AMD Radeon 760M vs NVIDIA GeForce GTX 680M Comparison

Head-to-Head Benchmarks

The single recorded head-to-head comparison between these two GPUs is the Geekbench OpenCL test, and the result is decisive. The AMD Radeon 760M scores 20,255 points, while the NVIDIA GeForce GTX 680M achieves 9,230 points. The delta percentage of -54.4% indicates that the GTX 680M trails by that margin, meaning the Radeon 760M is roughly 2.2 times faster in this compute-oriented workload. This is not a marginal difference; it is a generational leap in raw throughput.

Looking at the broader benchmark landscape, the GTX 680M's average score across all recorded tests is 7,023, placing it in the 39th percentile of all GPUs in the database. Its nearest rivals include the NVIDIA T600 (average score 7,035, delta -0.2%), the AMD Radeon R5 M240 (average score 6,975, delta 0.7%), and the NVIDIA GeForce GTX 675M (average score 6,946, delta 1.1%). The GTX 680M sits essentially in a dead heat with these three, with performance differences under 1.1%. It also trails the NVIDIA GeForce GTX 970 (average score 7,157) by 1.9%, a modest gap given that the GTX 970 is a far more recent desktop part.

The Radeon 760M, by contrast, has an average benchmark score of 6,019, which places it in the 35th percentile. Its nearest rivals are the AMD Radeon RX 6400 (average score 6,001, delta 0.3%), the NVIDIA GeForce GTX 770M (average score 6,000, delta 0.3%), and the NVIDIA RTX PRO 6000 Blackwell Server (average score 5,996, delta 0.4%). It also edges out the NVIDIA Quadro P2000 (average score 6,049) by -0.5%. The 760M's average score is actually lower than the 680M's average, yet in the OpenCL head-to-head, it wins by a massive 54.4%. This contradiction suggests that the two chips excel in very different test conditions.

The GTX 680M also has a Geekbench Metal score of 4,815, a test the Radeon 760M does not appear in. The Radeon 760M, however, has a much richer benchmark portfolio, including a 3DMark Steel Nomad DX12 score of 400, a Geekbench Vulkan score of 30,336, and multiple Passmark scores (DX9: 65, DX10: 19, DX11: 52, DX12: 25, G2D: 890, G3D: 5,310, GPU Compute: 2,840). These additional data points suggest the 760M is tested across a wider range of APIs and workloads, while the 680M's recorded metrics are limited to two Geekbench runs.

Where Each One Wins

The data indicates a clear split: the Radeon 760M wins in modern, compute-heavy, and API-diverse scenarios, while the GTX 680M appears more competitive in legacy or specific OpenCL contexts, though it loses that head-to-head too. The 760M's OpenCL score of 20,255 is more than double the 680M's 9,230, showing dominance in general-purpose GPU compute. Its Vulkan score of 30,336 further reinforces this, as Vulkan is a low-overhead API that rewards modern architecture efficiency. The 760M also supports DirectX 12 Ultimate (12_2), the latest feature level, whereas the 680M is limited to DirectX 12 (11_0), meaning the 760M can handle newer graphics features like ray tracing, as evidenced by its 8 ray tracing cores.

The GTX 680M's wins, if any, are less apparent from the recorded data. Its average benchmark score of 7,023 is higher than the 760M's 6,019, which suggests that in some aggregate workload, the older card performs better. This could be due to the specific tests included in the average: the 680M has only two Geekbench scores, both of which are substantial (4,815 Metal and 9,230 OpenCL). The 760M's average is pulled down by its Passmark scores, which are low (e.g., DX9: 65, DX10: 19, DX11: 52, DX12: 25) despite high Geekbench and Vulkan numbers. This implies the 760M is optimized for modern APIs but struggles in older DirectX 9/10/11 paths, where the 680M, being a 2012-era Kepler chip, likely has more mature drivers or hardware compatibility.

