AMD Radeon R7 M350 vs NVIDIA GeForce GTX 680M Comparison

The NVIDIA GeForce GTX 680M and the AMD Radeon R7 M350 are both end-of-life mobile GPUs built on the same 28 nm TSMC process, but they were aimed at very different segments of the laptop market. The recorded data shows the GTX 680M leading in every directly comparable benchmark, with an average benchmark score of 7023 against the R7 M350's 6327. That overall gap of roughly 11 percent understates the story in the one test where the two cards can be compared head to head, and the hardware spec sheet explains most of it.

Head-to-Head Benchmarks

The only direct comparison in the database is Geekbench OpenCL, and it is a clear win for the GTX 680M: 9230 points versus 6991 points, a 32 percent margin. OpenCL rewards raw compute throughput, and the GTX 680M's 2.038 TFLOPS of FP32 compute is more than two and a half times the R7 M350's 779.5 GFLOPS. The benchmark result tracks that hardware gap almost exactly as you would expect.

Context from the rest of the database reinforces the picture. The GTX 680M also posted 4815 points in Geekbench Metal, a test where the R7 M350 has no recorded score. The R7 M350's other recorded result, 5662 points in Geekbench Vulkan, sits below the GTX 680M's OpenCL figure and well below its average.

Against the wider GPU field, the two land surprisingly close in percentile terms: the GTX 680M sits in the 39th percentile of all GPUs in the database, the R7 M350 in the 36th. But the percentile similarity is partly an artifact of the scoring pool, not evidence of equivalent hardware. Look at each card's nearest rivals and the difference in character becomes obvious. The GTX 680M trades blows with the NVIDIA T600 (average score 7035, just 0.2 percent ahead of it), the AMD Radeon R5 M240 (6975, 0.7 percent behind), and the NVIDIA GeForce GTX 675M (6946, 1.1 percent behind). These are tightly clustered, performance-class mobile and entry workstation parts.

The R7 M350's rival list looks odd by comparison. It matches the AMD Radeon Pro WX 4100 exactly (6330, 0 percent delta), sits 0.7 percent ahead of the NVIDIA Quadro K620 (6282), and is recorded 0.9 percent ahead of the NVIDIA GeForce RTX 5070 Ti SUPER and the RTX 4070 Ti SUPER AD102 (both at 6270). Modern high-end desktop cards appearing in an entry-level mobile GPU's rival list is a quirk of how the averages are distributed across the database, and it should not be read as the R7 M350 competing with those cards in any practical sense. In the one controlled, same-test comparison available, the GTX 680M wins 1-0 in the head-to-head tally.

Architecture Differences

Despite the shared 28 nm TSMC node, the two GPUs are architecturally a generation and a market tier apart. The GTX 680M is the GK104 chip from NVIDIA's Kepler architecture, part of the GeForce 600M generation and released on 2012-06-03. The R7 M350 is the Meso chip from AMD's GCN 3.0 architecture, part of the Gem System (R7 M300) generation and released on 2015-05-04, nearly three years later.

The silicon budgets tell the real story. The GK104 die measures 294 mm² and packs 3,540 million transistors, a density of 12.0M per mm². The Meso die is 125 mm² with 1,550 million transistors at 12.4M per mm². Densities are nearly identical, which makes sense given the same node and foundry, but the GTX 680M simply has more than twice the silicon: 2.3 times the die area and 2.3 times the transistor count.

That scale shows up everywhere in the functional units. The GTX 680M carries 1344 shading units, 112 texture mapping units, and 32 render output units, against 384 shading units, 24 TMUs, and 8 ROPs on the R7 M350. That works out to 3.5 times the shader count, 4.7 times the TMUs, and 4 times the ROPs. The resulting throughput rates follow the same pattern: 21.22 GPixel/s versus 8.120 GPixel/s in pixel fill rate, and 84.90 GTexel/s versus 24.36 GTexel/s in texture fill rate.

Clocks run higher on the AMD part, 1000 MHz base and 1015 MHz boost versus 719 MHz base and 758 MHz boost on the NVIDIA card, but the clock advantage is nowhere near enough to offset the shader deficit. Memory is the other great divide. The GTX 680M uses 4 GB of GDDR5 on a 256 bit bus at 900 MHz (3.6 Gbps effective), delivering 115.2 GB/s of bandwidth. The R7 M350 uses 4 GB of DDR3 on a narrow 64 bit bus at 1000 MHz (2 Gbps effective), delivering just 16.00 GB/s. That is a 7.2-to-1 bandwidth advantage for the GTX 680M, and bandwidth is often the hard ceiling for entry-level parts.

Feature support has some nuances. The R7 M350 reports the newer DirectX 12 feature level 12_0, while the GTX 680M reports DirectX 12 at feature level 11_0. Vulkan support is close: 1.2.175 on the NVIDIA card versus 1.2.170 on the AMD card, and both report OpenGL 4.6. Neither card has RT cores or tensor cores. The R7 M350 also lists FP16 throughput at 779.5 GFLOPS, a 1:1 ratio with its FP32 rate, while no FP16 figure is recorded for the GTX 680M.

