AMD Radeon R7 M350 vs NVIDIA GeForce GTX 670M Comparison

Where Each One Wins

The benchmark data splits cleanly between these two mobile GPUs, but the split is not where the specifications would suggest. The AMD Radeon R7 M350 wins the only direct head-to-head benchmark available, taking the Geekbench OpenCL test with a score of 6991 against the NVIDIA GeForce GTX 670M’s 6513. That is a 6.8% margin in AMD’s favor, and it is the sole recorded win in the head-to-head results.

However, the picture becomes more nuanced when looking at average benchmark scores. The NVIDIA card’s average benchmark score is 6513, which is identical to its single OpenCL result. The AMD card’s average is 6327, which is lower than its OpenCL score of 6991 because it also has a Geekbench Vulkan result of 5662 dragging the average down. So while AMD wins the OpenCL test outright, its overall average is 2.9% below NVIDIA’s (6513 vs 6327). This means the R7 M350 is not uniformly faster; its advantage is specific to the OpenCL workload, while the Vulkan result pulls its aggregate down.

In terms of percentile ranking against all GPUs, the GTX 670M sits at the 38th percentile, while the R7 M350 sits at the 36th. The NVIDIA part is therefore ranked slightly higher overall despite losing the single head-to-head test. The nearest rivals for each card reinforce this split: the GTX 670M’s closest competitor is the GeForce GT 555M, which scores 6493 (0.3% behind), while the R7 M350’s closest rival is the Radeon Pro WX 4100 at 6330 (a 0.0% delta). The R7 M350 also has two NVIDIA RTX 40-series and 50-series parts listed as rivals, both scoring 6270 (0.9% behind), which is an artifact of the averaging method rather than a direct architectural comparison.

For use-case segmentation, the data suggests the R7 M350 is the better choice for OpenCL-compute workloads, where it leads by 6.8%. The GTX 670M, with its higher average score and better percentile, is the more consistent performer across the two benchmark types recorded for the AMD card. If a workload relies on Vulkan, the R7 M350’s 5662 score indicates a significant drop-off, whereas the GTX 670M has no recorded Vulkan result to compare. The practical takeaway: the AMD part wins where it wins by a clear margin, but the NVIDIA part is more stable in aggregate.

FAQ

Q: Which GPU has the higher OpenCL benchmark score?

A: The AMD Radeon R7 M350 scores 6991 in Geekbench OpenCL, which is 6.8% higher than the NVIDIA GeForce GTX 670M’s 6513.

Q: Does the AMD card also win on average benchmark score?

A: No. The R7 M350’s average benchmark score is 6327, which is lower than the GTX 670M’s 6513, because the AMD card’s Vulkan score of 5662 pulls its average down.

Q: What is the percentile ranking difference between the two?

A: The GTX 670M is at the 38th percentile of all GPUs, while the R7 M350 is at the 36th percentile. The NVIDIA card ranks two percentage points higher.

Q: Which GPU has the higher pixel fill rate?

A: The GTX 670M has a pixel rate of 8.372 GPixel/s, which is slightly higher than the R7 M350’s 8.120 GPixel/s.

Q: Does the R7 M350 support Vulkan?

A: Yes, the R7 M350 lists Vulkan 1.2.170 in its API support and has a Geekbench Vulkan score of 5662. The GTX 670M does not list a Vulkan version or score in the data.

Q: Which GPU has more shading units?

A: The R7 M350 has 384 shading units, while the GTX 670M has 336 shading units. AMD’s card has 48 more shading units.

Head-to-Head Benchmarks

The only direct comparison available is the Geekbench OpenCL test, and it is decisive. The AMD Radeon R7 M350 scores 6991, while the NVIDIA GeForce GTX 670M scores 6513. The delta is -6.8% from NVIDIA’s perspective, meaning AMD is ahead by that margin. This is a meaningful gap in a compute-oriented benchmark, and it aligns with the R7 M350’s higher shading unit count (384 vs 336) and its newer GCN 3.0 architecture.

The broader benchmark context complicates the story. The GTX 670M’s average score equals its OpenCL score (6513), indicating that this is its only recorded benchmark. The R7 M350 has two results: the strong OpenCL score of 6991 and a much weaker Vulkan score of 5662. The Vulkan result is 19.0% below the OpenCL result (6991 vs 5662), which is a substantial intra-card gap. This suggests that the R7 M350’s architecture is far more efficient in OpenCL than in Vulkan, at least in this data set.

Looking at the nearest rivals, the GTX 670M’s closest competitor (GeForce GT 555M) scores 6493, which is only 0.3% behind. The R7 M350’s closest rival (Radeon Pro WX 4100) scores 6330, a 0.0% delta. This means that in the aggregate, the R7 M350 is essentially tied with a professional workstation card, while the GTX 670M is barely ahead of an older mainstream mobile part. The head-to-head win for AMD is real, but it does not translate into a higher average or percentile ranking.

The pixel rate comparison is nearly a tie: 8.372 GPixel/s for NVIDIA vs 8.120 GPixel/s for AMD, a difference of 3.0%. The texture rate favors NVIDIA more clearly: 33.49 GTexel/s vs 24.36 GTexel/s, a 27.3% advantage for the GTX 670M. Yet the OpenCL score still goes to AMD, which indicates that compute performance is not solely determined by texture or pixel throughput.