For gaming, the 760M's Vulkan and DX12 scores suggest it excels in titles that use those APIs. Its 3DMark Steel Nomad DX12 score of 400, while low in absolute terms, is a modern benchmark that the 680M cannot even run. For older games that rely on DirectX 9 or 10, the 680M may hold its own, but the data does not include a direct comparison in those tests. The 760M's Passmark G3D score of 5,310 is close to its average, indicating that in DirectX 11 (a legacy but still common API), it performs at a level similar to its overall average, while the 680M's average is driven by Geekbench alone.

Architecture Differences

The fundamental difference lies in the process node and architecture generation. The GTX 680M uses the GK104 chip, built on a 28 nm process at TSMC, with a transistor count of 3,540 million and a die size of 294 mm². This yields a transistor density of 12.0M per mm². The Radeon 760M uses the Phoenix chip, fabricated on a 4 nm process, also at TSMC, with 25,390 million transistors on a 178 mm² die, giving a density of 142.6M per mm². The density difference is staggering: the 760M packs over 11 times more transistors per area, enabling far more complex logic in a smaller space.

The 680M is based on Kepler architecture, a design from 2012, while the 760M uses RDNA 3.0, a modern architecture from 2024. Kepler was NVIDIA's third unified architecture, focusing on efficiency and scalability. RDNA 3.0 is AMD's latest gaming architecture, designed for high clock speeds and ray tracing. The 760M has 512 shading units, 32 TMUs, and 16 ROPs, while the 680M has 1,344 shading units, 112 TMUs, and 32 ROPs. Despite having fewer shading units, the 760M achieves higher pixel and texture rates: 41.58 GPixel/s and 83.17 GTexel/s, respectively, versus the 680M's 21.22 GPixel/s and 84.90 GTexel/s. The pixel rate is nearly double, thanks to the 760M's much higher boost clock.

Clock speeds illustrate the architectural leap: the 680M runs at a base of 719 MHz and a boost of 758 MHz, while the 760M runs at 800 MHz base and 2,599 MHz boost. The 760M's boost clock is over 3.4 times higher, leading to an FP32 throughput of 5.323 TFLOPS versus the 680M's 2.038 TFLOPS. The 760M also supports FP16 at 5.323 TFLOPS (1:1), while the 680M has no recorded FP16 capability. Memory configuration differs fundamentally: the 680M has 4 GB of dedicated GDDR5 on a 256-bit bus with 115.2 GB/s bandwidth, while the 760M uses system shared memory, with its size, type, bus width, and bandwidth all marked as "System Dependent" or "System Shared." This means the 760M's memory performance depends entirely on the host system's RAM, a common trade-off for integrated GPUs.

The 760M includes 8 ray tracing cores, a feature entirely absent from the 680M. Its API support extends to DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, while the 680M supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The 680M uses an MXM-B (3.0) bus interface and is a discrete MXM module, while the 760M is an IGP (integrated graphics processor) using PCIe 4.0 x8. Power consumption differs drastically: the 680M has a TDP of 100 W, while the 760M is rated at 15 W, a 85% reduction in power draw for significantly higher compute performance.

FAQ

Q: Why does the AMD Radeon 760M have a lower average benchmark score than the NVIDIA GeForce GTX 680M, yet win the head-to-head OpenCL test?

A: The 680M's average of 7,023 is based on only two Geekbench tests, both of which are relatively strong. The 760M's average of 6,019 includes ten tests, several of which (Passmark DX9, DX10, DX11, DX12) score very low (19 to 65), pulling its average down despite a dominant OpenCL score of 20,255 and a Vulkan score of 30,336.

Q: Can the GTX 680M run modern ray tracing games?

A: No. The GTX 680M has no ray tracing cores and only supports DirectX 12 (11_0), which lacks the ray tracing features found in DirectX 12 Ultimate. The Radeon 760M, with 8 ray tracing cores and DirectX 12 Ultimate (12_2), is the only one of the two capable of hardware-accelerated ray tracing.

