AMD Radeon HD 8950M vs NVIDIA GeForce GTX 670 Comparison

The AMD Radeon HD 8950M and NVIDIA GeForce GTX 670 are both end-of-life graphics solutions from the 2012–2013 era, built on the same 28 nm TSMC process node but representing fundamentally different design philosophies. The data shows a clear performance hierarchy in the two shared benchmark tests, with the NVIDIA card winning both, though the margin varies significantly by workload. This analysis walks through the head-to-head numbers, what they mean relative to each card’s nearest rivals, and where each part remains relevant based strictly on the factual record.

Head-to-Head Benchmarks

The two GPUs were tested in two identical compute workloads: Geekbench OpenCL and Geekbench Vulkan. In both cases, the NVIDIA GeForce GTX 670 emerges as the winner, but the size of the victory is not uniform. In OpenCL, the GTX 670 scores 15,341 against the HD 8950M’s 13,772, a delta of -10.2% from the AMD card’s perspective. In Vulkan, the gap widens considerably: the GTX 670 scores 15,553 while the HD 8950M manages 12,979, producing a -16.5% delta. The pattern suggests that NVIDIA’s advantage grows in API workloads that lean more heavily on driver overhead and raw geometry throughput, areas where the GTX 670’s Kepler architecture with 1,344 shading units and 112 texture mapping units holds a structural edge.

Looking at the absolute scores, the GTX 670’s Vulkan result of 15,553 is its strongest showing, while its OpenCL score of 15,341 is nearly identical. The HD 8950M, by contrast, performs better in OpenCL (13,772) than in Vulkan (12,979), a difference of 793 points. This 5.8% internal drop for the AMD card between the two tests is notable because it indicates the GCN 2.0 architecture does not scale as well into Vulkan, whereas the Kepler-based GTX 670 remains flat across both APIs. For the HD 8950M, the OpenCL result is the better metric to cite when comparing its capability, as it represents the higher of its two scores.

When placed against their respective nearest rivals, the two cards occupy similar tiers but with different competitive positioning. The HD 8950M has an average benchmark score of 13,376, placing it 0.2% ahead of the AMD Radeon RX 5500M (13,356) and 0.2% ahead of the AMD FirePro M6100 (13,354), while trailing the AMD Radeon Pro 555 (13,407) by -0.2% and beating the AMD Radeon Pro 555X (13,321) by 0.4%. This cluster of rivals spans a narrow 86-point range, showing that the HD 8950M sits at the exact center of a tightly packed group of mid-range mobile GPUs. The GTX 670’s average score of 12,773 places it 0.4% behind the NVIDIA GeForce GTX 590 (12,830) and the AMD Radeon Pro 455 (12,831), while leading the AMD Radeon RX 7600M XT (12,710) by 0.5% and the AMD FirePro W5100 (12,847) by -0.6%. Here, the rival spread is 137 points, slightly wider, but the GTX 670 still holds a position within a few tenths of a percent of its nearest competitors.

The percentile rankings reinforce the head-to-head outcome. The HD 8950M sits at the 54th percentile of all GPUs, while the GTX 670 is at the 52nd percentile. Despite the GTX 670 winning both direct comparisons, its lower percentile reflects the fact that its average score (12,773) is dragged down by the absence of a Vulkan score in its average? No, that is not the case — the GTX 670 has both OpenCL and Vulkan scores, and its average is computed from those two, while the HD 8950M also has two scores. The percentile difference is simply a function of the broader GPU population distribution. The key takeaway is that in the two tests where both were measured, the GTX 670 is 10.2% to 16.5% faster, a meaningful margin that should be considered decisive for compute-oriented tasks.

