AMD Radeon HD 6950 vs NVIDIA GeForce GTX 670M Comparison

The NVIDIA GeForce GTX 670M and the AMD Radeon HD 6950 represent two distinct philosophies in GPU design from the early 2010s, separated by over a year of release timing and aimed at different market segments. The GTX 670M is a mobile-first MXM module, while the HD 6950 is a full-length desktop card. Benchmark data from the Geekbench OpenCL suite shows the GTX 670M scoring 6513, which places it in the 38th percentile of all GPUs, while the HD 6950 scores 6210, sitting in the 36th percentile. Despite the HD 6950’s larger die and higher power envelope, the GTX 670M emerges as the winner in the only head-to-head benchmark available, posting a 4.9% higher score. This outcome is surprising given the substantial specification differences, and the analysis below explores where each card wins, answers common questions about their performance, and breaks down the architectural and specification gaps.

Where Each One Wins

The data shows a clear, albeit narrow, victory for the NVIDIA GeForce GTX 670M in the sole benchmark recorded. In the Geekbench OpenCL test, the GTX 670M scores 6513, while the AMD Radeon HD 6950 scores 6210. This 4.9% delta is the only direct comparison available, and it indicates that the GTX 670M holds a performance edge in compute workloads that utilize OpenCL. The GTX 670M’s average benchmark score of 6513 places it comfortably ahead of its nearest rivals, including the NVIDIA GeForce GT 555M (6493, a 0.3% gap) and the AMD Radeon Vega 10 Mobile (6476, a 0.6% gap). It also narrowly outperforms the NVIDIA Quadro M5000M (6481, a 0.5% gap) and the Intel UHD Graphics P750 (6554, a -0.6% gap, meaning the Intel chip is slightly behind).

On the other hand, the AMD Radeon HD 6950’s score of 6210 puts it in a different competitive tier. Its nearest rivals include the AMD FirePro W600 (6223, a -0.2% delta, meaning the HD 6950 trails by 0.2%) and the NVIDIA Quadro K620 (6282, a -1.1% delta). Notably, the HD 6950 is also listed against the NVIDIA GeForce RTX 4070 Ti SUPER AD102 and the NVIDIA GeForce RTX 5070 Ti SUPER, both scoring 6270, with the HD 6950 trailing by 0.9% in both cases. This suggests that while the HD 6950 is an older card, its OpenCL performance is not entirely obsolete, though it lacks the top-tier placement of the GTX 670M.

In terms of use-case wins, the GTX 670M wins the compute benchmark outright. However, the HD 6950’s specification sheet suggests strengths in raw throughput metrics. The HD 6950 has a pixel rate of 25.60 GPixel/s and a texture rate of 70.40 GTexel/s, which are significantly higher than the GTX 670M’s 8.372 GPixel/s and 33.49 GTexel/s. These figures imply that the HD 6950 would likely win in rasterization-heavy scenarios, such as traditional gaming at high resolutions, if benchmark data were available. The GTX 670M, with its lower power envelope of 75 W versus the HD 6950’s 200 W, is clearly designed for mobile efficiency, making it the winner in power-constrained environments. The HD 6950, requiring a 550 W suggested PSU and dual 6-pin connectors, is a desktop-only part.

FAQ

Q: Which GPU has a higher OpenCL benchmark score?

A: The NVIDIA GeForce GTX 670M scores 6513, which is 4.9% higher than the AMD Radeon HD 6950’s 6210 in the Geekbench OpenCL test.

Q: How does the GTX 670M compare to its closest rivals?

A: The GTX 670M’s average score of 6513 places it 0.3% ahead of the NVIDIA GeForce GT 555M (6493) and 0.6% ahead of the AMD Radeon Vega 10 Mobile (6476). It trails the Intel UHD Graphics P750 (6554) by 0.6%.

Q: What is the memory bandwidth difference between the two cards?

A: The AMD Radeon HD 6950 has a memory bandwidth of 160.0 GB/s, which is more than double the GTX 670M’s 72.00 GB/s. The HD 6950 also features a 256-bit bus width compared to the GTX 670M’s 192-bit bus.

Q: Which card has a higher transistor count and what does that indicate?

A: The AMD Radeon HD 6950 has 2,640 million transistors, while the NVIDIA GeForce GTX 670M has 1,950 million. The HD 6950’s higher count reflects a larger die size of 389 mm² versus 332 mm², indicating a more complex architecture.

Q: What is the launch MSRP of the AMD Radeon HD 6950?

A: The launch MSRP for the AMD Radeon HD 6950 is 299 USD.

Q: Does the GTX 670M support a higher DirectX version than the HD 6950?

A: Yes, the GTX 670M supports DirectX 12 (11_0), while the HD 6950 supports DirectX 11.2 (11_0). The GTX 670M also supports OpenGL 4.6, whereas the HD 6950 supports OpenGL 4.4.

Head-to-Head Benchmarks

The only direct benchmark comparison in the data is the Geekbench OpenCL test, where the NVIDIA GeForce GTX 670M scores 6513 against the AMD Radeon HD 6950’s 6210. This yields a 4.9% delta in favor of the GTX 670M. While this is a single data point, it is significant because it shows a mobile GPU beating a desktop card with a much larger die and higher power draw. The GTX 670M achieves this with 336 shading units, 56 texture mapping units, and 24 render output units. In contrast, the HD 6950 has 1408 shading units, 88 TMUs, and 32 ROPs. The HD 6950’s raw compute capacity is substantially higher, with an FP32 throughput of 2.253 TFLOPS versus the GTX 670M’s 803.7 GFLOPS. Yet, the OpenCL result suggests that the GTX 670M’s architecture is more efficient in this specific workload.