Q: Which GPU is more power-efficient based on the data?

A: The Radeon 760M is dramatically more efficient. Its TDP is 15 W compared to the GTX 680M's 100 W, yet it delivers 5.323 TFLOPS of FP32 performance versus 2.038 TFLOPS. This means the 760M achieves over 2.6 times the compute per watt.

Q: How does memory bandwidth compare between the two?

A: The GTX 680M has a fixed 115.2 GB/s bandwidth from 4 GB of GDDR5 on a 256-bit bus. The Radeon 760M uses system shared memory, and its bandwidth is listed as "System Dependent," meaning it varies based on the host laptop or desktop's RAM configuration.

Q: What is the transistor density difference and why does it matter?

A: The 760M has a transistor density of 142.6M per mm² on a 4 nm node, versus the 680M's 12.0M per mm² on 28 nm. This 11-fold density increase allows the 760M to fit far more logic (including ray tracing cores and higher clock speeds) into a smaller die (178 mm² vs 294 mm²).

Q: Which GPU has better API support for future software?

A: The Radeon 760M, with DirectX 12 Ultimate (12_2), Vulkan 1.4, and FP16 support, is future-proof. The GTX 680M tops out at DirectX 12 (11_0) and Vulkan 1.2.175, missing key modern features.

Specification Differences

The two GPUs differ in nearly every measurable specification. The process node moves from 28 nm (680M) to 4 nm (760M), and the die size shrinks from 294 mm² to 178 mm² despite the transistor count rising from 3,540 million to 25,390 million. Shading units drop from 1,344 to 512, TMUs from 112 to 32, and ROPs from 32 to 16, yet the 760M's pixel rate more than doubles (41.58 GPixel/s vs 21.22 GPixel/s) and texture rate is nearly identical (83.17 vs 84.90 GTexel/s). Clock speeds: base goes from 719 MHz to 800 MHz, boost from 758 MHz to 2,599 MHz. FP32 doubles-plus from 2.038 TFLOPS to 5.323 TFLOPS, with FP16 only on the 760M (5.323 TFLOPS 1:1). Memory shifts from 4 GB GDDR5 on a 256-bit bus with 115.2 GB/s to system shared memory. TDP drops from 100 W to 15 W. The 680M is an MXM module with MXM-B (3.0) interface; the 760M is an IGP with PCIe 4.0 x8. The 760M adds 8 ray tracing cores. API support: DirectX 12 (11_0) vs 12 Ultimate (12_2), Vulkan 1.2.175 vs 1.4. Production status: end-of-life (2012 release) vs active (2024 release).

The Verdict

The data points to a clear generational shift. The AMD Radeon 760M is the superior compute and modern-graphics processor, winning the only direct head-to-head test by 54.4% and offering more than double the FP32 throughput, ray tracing support, and dramatically better power efficiency (15 W vs 100 W). Its higher boost clock (2,599 MHz vs 758 MHz) and newer architecture (RDNA 3.0 vs Kepler) make it the obvious choice for anyone running Vulkan, DirectX 12 Ultimate, or compute workloads.

However, the NVIDIA GeForce GTX 680M should not be dismissed. Its average benchmark score is higher (7,023 vs 6,019), and it has dedicated GDDR5 memory with fixed bandwidth, which can be an advantage in scenarios where system RAM is slow or shared with the CPU. The 680M also has more shading units (1,344 vs 512), which may help in certain legacy DirectX 9/10/11 titles where raw shader count matters more than clock speed. The recorded data shows the 680M is competitive with mid-range GPUs from its era, but it is end-of-life and limited to older APIs.

For a new purchase or upgrade, the Radeon 760M is the rational pick based on performance, features, and efficiency. For a collector or someone running older software that favors high shader counts and dedicated VRAM, the GTX 680M remains a functional, if obsolete, option. The database records one decisive win for AMD, and the architectural evidence supports that outcome across most modern workloads.