Specification Differences

The two GPUs differ in nearly every major specification field. The GTX 670M uses a 40 nm process node, while the R7 M350 uses a 28 nm node. The transistor counts are close — 1,950 million for NVIDIA vs 1,550 million for AMD — but the die sizes diverge sharply: 332 mm² for the GTX 670M vs 125 mm² for the R7 M350. This produces a transistor density of 5.9M per mm² for NVIDIA and 12.4M per mm² for AMD, meaning AMD packs more than twice as many transistors per square millimeter.

Memory configurations are fundamentally different. The GTX 670M has 1536 MB of GDDR5 on a 192-bit bus, delivering 72.00 GB/s of bandwidth. The R7 M350 has 4 GB of DDR3 on a 64-bit bus, delivering only 16.00 GB/s. That is a 4.5x bandwidth advantage for NVIDIA (72.00 vs 16.00 GB/s), while AMD has 2.67x more memory capacity (4 GB vs 1536 MB). The memory clock differs as well: 750 MHz (3 Gbps effective) for NVIDIA vs 1000 MHz (2 Gbps effective) for AMD.

The compute units also differ. The GTX 670M has 336 shading units, 56 TMUs, and 24 ROPs. The R7 M350 has 384 shading units, 24 TMUs, and 8 ROPs. NVIDIA has more than double the TMUs (56 vs 24) and triple the ROPs (24 vs 8), while AMD has 14.3% more shading units (384 vs 336). The R7 M350 also has a base clock of 1000 MHz and a boost clock of 1015 MHz, while the GTX 670M lists no base or boost clock, only a memory clock.

The bus interface differs: the GTX 670M uses MXM-B (3.0), while the R7 M350 uses PCIe 3.0 x8. The NVIDIA card is a 75 W MXM module with no power connectors, while the AMD card lists no TDP, slot width, or power connector information. The GTX 670M supports DirectX 12 (11_0) and OpenGL 4.6, while the R7 M350 supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.2.170.

Architecture Differences

The architectural gap is generational. The GTX 670M is built on NVIDIA’s Fermi 2.0 architecture, using the GF114 chip, and belongs to the GeForce 600M generation. The R7 M350 is built on AMD’s GCN 3.0 architecture, using the Meso chip, and belongs to the Gem System (R7 M300) generation. Fermi 2.0 is a 2012-era design, while GCN 3.0 is a 2015-era design, which explains the process node difference (40 nm vs 28 nm).

The foundry is the same — TSMC — but the design philosophies differ. NVIDIA’s GF114 die is 332 mm², which is 2.66x larger than AMD’s 125 mm² Meso die. Despite the smaller die, AMD achieves a higher transistor density (12.4M vs 5.9M per mm²) because of the newer process. The R7 M350 also has a higher shading unit count (384 vs 336), suggesting a wider SIMD design, but it has far fewer TMUs and ROPs, which limits its texture and pixel throughput.

The memory architecture reflects the different target markets. The GTX 670M uses GDDR5 on a 192-bit bus, which is a classic performance-oriented configuration for its era. The R7 M350 uses DDR3 on a 64-bit bus, which is a power-efficient, cost-conscious configuration. The R7 M350’s 1:1 FP16 to FP32 ratio (779.5 GFLOPS each) indicates that it can handle half-precision workloads at full rate, while the GTX 670M lists no FP16 support. The GTX 670M’s FP32 throughput is 803.7 GFLOPS, which is 3.1% higher than the R7 M350’s 779.5 GFLOPS.

The API support is another differentiator. The R7 M350 supports DirectX 12 (12_0) and Vulkan 1.2.170, both of which are more modern than the GTX 670M’s DirectX 12 (11_0) and lack of Vulkan. The GTX 670M’s predecessor is GeForce 500M and its successor is GeForce 700M, while the R7 M350’s predecessor is Solar System and its successor is Polaris Mobile. The release dates reflect the gap: March 2012 for NVIDIA, May 2015 for AMD.

The Verdict

The data does not support a single winner; it supports two different winners depending on the metric. For raw OpenCL compute, the AMD Radeon R7 M350 is the clear choice, scoring 6991 against the GTX 670M’s 6513, a 6.8% advantage. For aggregate consistency and percentile ranking, the NVIDIA GeForce GTX 670M is ahead, with an average score of 6513 vs 6327 and a 38th percentile rank vs 36th.

The GTX 670M also holds decisive leads in bandwidth (72.00 GB/s vs 16.00 GB/s), texture rate (33.49 GTexel/s vs 24.36 GTexel/s), and pixel rate (8.372 GPixel/s vs 8.120 GPixel/s). These are classic gaming-oriented strengths. The R7 M350 counters with more memory (4 GB vs 1536 MB), more shading units (384 vs 336), and a modern API set including Vulkan 1.2.170.

A user prioritizing compute workloads in OpenCL should pick the R7 M350. A user prioritizing consistent performance across a broader set of metrics, or who values higher memory bandwidth and texture throughput, should pick the GTX 670M. The R7 M350’s Vulkan score of 5662 is a warning that its performance is not uniform across APIs, while the GTX 670M has no such recorded weakness — but also no recorded Vulkan capability at all.

Given the end-of-life status of both parts, the choice is academic for new purchases, but for benchmark-based comparisons, the data says: the AMD part wins the compute test, the NVIDIA part wins the aggregate. Neither dominates. The 6.8% OpenCL win for AMD is offset by a 2.9% average-score deficit, and the percentile gap (38 vs 36) is narrow. The verdict is that the R7 M350 is the better compute part in OpenCL, while the GTX 670M is the more balanced part overall.