The Verdict

The benchmark data is unambiguous: the NVIDIA GeForce GTX 670 is the faster GPU in both available tests. If a user is choosing between these two solely on performance, the GTX 670 wins on OpenCL by 10.2% and on Vulkan by 16.5%. The GTX 670 also offers a higher raw specification ceiling, with 1,344 shading units versus 768, 112 TMUs versus 48, and 32 ROPs versus 16. Its pixel rate of 27.44 GPixel/s and texture rate of 109.8 GTexel/s dwarf the HD 8950M’s 17.20 GPixel/s and 51.60 GTexel/s. In floating-point compute, the GTX 670 delivers 2.634 TFLOPS against the HD 8950M’s 1.651 TFLOPS, a 59.5% advantage.

However, the HD 8950M is not without justification. It consumes 100 W of power compared to the GTX 670’s 170 W, a 70 W difference that matters in portable or thermally constrained systems. The HD 8950M uses an MXM Module slot width, while the GTX 670 is a Dual-slot desktop card requiring 2x 6-pin power connectors and a 450 W suggested PSU. The HD 8950M also supports DirectX 12 (12_0), while the GTX 670 only reaches DirectX 12 (11_0). For users prioritizing API compliance on paper, the AMD card has the edge.

Who should pick which? From the data alone, the GTX 670 is the choice for anyone seeking maximum compute performance in OpenCL or Vulkan workloads, particularly where the 16.5% Vulkan lead can be exploited. The HD 8950M is the choice for those who need a lower-power, mobile-form-factor solution with higher DirectX feature level support, accepting a 10-16% performance deficit. The GTX 670’s launch MSRP was 399 USD, a figure that reflects its original desktop positioning, while the HD 8950M had no stated launch MSRP. For a workstation or desktop build where power and space are available, the GTX 670 is the data-backed winner. For a laptop or compact system, the HD 8950M’s 100 W TDP and MXM form factor make it the only realistic option of the two.

FAQ

Q: Which GPU wins in Geekbench OpenCL?

A: The NVIDIA GeForce GTX 670 wins with a score of 15,341 against the AMD Radeon HD 8950M’s 13,772, a delta of -10.2% for the AMD card.

Q: How much faster is the GTX 670 in Vulkan compared to the HD 8950M?

A: The GTX 670 scores 15,553 in Geekbench Vulkan, while the HD 8950M scores 12,979. The GTX 670 is 16.5% faster in this test.

Q: What is the average benchmark score for each GPU?

A: The AMD Radeon HD 8950M has an average benchmark score of 13,376. The NVIDIA GeForce GTX 670 has an average benchmark score of 12,773.

Q: How do the two cards compare to their nearest rivals?

A: The HD 8950M is 0.2% ahead of the AMD Radeon RX 5500M and the AMD FirePro M6100, and 0.4% ahead of the AMD Radeon Pro 555X, while trailing the AMD Radeon Pro 555 by -0.2%. The GTX 670 is 0.5% ahead of the AMD Radeon RX 7600M XT, but trails the GTX 590 and Radeon Pro 455 by -0.4% and the FirePro W5100 by -0.6%.

Q: What is the power consumption difference?

A: The HD 8950M has a TDP of 100 W, while the GTX 670 has a TDP of 170 W. The GTX 670 also requires a 450 W suggested PSU and two 6-pin power connectors.

Q: Which GPU supports a higher DirectX version?

A: The AMD Radeon HD 8950M supports DirectX 12 (12_0), while the NVIDIA GeForce GTX 670 supports DirectX 12 (11_0).

Specification Differences

The two GPUs differ across nearly every core specification. The HD 8950M uses the Saturn chip with 2,080 million transistors on a 160 mm² die, while the GTX 670 uses the GK104 chip with 3,540 million transistors on a 294 mm² die. Transistor density favors the AMD card slightly at 13.0M / mm² versus 12.0M / mm² for NVIDIA. The GTX 670 has a base clock of 915 MHz and a boost clock of 980 MHz, while the HD 8950M lists no base or boost clock. Memory clocks also differ: the HD 8950M runs at 1375 MHz (5.5 Gbps effective), and the GTX 670 runs at 1502 MHz (6 Gbps effective).