The delta of 4.9% is modest but consistent with the GTX 670M’s percentile ranking. The GTX 670M sits at the 38th percentile of all GPUs, while the HD 6950 sits at the 36th. This two-percentile gap aligns with the benchmark delta, indicating that the GTX 670M is not only ahead in this test but also in the broader distribution of GPU performance. The HD 6950’s nearest rival, the AMD FirePro W600, scores 6223, which is only 0.2% higher than the HD 6950, showing that the HD 6950 is in a tight cluster of similarly performing cards. The GTX 670M, by contrast, is at the top of its immediate cluster, with its closest competitor being the Intel UHD Graphics P750 at 6554, which is 0.6% higher.

The head-to-head result also highlights the efficiency of the GTX 670M. It achieves its score with a 75 W TDP, whereas the HD 6950 has a 200 W TDP. This means the GTX 670M delivers competitive OpenCL performance at a fraction of the power consumption. The HD 6950’s higher bandwidth (160.0 GB/s versus 72.00 GB/s) and faster memory clock (5 Gbps effective versus 3 Gbps effective) do not translate into a win in this test, suggesting that memory bandwidth is not the limiting factor in OpenCL compute for these cards. The GTX 670M’s advantage may stem from its architecture’s scheduling or driver optimizations, but the data does not specify the cause.

Specification Differences

The two cards differ across nearly every major specification category. The GTX 670M is built on a 40 nm process at TSMC, with 1,950 million transistors on a 332 mm² die, resulting in a transistor density of 5.9M / mm². The HD 6950 also uses a 40 nm process at TSMC, but packs 2,640 million transistors into a 389 mm² die, achieving a density of 6.8M / mm². The HD 6950’s die is 57 mm² larger and contains 690 million more transistors.

Memory configurations diverge sharply. The GTX 670M has 1536 MB of GDDR5 on a 192-bit bus, yielding a bandwidth of 72.00 GB/s at a memory clock of 750 MHz (3 Gbps effective). The HD 6950 offers 2 GB of GDDR5 on a 256-bit bus, producing 160.0 GB/s of bandwidth at a memory clock of 1250 MHz (5 Gbps effective). The HD 6950 has over twice the bandwidth and 33% more memory capacity.

Compute resources also favor the HD 6950 on paper. The HD 6950 has 1408 shading units, 88 TMUs, and 32 ROPs, while the GTX 670M has 336 shading units, 56 TMUs, and 24 ROPs. Pixel rate and texture rate follow suit: the HD 6950 hits 25.60 GPixel/s and 70.40 GTexel/s, versus the GTX 670M’s 8.372 GPixel/s and 33.49 GTexel/s. FP32 performance is 2.253 TFLOPS for the HD 6950 and 803.7 GFLOPS for the GTX 670M.

Power and physical characteristics differ dramatically. The GTX 670M has a 75 W TDP, uses an MXM-B (3.0) bus interface, and is an MXM Module with no power connectors, making it suitable for laptops. The HD 6950 has a 200 W TDP, uses a PCIe 2.0 x16 interface, is a dual-slot card, and requires 2x 6-pin power connectors and a 550 W suggested PSU. The HD 6950 measures 286 mm in length, 126 mm in height, and 42 mm in width, while the GTX 670M’s dimensions are listed as portable-device dependent. Display outputs also differ: the HD 6950 offers 2x DVI, 1x HDMI 1.4a, and 2x mini-DisplayPort 1.2, whereas the GTX 670M’s outputs are portable device dependent. Release dates are separated by over a year, with the HD 6950 launching on 2010-12-13 and the GTX 670M on 2012-03-21.

Architecture Differences

The architectural divide is fundamental. The GTX 670M is based on NVIDIA’s Fermi 2.0 architecture, using the GF114 chip, and belongs to the GeForce 600M generation. The HD 6950 uses AMD’s TeraScale 3 architecture, built on the Cayman chip, and is part of the Northern Islands (HD 6900) generation. These are entirely different design philosophies: Fermi 2.0 focuses on compute and geometry processing, while TeraScale 3 uses a VLIW4 design optimized for shader throughput.

The process node is identical (40 nm TSMC), but the transistor counts reflect different scaling strategies. The HD 6950’s 2,640 million transistors on a 389 mm² die give it a density of 6.8M / mm², while the GTX 670M’s 1,950 million transistors on a 332 mm² die yield a lower density of 5.9M / mm². This indicates that AMD packed more logic into a larger area, likely for the 1408 shading units, while NVIDIA’s Fermi 2.0 used fewer, more complex cores.

API support differs as well. The GTX 670M supports DirectX 12 (11_0) and OpenGL 4.6, while the HD 6950 supports DirectX 11.2 (11_0) and OpenGL 4.4. Neither card lists Vulkan support. The GTX 670M’s higher OpenGL version suggests better support for modern graphics APIs, even though both are end-of-life products. The HD 6950’s predecessor is Evergreen and its successor is Southern Islands, while the GTX 670M’s predecessor is GeForce 500M and its successor is GeForce 700M, placing them in different product lineage trees.

The TDP difference of 125 W (75 W versus 200 W) is a direct result of the architectural choices. The GTX 670M’s Fermi 2.0 design achieves its benchmark win while consuming far less power, making it a superior choice for mobile computing. The HD 6950’s TeraScale 3 architecture is built for maximum throughput, but it requires a desktop chassis and substantial cooling. The slot width confirms this: the GTX 670M is an MXM Module, while the HD 6950 is dual-slot. In essence, the GTX 670M wins on efficiency and compute benchmark performance, while the HD 6950 dominates on raw specification numbers, though those numbers do not translate to a win in the recorded test.