AMD Radeon R7 350 vs NVIDIA GeForce GTX 670M Comparison

Head-to-Head Benchmarks

The only directly comparable measurement in the database is the Geekbench OpenCL compute test, and the result is decisive. The AMD Radeon R7 350 scores 7792, while the NVIDIA GeForce GTX 670M scores 6513. That puts the R7 350 ahead by 19.6%, a substantial margin for a synthetic compute workload. In practical terms, the R7 350 delivers nearly a fifth more raw compute throughput in OpenCL, which matters for applications that offload parallel math to the GPU rather than relying on the CPU.

Looking at the wider database context, the R7 350 sits at the 40th percentile among all GPUs, while the GTX 670M sits at the 38th percentile. That two-point gap in percentile ranking reflects the measured 19.6% delta in the head-to-head test. The R7 350's average benchmark score across all recorded tests is 7425, whereas the GTX 670M's average is 6513, a difference of 912 points. The R7 350 has the advantage of a second recorded benchmark, a Geekbench Vulkan score of 7057, which the GTX 670M cannot match because it has no Vulkan result in the database. That missing data point is itself a meaningful distinction: the R7 350 is a more modern part with broader API support.

The nearest rival data reinforces where each card stands in the broader landscape. The R7 350's closest competitors by average score are the Intel UHD Graphics 750 at 7441 (0.2% behind), the AMD Radeon HD 8850M at 7447 (0.3% behind), the NVIDIA GeForce GTX 1650 at 7472 (0.6% behind), and the Intel Arc A310 at 7550 (1.7% behind). The GTX 670M's nearest rivals are the NVIDIA GeForce GT 555M at 6493 (0.3% ahead), the NVIDIA Quadro M5000M at 6481 (0.5% ahead), the AMD Radeon Vega 10 Mobile at 6476 (0.6% ahead), and the Intel UHD Graphics P750 at 6554 (0.6% behind). The R7 350 is trading with much newer integrated and entry-level discrete parts, whereas the GTX 670M is surrounded by older mobile and workstation silicon.

The win count is one to zero in favor of the R7 350. There is no benchmark in the database where the GTX 670M comes out ahead. The single head-to-head test, Geekbench OpenCL, is the only direct comparison available, and the R7 350 takes it by a clear margin. The absence of a Vulkan result for the GTX 670M means the R7 350 also wins the breadth of testing argument by default.

Where Each One Wins

The R7 350 wins where modern API support and compute throughput matter. Its OpenCL score of 7792 versus 6513 for the GTX 670M shows a 19.6% advantage in heterogeneous compute workloads. The Vulkan score of 7057, while not directly comparable to anything the GTX 670M has in the database, demonstrates that the R7 350 can run modern graphics APIs. This makes it the better choice for OpenCL compute tasks, Vulkan-based workloads, and any application that can leverage the newer API stack. The R7 350 also has a higher pixel rate, 12.80 GPixel/s versus 8.372 GPixel/s, which indicates faster rasterization throughput at the output stage.

The GTX 670M wins on texture throughput. Its texture rate of 33.49 GTexel/s exceeds the R7 350's 25.60 GTexel/s, a difference that favors the NVIDIA part in texture-heavy scenes where the GPU is feeding the pixel pipeline with filtered samples. The GTX 670M also has more TMUs, 56 versus 32, and more ROPs, 24 versus 16, which gives it a structural advantage in certain fill-rate-limited scenarios. Its FP32 throughput of 803.7 GFLOPS is close to the R7 350's 819.2 GFLOPS, so the compute gap is not as large as the OpenCL benchmark might suggest on paper.

For gaming, the picture is mixed. The R7 350 has the modern architecture and API support, but the GTX 670M has the higher texture rate and more ROPs. The database does not include gaming benchmarks, so the practical outcome depends on the specific title and API. The R7 350's DirectX 12 (11_1) support versus the GTX 670M's DirectX 12 (11_0) support gives the AMD part a slight edge in titles that use newer DirectX 12 features. The GTX 670M's lack of Vulkan support in the database means any Vulkan-based game or application is automatically out of reach for the NVIDIA card.

FAQ

Q: Which GPU has the higher OpenCL benchmark score?

A: The AMD Radeon R7 350 scores 7792 in Geekbench OpenCL, while the NVIDIA GeForce GTX 670M scores 6513. The R7 350 wins by 19.6%.

Q: Does the NVIDIA GeForce GTX 670M support Vulkan?

A: The database lists no Vulkan API support for the GTX 670M. The R7 350 supports Vulkan 1.2.170 and has a recorded Geekbench Vulkan score of 7057.

Q: How do the memory bandwidth figures compare?

A: Both cards have identical memory bandwidth at 72.00 GB/s. The R7 350 uses 2 GB of GDDR5 on a 128-bit bus, while the GTX 670M uses 1536 MB of GDDR5 on a 192-bit bus. The R7 350's memory runs at 1125 MHz (4.5 Gbps effective), and the GTX 670M's runs at 750 MHz (3 Gbps effective).

Q: Which card has the higher transistor density?

A: The R7 350 has a transistor density of 12.2M per square millimeter on its 123 mm² die, which holds 1,500 million transistors. The GTX 670M has a density of 5.9M per square millimeter on its 332 mm² die, which holds 1,950 million transistors.

Q: What are the nearest rivals for each card?