Memory configuration shows a split: both have 2 GB of GDDR5, but the bus widths differ substantially. The HD 8950M uses a 128-bit bus yielding 88.00 GB/s bandwidth, while the GTX 670 uses a 256-bit bus yielding 192.3 GB/s bandwidth — more than double. Shading units are 768 on the AMD card versus 1,344 on the NVIDIA card. TMUs are 48 versus 112, and ROPs are 16 versus 32. Pixel rate is 17.20 GPixel/s versus 27.44 GPixel/s, and texture rate is 51.60 GTexel/s versus 109.8 GTexel/s. FP32 compute is 1.651 TFLOPS versus 2.634 TFLOPS. The TDP is 100 W versus 170 W. The HD 8950M uses an MXM Module slot, while the GTX 670 is Dual-slot with 2x 6-pin power connectors and a 450 W suggested PSU. The GTX 670 measures 241 mm (9.5 inches) long, 111 mm (4.4 inches) tall, and 38 mm (1.5 inches) wide, while the HD 8950M has no listed dimensions.

Architecture Differences

The HD 8950M is built on AMD’s GCN 2.0 architecture, using the Saturn chip, and belongs to the Solar System generation (HD 8900M series). The GTX 670 is built on NVIDIA’s Kepler architecture, using the GK104 chip, and belongs to the GeForce 600 generation. Both use a 28 nm process node from TSMC, but the transistor counts diverge sharply: 2,080 million for AMD versus 3,540 million for NVIDIA. The die size also differs, 160 mm² versus 294 mm², leading to a slightly higher transistor density for the AMD chip.

Cache and feature differences are not listed in the data, but the API support reveals a key architectural split. The HD 8950M supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.2.170. The GTX 670 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The Vulkan version is marginally newer on the NVIDIA card. The GTX 670 has display outputs of 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2, while the HD 8950M’s outputs are listed as “Portable Device Dependent,” reflecting its mobile MXM form factor. Neither card has ray tracing or tensor cores. The HD 8950M’s predecessor is “London” and successor is “Gem System,” while the GTX 670’s predecessor is “GeForce 500” and successor is “GeForce 700.” Release dates are 2013-05-13 for the AMD card and 2012-05-09 for the NVIDIA card, making the GTX 670 the older part by roughly one year.

Where Each One Wins

The NVIDIA GeForce GTX 670 wins outright in both measured benchmarks, so the performance-based use case is clear. Its 16.5% Vulkan advantage and 10.2% OpenCL advantage make it the superior choice for any compute workload that relies on those APIs. The GTX 670 also has a massive memory bandwidth lead at 192.3 GB/s versus 88.00 GB/s, which suggests it will handle higher-resolution textures and memory-intensive tasks more effectively. Its higher pixel rate (27.44 GPixel/s) and texture rate (109.8 GTexel/s) indicate stronger fill-rate performance for traditional rasterization.

The AMD Radeon HD 8950M wins in efficiency and form factor. Its 100 W TDP is 70 W lower than the GTX 670’s 170 W, making it viable for systems with limited power budgets. The MXM Module slot width means it is designed for laptops and small-form-factor systems, whereas the GTX 670 is a Dual-slot desktop card. The HD 8950M also supports DirectX 12 (12_0) versus the GTX 670’s DirectX 12 (11_0), providing a higher feature level for applications that require it. In terms of nearest rivals, the HD 8950M’s average score places it above the RX 5500M and FirePro M6100, while the GTX 670 sits below the GTX 590 and Radeon Pro 455 — but those rival comparisons are within fractions of a percent and do not change the direct head-to-head result.

For a desktop user with space and power available, the GTX 670 is the data-backed winner. For a mobile user or a system constrained to 100 W and an MXM slot, the HD 8950M is the only feasible option, and its performance deficit of 10-16% is the trade-off for that portability. The GTX 670’s launch MSRP was 399 USD, but no pricing should be inferred for the HD 8950M. The verdict is straightforward: performance favors NVIDIA, portability and efficiency favor AMD.