTX 670MX does not win any benchmark category in the database, but its specification sheet suggests scenarios where it could be preferable despite the scores. The 67.20 GB/s memory bandwidth is 4.2 times higher than the R7 M350's, which would matter in bandwidth-bound tasks such as large texture streaming or certain compute kernels that the Geekbench tests may not fully exercise. The higher ROP count (24 versus 8) and TMU count (80 versus 24) indicate stronger fill-rate capabilities for traditional rasterization, even though the clock speed penalty (601 MHz versus 1000 MHz) limits realized performance.

For use-case selection, the data supports the R7 M350 for general compute workloads and OpenCL/Vulkan applications. The GTX 670MX, despite its losses, retains a theoretical edge in memory-intensive and fill-rate-limited scenarios, though the recorded benchmarks do not confirm this. The R7 M350's lower transistor count (1,550 million versus 3,540 million) and smaller die (125 mm² versus 294 mm²) also suggest it may be more efficient in terms of silicon cost, though power data is only available for the NVIDIA card (75 W TDP). Ultimately, the database records 2 wins for the R7 M350 and 0 for the GTX 670MX, making the AMD card the statistically superior choice in every measured dimension.