AMD Radeon 740M vs NVIDIA GeForce GTX 670 Comparison

The AMD Radeon 740M and NVIDIA GeForce GTX 670 represent two very different eras of GPU design, yet their aggregate benchmark scores land them within a razor-thin margin of each other. The data shows a near-total stalemate in overall performance, with the AMD part averaging 12,870 points and the NVIDIA part averaging 12,773 points — a difference of less than one percent. This parity is remarkable given that the GTX 670 is a legacy discrete card from a previous decade, while the 740M is a modern integrated processor. The head-to-head results, however, reveal a more nuanced story about workload-specific strengths, architectural efficiency, and the shifting priorities of GPU development.

Head-to-Head Benchmarks

The most decisive data point in this comparison comes from the Geekbench OpenCL test, where the NVIDIA GeForce GTX 670 delivers a commanding win. The GTX 670 scores 15,341, while the AMD Radeon 740M manages only 10,172 — a delta of -33.7% for the AMD part. This is a substantial margin that highlights the GTX 670's raw compute throughput in a legacy API context. The NVIDIA card's 1,344 shading units and 112 texture mapping units provide a significant parallel execution advantage that the integrated AMD solution cannot match in this specific workload. For compute-heavy tasks that rely on OpenCL, the data suggests the older discrete card remains the superior choice.

The Vulkan benchmark tells an almost opposite story, though with a far smaller margin. The AMD Radeon 740M edges out the GTX 670 by a score of 15,568 to 15,553, a delta of just 0.1%. While this is technically a win for the AMD part, the practical difference is negligible — within the noise of any benchmark run. What is significant is that the 740M, despite having only 256 shading units and 16 TMUs, manages to match a card with over five times the shader count in a modern graphics API. This suggests that architectural efficiency and driver optimization for Vulkan have largely closed the gap that raw hardware specifications once dictated. The 740M's RDNA 3.0 architecture appears to extract far more performance per shading unit than the older Kepler design.

Looking at the broader rival context, both GPUs sit in a tight cluster of similarly performing parts. The 740M's nearest rivals include the AMD Radeon RX 580, which scores 12,928 and is 0.4% ahead, and the NVIDIA GeForce GTX 590 at 12,830, which is 0.3% behind. The GTX 670's rivals include the same GTX 590 (0.4% behind) and the AMD Radeon RX 7600M XT at 12,710, which is 0.5% ahead. This clustering reinforces that the two cards occupy the same performance tier on average, even though their individual benchmark profiles diverge sharply. The average benchmark score of 12,870 for the 740M versus 12,773 for the GTX 670 places both in the 52nd and 53rd percentiles of all GPUs, respectively — effectively identical standings.

The Verdict

From the data alone, the choice between these two GPUs depends almost entirely on the software environment. For OpenCL-based workloads, the NVIDIA GeForce GTX 670 is the clear winner, offering a 33.7% performance advantage that would be noticeable in any compute task leveraging that API. The GTX 670 also holds a significant bandwidth advantage with its dedicated 192.3 GB/s GDDR5 memory, which is critical for texture-heavy applications and large dataset operations. However, the GTX 670 is end-of-life, consumes 170 W versus the 740M's 45 W, and requires a dual-slot form factor with two 6-pin power connectors.

The AMD Radeon 740M, conversely, demonstrates that modern integrated graphics have reached parity with older mid-range discrete cards in modern APIs like Vulkan. Its 0.1% lead in that benchmark, while statistically insignificant, signals that the architectural advancements in RDNA 3.0 — including a 4 nm process node and 4 ray tracing cores — have effectively neutralized the GTX 670's hardware advantage in contemporary workloads. The 740M's system-shared memory and motherboard-dependent display outputs make it a flexible, low-power solution for compact systems, though its performance will vary with system memory configuration.

Strictly from the data, users prioritizing raw OpenCL compute should choose the GTX 670, while those targeting modern Vulkan applications or seeking an efficient integrated solution should favor the 740M. The GTX 670's launch MSRP was 399 USD, though it has long since been discontinued. The 740M, being active and current, represents forward-looking compatibility with DirectX 12 Ultimate and Vulkan 1.4, whereas the GTX 670 is capped at DirectX 12 (11_0) and Vulkan 1.2.175.

Architecture Differences

The architectural divide between these two GPUs is generational. The AMD Radeon 740M is built on TSMC's 4 nm process node, packing 20,900 million transistors into a 137 mm² die — a transistor density of 152.6 million per square millimeter. The NVIDIA GeForce GTX 670, by contrast, uses a 28 nm process from the same foundry, with 3,540 million transistors on a 294 mm² die, yielding a density of just 12.0 million per square millimeter. This represents a 12.7x difference in transistor density, which explains how the 740M achieves competitive performance with far fewer resources.

The core configurations reflect this efficiency gap. The 740M utilizes 256 shading units, 16 TMUs, and 8 ROPs, alongside 4 dedicated ray tracing cores. The GTX 670 employs 1,344 shading units, 112 TMUs, and 32 ROPs — dramatically higher counts across the board. Yet the FP32 compute output tells a different story: the 740M delivers 2.867 TFLOPS, slightly exceeding the GTX 670's 2.634 TFLOPS. This means each 740M shading unit produces over 5x the floating-point work of each GTX 670 unit, evidence of the IPC improvements in RDNA 3.0 over Kepler.