A: The R7 350's closest rival is the Intel UHD Graphics 750 with an average score of 7441, just 0.2% behind. The GTX 670M's closest rival is the NVIDIA GeForce GT 555M with an average score of 6493, which the GTX 670M leads by 0.3%.

Q: Which card has the higher pixel rate?

A: The R7 350 has a pixel rate of 12.80 GPixel/s, while the GTX 670M has a pixel rate of 8.372 GPixel/s. The R7 350 is about 53% faster in this metric.

Specification Differences

The two cards differ in nearly every major specification category. The R7 350 uses a 28 nm process node, while the GTX 670M uses a 40 nm node. The R7 350 has a die size of 123 mm² with 1,500 million transistors, giving a density of 12.2M per square millimeter. The GTX 670M has a die size of 332 mm² with 1,950 million transistors, for a density of 5.9M per square millimeter. The R7 350 is a much denser, more efficient design.

Memory configurations differ substantially. The R7 350 has 2 GB of GDDR5 on a 128-bit bus with a memory clock of 1125 MHz (4.5 Gbps effective). The GTX 670M has 1536 MB of GDDR5 on a 192-bit bus with a memory clock of 750 MHz (3 Gbps effective). Both end up with identical bandwidth at 72.00 GB/s, but the R7 350 achieves it with a narrower bus and faster memory.

The compute resources are arranged differently. The R7 350 has 512 shading units, 32 TMUs, and 16 ROPs. The GTX 670M has 336 shading units, 56 TMUs, and 24 ROPs. The R7 350 has more shading units but fewer texture and raster units. The pixel rate favors the R7 350 at 12.80 GPixel/s versus 8.372 GPixel/s, while the texture rate favors the GTX 670M at 33.49 GTexel/s versus 25.60 GTexel/s. FP32 throughput is nearly identical: 819.2 GFLOPS for the R7 350 and 803.7 GFLOPS for the GTX 670M.

Power and physical specifications also differ. The R7 350 has a TDP of 55 W and a suggested PSU of 250 W. The GTX 670M has a TDP of 75 W and no suggested PSU listed. The R7 350 is a single-slot card with no power connectors, while the GTX 670M is an MXM module. The R7 350 uses a PCIe 3.0 x16 interface, while the GTX 670M uses MXM-B (3.0). The R7 350 has display outputs (1x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2), while the GTX 670M's outputs are listed as portable device dependent. The R7 350 is 168 mm (6.6 inches) long, while the GTX 670M has no listed dimensions.

API support differs in important ways. The R7 350 supports DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170. The GTX 670M supports DirectX 12 (11_0) and OpenGL 4.6, but has no Vulkan support listed.

Architecture Differences

The R7 350 is built on GCN 1.0 architecture using the Cape Verde chip, part of the Pirate Islands (R7 300) generation. It was released in 2016 and is a 28 nm TSMC part. Its predecessor is the Volcanic Islands generation and its successor is Arctic Islands. The GTX 670M is built on Fermi 2.0 architecture using the GF114 chip, part of the GeForce 600M generation. It was released in 2012 and is a 40 nm TSMC part. Its predecessor is the GeForce 500M generation and its successor is GeForce 700M.

The architectural gap is significant. GCN 1.0 is a compute-oriented design that scales well across a wide range of workloads, and the 28 nm process node allows a much smaller die with higher transistor density. Fermi 2.0 is an older design from a larger process node, which explains the 332 mm² die size and lower density. The R7 350's 1,500 million transistors on 123 mm² versus the GTX 670M's 1,950 million transistors on 332 mm² shows how much more efficiently the newer process packs transistors.

The R7 350 has 512 shading units, while the GTX 670M has 336. Despite having 176 fewer shading units, the GTX 670M achieves similar FP32 throughput due to its higher texture rate and different execution structure. The R7 350 has 32 TMUs and 16 ROPs, while the GTX 670M has 56 TMUs and 24 ROPs. The R7 350 compensates with a higher pixel rate, 12.80 GPixel/s versus 8.372 GPixel/s, but falls behind in texture rate, 25.60 GTexel/s versus 33.49 GTexel/s.

Neither card has ray tracing cores or tensor cores. Both are pure rasterization and compute designs. The R7 350's Vulkan support is a major architectural advantage, as it allows access to modern cross-platform graphics APIs that the GTX 670M cannot use. The DirectX 12 support also differs slightly, with the R7 350 supporting 11_1 features versus 11_0 for the GTX 670M.

The Verdict

The data points to a clear overall winner. The AMD Radeon R7 350 wins the only head-to-head benchmark by 19.6%, has a higher average benchmark score (7425 versus 6513), sits at a higher percentile (40 versus 38), and supports Vulkan while the GTX 670M does not. The R7 350 also has a higher pixel rate, more shading units, a smaller and more efficient die, and a lower TDP at 55 W versus 75 W.

The GTX 670M retains advantages in texture rate (33.49 GTexel/s versus 25.60 GTexel/s), more TMUs (56 versus 32), more ROPs (24 versus 16), and a wider memory bus (192-bit versus 128-bit). These are real strengths for texture-heavy workloads, but they do not translate into a benchmark win in the recorded data.

Choose the AMD Radeon R7 350 if you need modern API support, Vulkan compatibility, higher compute throughput, and better power efficiency. It is the faster card in the database's measurements and the more future-proof option for current software.

Choose the NVIDIA GeForce GTX 670M only if you specifically need its higher texture rate and larger number of ROPs, or if you are constrained to an MXM form factor. For everything else, the R7 350 is the stronger choice based on the recorded benchmark data.