Platform integration differs too. The GTX 680M is an MXM module using the MXM-B (3.0) interface, the form factor of a big, replaceable mobile graphics card with a 100 W TDP. The R7 M350 connects over PCIe 3.0 x8 and has no recorded TDP, consistent with a small chip soldered into thin or mainstream laptops. Neither card needs power connectors.

Where Each One Wins

The data shows exactly one winner. The GTX 680M takes the only shared benchmark, Geekbench OpenCL, by 32 percent, and its single-test scores bracket the R7 M350's on both ends: 9230 in OpenCL above, 4815 in Metal below, against the AMD card's 6991 OpenCL and 5662 Vulkan.

Where the R7 M350 has genuine advantages, they are all in platform features rather than performance. Its DirectX 12 feature level 12_0 support is the more modern API tier, its clocks are higher, and its small 125 mm² die and PCIe-based integration suit thin machines where a 100 W MXM module is physically impossible. Its FP16 support at full FP32 rate is also recorded, something the database does not list for the Kepler part.

But in any workload measured by the database, compute, rendering, or fill-rate-bound gaming, the GTX 680M's 4 times the ROPs, 4.7 times the TMUs, 3.5 times the shaders, and 7.2 times the memory bandwidth produce results that consistently favor it. The 21.22 GPixel/s versus 8.120 GPixel/s pixel rate alone predicts a large gap in any resolution-limited scenario.

FAQ

Q: Which GPU is faster overall?

A: The NVIDIA GeForce GTX 680M. It leads the only shared benchmark, Geekbench OpenCL, 9230 to 6991, a 32 percent margin, and its average benchmark score of 7023 beats the R7 M350's 6327.

Q: Do they use the same manufacturing process?

A: Yes. Both are built on a 28 nm process at TSMC, with transistor densities of 12.0M per mm² for the GTX 680M and 12.4M per mm² for the R7 M350.

Q: How big is the memory bandwidth difference?

A: Very large. The GTX 680M provides 115.2 GB/s from 4 GB of GDDR5 on a 256 bit bus, while the R7 M350 provides 16.00 GB/s from 4 GB of DDR3 on a 64 bit bus.

Q: Which card has better API support?

A: The R7 M350 reports DirectX 12 at feature level 12_0, ahead of the GTX 680M's feature level 11_0. Vulkan is essentially a wash at 1.2.175 versus 1.2.170, and both support OpenGL 4.6.

Q: Are these cards still in production?

A: No. Both are listed as end-of-life. The GTX 680M was released on 2012-06-03 and the R7 M350 on 2015-05-04.

Q: How do they compare to other GPUs in the database?

A: The GTX 680M ranks in the 39th percentile and sits within about 2 percent of cards like the NVIDIA T600 (7035) and the GTX 675M (6946). The R7 M350 ranks in the 36th percentile and matches the Radeon Pro WX 4100 at 6330.

The Verdict

Strictly from the recorded data, the GTX 680M is the pick for anyone who cares about GPU performance. It wins the only head-to-head test by 32 percent, it delivers 2.6 times the FP32 compute, 4 times the ROPs, and 7.2 times the memory bandwidth, and its fill rates are more than double or triple the AMD part's across the board. The cost of that performance is the 100 W TDP and the MXM-B (3.0) module format, which confines it to larger, performance-oriented notebooks.

The R7 M350 makes sense only where the GTX 680M cannot physically exist: it is a 125 mm² die with no listed TDP, integrated over PCIe 3.0 x8, with the newer DirectX 12 feature level 12_0. Its 36th percentile ranking and 6327 average score put it slightly below the GTX 680M in the database, and its 6991 OpenCL score is not close to the NVIDIA card's 9230. If the decision is purely about measured performance, the data points one way.

Specification Differences

• Manufacturer: NVIDIA versus AMD
• Chip: GK104 versus Meso
• Architecture: Kepler versus GCN 3.0
• Generation: GeForce 600M versus Gem System (R7 M300)
• Transistors: 3,540 million versus 1,550 million
• Die size: 294 mm² versus 125 mm²
• Transistor density: 12.0M / mm² versus 12.4M / mm²
• Base clock: 719 MHz versus 1000 MHz
• Boost clock: 758 MHz versus 1015 MHz
• Memory type: GDDR5 versus DDR3
• Memory speed: 3.6 Gbps effective versus 2 Gbps effective
• Bus width: 256 bit versus 64 bit
• Bandwidth: 115.2 GB/s versus 16.00 GB/s
• Shading units: 1344 versus 384
• TMUs: 112 versus 24
• ROPs: 32 versus 8
• Pixel rate: 21.22 GPixel/s versus 8.120 GPixel/s
• Texture rate: 84.90 GTexel/s versus 24.36 GTexel/s
• FP32: 2.038 TFLOPS versus 779.5 GFLOPS
• FP16: not recorded versus 779.5 GFLOPS (1:1)
• TDP: 100 W versus not recorded
• Bus interface: MXM-B (3.0) versus PCIe 3.0 x8
• DirectX: 12 (11_0) versus 12 (12_0)
• Vulkan: 1.2.175 versus 1.2.170
• Release date: 2012-06-03 versus 2015-05-04
• Predecessor: GeForce 500M versus Solar System
• Successor: GeForce 700M versus Polaris Mobile