Memory architecture also diverges fundamentally. The 740M uses system-shared memory with bandwidth that is system-dependent, while the GTX 670 has a dedicated 2 GB GDDR5 frame buffer on a 256-bit bus delivering 192.3 GB/s. The GTX 670's memory clock runs at 1502 MHz (6 Gbps effective), whereas the 740M's memory clock is listed as system shared. The 740M's pixel rate of 22.40 GPixel/s and texture rate of 44.80 GTexel/s are lower than the GTX 670's 27.44 GPixel/s and 109.8 GTexel/s, but the 740M's higher clock speeds — 2800 MHz boost versus 980 MHz boost — partially compensate in compute-bound scenarios.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The AMD Radeon 740M has an average benchmark score of 12,870, while the NVIDIA GeForce GTX 670 scores 12,773 — a difference of 97 points, or roughly 0.8%.

Q: How do they compare in OpenCL performance?

A: The NVIDIA GeForce GTX 670 wins decisively in OpenCL, scoring 15,341 versus the AMD Radeon 740M's 10,172, representing a 33.7% advantage for the NVIDIA card.

Q: What about Vulkan performance?

A: The AMD Radeon 740M narrowly wins the Vulkan test with a score of 15,568, compared to the GTX 670's 15,553 — a margin of just 0.1%, making the two effectively tied in this API.

Q: What are the power consumption differences?

A: The AMD Radeon 740M has a TDP of 45 W and requires no power connectors, while the NVIDIA GeForce GTX 670 has a TDP of 170 W and requires two 6-pin power connectors, with a suggested PSU of 450 W.

Q: Do these GPUs support ray tracing?

A: The AMD Radeon 740M includes 4 ray tracing cores, while the NVIDIA GeForce GTX 670 has no ray tracing cores listed in its specifications.

Q: What is the production status of each GPU?

A: The AMD Radeon 740M is listed as active and was released on 2024-01-30, while the NVIDIA GeForce GTX 670 is end-of-life, having been released on 2012-05-09.

Where Each One Wins

The NVIDIA GeForce GTX 670 wins in scenarios that demand raw compute throughput in legacy APIs. Its 33.7% OpenCL advantage makes it the better choice for users running OpenCL-based compute workloads, scientific simulations, or older professional applications that have not been optimized for modern APIs. The GTX 670's dedicated 192.3 GB/s memory bandwidth also gives it a clear edge in texture-heavy and bandwidth-sensitive tasks, where the 740M's system-shared memory could become a bottleneck. Additionally, the GTX 670's higher pixel rate of 27.44 GPixel/s and texture rate of 109.8 GTexel/s suggest superior performance in fill-rate-limited scenarios, such as high-resolution rendering with heavy overdraw.

The AMD Radeon 740M wins in modern API environments and efficiency-critical applications. Its Vulkan performance, while only 0.1% ahead, demonstrates that its architecture is better suited to contemporary graphics standards, including DirectX 12 Ultimate and Vulkan 1.4. The 740M's 4 ray tracing cores provide hardware acceleration that the GTX 670 completely lacks, making it the only option here for ray-traced effects in supported games. The 740M's 45 W TDP, IGP form factor, and lack of power connectors make it ideal for thin-and-light laptops, mini PCs, and systems where space and thermal headroom are at a premium. Its 4 nm process node also positions it as a forward-looking solution with active production status and ongoing driver support.

The 740M's higher FP32 throughput of 2.867 TFLOPS versus the GTX 670's 2.634 TFLOPS suggests it may also edge ahead in compute tasks that are well-optimized for RDNA architecture, even if the OpenCL test does not reflect this. The 740M's transistor density advantage — 152.6M per mm² versus 12.0M per mm² — implies far better power efficiency per unit of performance, which is critical for sustained workloads in thermally constrained environments.

Specification Differences

The following specifications differ between the AMD Radeon 740M and NVIDIA GeForce GTX 670:

• Process Node: 4 nm (TSMC) versus 28 nm (TSMC)
• Transistors: 20,900 million versus 3,540 million
• Die Size: 137 mm² versus 294 mm²
• Transistor Density: 152.6M / mm² versus 12.0M / mm²
• Base Clock: 800 MHz versus 915 MHz
• Boost Clock: 2800 MHz versus 980 MHz
• Memory Size: System Shared versus 2 GB
• Memory Type: System Shared versus GDDR5
• Memory Bus Width: System Shared versus 256 bit
• Memory Bandwidth: System Dependent versus 192.3 GB/s
• Memory Clock: System Shared versus 1502 MHz (6 Gbps effective)
• Shading Units: 256 versus 1,344
• TMUs: 16 versus 112
• ROPs: 8 versus 32
• Ray Tracing Cores: 4 versus None
• Pixel Rate: 22.40 GPixel/s versus 27.44 GPixel/s
• Texture Rate: 44.80 GTexel/s versus 109.8 GTexel/s
• FP32: 2.867 TFLOPS versus 2.634 TFLOPS
• FP16: 2.867 TFLOPS (1:1) versus None
• TDP: 45 W versus 170 W
• Slot Width: IGP versus Dual-slot
• Power Connectors: None versus 2x 6-pin
• Suggested PSU: None versus 450 W
• Bus Interface: PCIe 4.0 x8 versus PCIe 3.0 x16
• Display Outputs: Motherboard Dependent versus 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2
• DirectX Support: 12 Ultimate (12_2) versus 12 (11_0)
• Vulkan Support: 1.4 versus 1.2.175
• Dimensions: Not specified versus 241 mm length, 111 mm height, 38 mm width
• Production Status: Active versus End-of-life
• Release Date: 2024-01-30 versus 2012-05-09
• Predecessor: Navi II IGP versus GeForce 500
• Successor: Navi III IGP versus GeForce 700
• Launch MSRP: None versus 399 USD